Thank you thank you.

But I'll start by saying a little
bit about the supply chain and

logistics Institute and

that may help explain that sort of
sequence of additional interests but

before I do that I should tell my my
students that I recognize that you don't

really don't have to come to this you get
good sort of material in class to do it.

So you can sneak out quietly if you want.

OK so

if you'll forgive me for talking a little
bit about myself just as a way to segue

into the supply chain and logistics
Institute which is within the school

of industrial systems engineering Anybody
here from industrial systems engineering.

OK Goodness.

So you guys already know what most of
this my background is actually math and

the wonderful thing about industrial
engineering in the supply chain and

logistics Institute is
it's a wonderful place for

people who are interested in math but
perhaps not in pill to work on pure math.

If you like to work on applied math and
it's sort of different from

the Applied math that you
might think of with physics or

things like that that instead it's
really an Applied Math devoted to

human systems the kinds of
systems that human beings build.

And the techniques are different in
the issues are different and it's

an altogether satisfying field because you
can affect the way that people live and

your laboratory instead of being
you know having a testbench

in and pushing tiny things around and
looking through electron microscope.

What you do is you get out and
go out into the community and

look at the kinds of systems
that need to be improved.

My colleagues but
to Kilian the supply chain and

logistics Institute work on just
incredible variety of problems.

I've got one group of colleagues that
is working on this air taxi service.

Yes that you may have read about the some
of them are just beginning to start

the idea being that in order to avoid
congestion at Hartsfield Jackson you

should be able to call this company and
book a seat on a small jet.

That will.

They will incorporate you into their
schedule quote you a price on it and

you may not fly directly there but when
you add up the overall travel time it may

be rather faster than going through
a major carrier at a major airport and so

this is basically a big math problem and
it's similar to a lot of the kinds of math

problems that we work at which are work
on which are problems of coordination.

So.

We have a lot of folks in our group that
have to background like myself in math or

in computer science.

And we end up using those
tools on human systems.

So that's what we do.

I'll tell you very quickly to so
you can get a little bit more

sense of the range of stuff I'll tell you
one thing about things that I know about

more personally than some of
the things that I've worked on.

One thing that I did was build
a routing system for Meals on Wheels

program that senior citizen services runs
of Liber hot meals to people who are too

old or sick to to cook for themselves they
make about three hundred fifty deliveries

that day which is a pretty good
sized commercial routing system but

they couldn't afford
anything to do the routing.

And so what we did was use the concept
from mathematics of a space filling curve

to build a way of generating routes

that was indistinguishable from simply
operating a rule of X. card file.

And so there was no technology
that was visible the the total

cost of the system we built for Meals
on Wheels was about forty dollars for

two Rolodex card files and
it produces pretty darn good routes.

The Exactly our same idea
was later adopted by T.R.W.

systems a contractor for this.

Strategic Defense Initiative
this idea that sort of

has continued on since Reagan and.

Which I personally think is idiotic.

But the.

U.S. has been working on for
a long time about controlling a space

based laser weapon and the idea is that
it would destroy incoming missiles and

then it would travel to the next
missile as it really aimed and

then destroy that missile and so on and
so that's just like a routing problem for

Meals on Wheels you make a delivery which
in this case is blow up the missile and

then you travel to the next delivery point
which in this case is travel to the next

game to the next missile and in both cases
you want to minimize the travel distance.

And so that turns out to be
a very interesting problem and.

You can see that it has sort of wildly
different applications a question.

Yes please.

But it's traveling many times
slower than the speed of light.

So it's that because of the huge
difference in the speed of

light the love the laser beam so
that in effect you can think of a bunch

of points on a two dimensional focal plane
because also the laser weight weapon

does not appreciably degrade
over those distances.

So one of the challenges here is that
you don't want to send up a doomsday

weapon floating in space controlled
by a genetic algorithm or.

Artificial intelligence because
you don't know what it's going

to do You can't be sure and
so in this case what

the Department of Defense wants is
something which is an understandable.

Analyzable they want something that.

Theorems can be proved about it that
guarantee how it's going to behave and

so it all of these things make this idea
based on space filling curves particularly

appropriate in the nice thing about them
a space filling curves are such bizarre

mathematical creatures that it's a very
satisfying from in a static point if you.

You might guess wonder about
the name space filling curve.

It's essentially.

Way a pattern you can
trace out with a pencil.

So that you can use a one
dimensional line to fill completely

a three dimensional space.

It's an example of a fractal OK So
that is sort of and

shows you the kind of range
of topics from mathematics.

To.

You know how do we care for our elderly.

What are the resources that they have
available what are the solutions that they

need and sort of bringing all
those kinds of things to bear

along with just some sort
of engineering judgment and

common sense to produce a usable solution
that's appropriate for the customer.

So that's the.

An example of the sort of thing.

Some of those other topics like in
politics and biology and so on.

You'll see some of those touch here.

In particular the biology and so one of
the great things I think about being at

a university is the encouragement to do
the think really work on anything and

this is one of the particular things I
value about this school of industrial

systems engineering is that I'll bet
more than most places you could come and

do graduate study in almost anything you
can imagine because I know I have been

supported by my department chair and
encouraged to write papers in biology or

to write papers on the theory of
voting and political science and

I've done that and had a wonderful time
doing it and feel that it's really.

Enrich my life and so
it's the kind of thing that that sort of

flexibility is sort of built into
the Department in Georgia Tech.

It's very good about
encouraging that sort of thing.

OK so what I want to talk
to you today about is self

organizing logistics this is something
I've spent a lot of time thinking about

over the last ten years or so.

And let me begin by I said
supply chain and logistics

Institute for those of you who don't know
this is what a supply chain is and it's

sort of the path by which mainly affected
products gets to us the consumers.

So almost.

Everything that you have with
you right now your clothes

your pencil your note pad your shoes
your bookcase has all traveled

down a path like this in order to get to
you in the interesting thing about it.

These supply chains says that
there are international and

so the manufacturing may very
well be in East Asia and

the coast of east coast of China
in particular and other segments

of the supply chain may bring the product
closer and closer to you and kill you.

Finally purchase it in a retail
store someplace in Georgia.

And so this product has been moving
towards you and what we do with the supply

chain and logistics Institute is trying
to understand the dynamics the and

the economics of flow through a system
like this because everyone has got to work

in the real world and there are you
know labor issues and labor rates and

equipment issues and the cost of
capital and the cost of land and

how much money you have tied up in
inventory and all these business issues

that determine how we can move
product is inefficient in a fish.

As if I got my maxim of the minimisation
mix up there as efficiently as possible

towards you the consumer.

Now of course the masters of
this these days is Wal-Mart and

they've basically built their dominance
as the world's largest retailer based

on focusing on this.

They don't make pretty displays or
worry about things like that so

much as they worry about how can I move
the product cheaply here because when you

look at the total cost of the product
it's frequently the case that the cost of

moving it dwarfs the cost of manufacture
and so there's a huge amount

of efficiencies to be gained by moving
the product quickly to the customer.

And so that's what we do we think a lot
about the moving of the product here and

that's what I'm going
to talk about today and

I'm going to chose I've chosen a very
small part of the whole of this.

I'm going to look just at a regional D.C.

which I mean distribution center or
and sort.

Older take terminology warehouse.

There are many people in the supply chain
and logistics that no longer like the term

warehouse because it sounds too static
like stuff is just being stored there.

They like the word distribution
center better because

it emphasizes I've got all
the stuff in the building.

I've got customers waiting for product.

How can I move this product to
the customers as quickly as possible.

And indeed that is the focus of
most modern distribution centers.

If you look at them in the United States.

You'll see ones that typically look like
this though this might be an Amazon or

a Cosco or a Wal-Mart.

And they may have ten or
twenty or thirty of these

things scattered around North America and
you can see they're quite large.

There may be five hundred to fifteen
hundred workers in a place like

this huge labor costs.

But they're devoted solely to once you
make that click on Amazon dot com These

people spring into action to get that
product to you as quickly as possible.

Now one of the interesting
things the challenging things

about this part of the supply chain in
North America is that we are driven.

So strongly by the holiday selling
season which is beginning to about now.

It will just intensify
all the way to Christmas.

But you know the Port of Los Angeles
has been congested for weeks

already with containers coming through
bringing manufactured goods from China.

Which are being distributed
throughout the stores and

they'll be stuffed into
the distribution centers here and.

Ready to come to you.

During these times.

There's a huge increase in temporary
labor for example Amazon which normally

has about five hundred order pickers
may go to one thousand five hundred or

pickers that me triple their work force
here to the staff distribution center

to get the product to you.

So they've got this mix of
trained workers and then.

Temporary were.

They don't really know what they're
doing and will only be working for

a few months here during
the busiest season of the year.

Let's go inside this distribution
center and see what's going on.

Hoops.

OK I've gotten ahead of myself.

Let's don't go inside this
distribution center just yet.

Instead what I want to talk
about is the challenge

of coordinating thousands of workers.

You can imagine all of this product is.

In this warehouse and
we've got one thousands of workers and

I'm getting customer orders
pouring in over the Internet for

example or from stores from consumers.

From the retail stores that I support.

How can I coordinate my my work
force to get this product out.

So that's the challenge I want
to particularly mention and

I want to tell you about
a marvelous system that does this.

It's a logistics system that has
tens of thousands of workers.

The amazing thing is that operates at near
optimality it's almost impossible to make

an improvement in yet nobody is in
charge there is no I.T. department.

There's no management science.

There's no consultants no engineers.

It sounds too good to be true but
we've all seen these things.

And they are the social insects.

Here is the honeypot and room assistants
Milledge are native to the high

plateau of Colorado and distinctive for
the fact that it runs a warehouse.

The ants that you see hanging from the
ceiling with a grossly distended abdomens

are ants that the vote themselves
entirely to the storage of nectar

food for their colleagues.

Other ants and shown in this
picture are out gathering nectar.

OK Now think about the problem
faced by this species of ant

if they have too many ants
out gathering nectar and

insufficiently many devoted
to the storage of that.

There's a imbalance and
the survival of the colonies at risk

because they're not able to
store all the food they collect

in the same way if they have too many ants
devoted to the storage and not enough

gathering again there's an imbalance which
imperils the existence of the colony.

So this is just the most dramatic of the
things that the ants have got to get just

right there are many others as well you
could there are other tasks going on here.

The ant in the center is the queen and
the large one on the bottom here.

The queen ant but
the queen ant is not despite the name is

not really in charge in any meaningful
sense of the word the queen and

the simply an egg laying machine.

The other and
there's ants here in gauged in Queen

care they groom the queen constantly and
feeder.

There are ants engaged in brood care there
are caring for the larvae in the P.P.

in the distance there ants
in gauged in nest defense in

mention foraging already nest building
maintenance many different tasks and

in fact generations of Ph D.
students in biology have studied

exactly what tasks and stew over the
course of their lifetime lifetime in for

an amp like this that the runs anywhere
from range six to eight to ten or

twelve different tasks specific tasks
that ants will work on in their lifetime.

But the odd thing about it is that.

To a first approximation these
ants are remarkably accurate

in allocating just the right
number of workers to each task.

Just the right number of
workers now believe it or not.

Biologists have a way of putting these
tiny bar codes on the backs of ants.

And what they will do is
take a colony of ants

like this put them in the refrigerator so

they become sluggish and then put these
bar codes on the backs of all the ants.

And then set up a video camera up above
them and sort of track them as they go and

then they can read these bar codes and
figure out which each individual is doing

and how many answer doing this and
how many answer doing that.

And lo and behold when they do the
arithmetic here about balancing the flows

of you know incoming food and storage
they get it right but nobody's in charge.

How do they do that.

OK this is the phenomenon
of self organisation.

It's been sort of popular in the press
here for the last couple of years.

So there's a lot of buzz words
that you may hear in ization is

probably the most common swarm
intelligence is another popular one.

The interesting thing about this
is that he had arises from below.

So it's not imposed from above
by somebody in charge but

arises from the participants themselves.

No ant is very intelligent but

when they work together things that
are produced that appear intelligent.

This is a.

Cross-section of a nest of
a coast Rican paper wasp.

And it's safe to say that when the wasp
is laying down the substrate for

this nest the wasp doesn't
know what it's building.

It doesn't have a blueprint in mind.

It's not working with an architect.

It's simply following some very simple and
primitive impulse but

what emerges is something that has got
a larger pattern that when we see it.

We think of as intelligent
This satisfies the purpose.

It's tempting to say for
which it was designed.

Although a biologist would say well
wait it wasn't designed at all.

At least not by design or
that it has emerged from

the joint operations of
many many participants.

So.

The other important thing to say about
this is the agents themselves are unaware

of what it is they're producing.

Now you can begin to see already
that boy if we could get.

Something like this working in human
logistic systems how practical that

would be because it would put
people like myself out of business.

If we weren't needed
as consultants to fine

tune operations if you could just simply
set it up and it would work on its own.

Well that's one of the very nice
things about self organizing behavior

because if it's self organizing that
means it's easy to implement because

you just set it up and
walk away and it adjusts itself.

And furthermore it continually adjusts
itself so you don't need to manage it.

I'll mention that this
is the the weaver ant.

And you can find these all over Singapore
where the School of Industrial and

systems engineering has a.

In the supply chain logistics Institute
has a sister organization there called

the logistics Institute Asia Pacific and
so if you go there to study

you might see the weaver and of Singapore
which is distinctive for this reason.

You can see these ants are working
together they form and chains each one

uses its mandibles to grasp the waste
of an ant in front of it so

they form these chains of ants and
they pull leaves together.

And once the leaves are together and
touching.

Another ant will take a larva.

One of their own larva.

Which looks sort of like a caterpillar and

has the capability of spinning
silk just like a caterpillar.

And it will touch the larva
with its antennae.

Causing the larva to spin silk and

then this ant will just take the larva and
stitch the two leaves together

by basically sewing the two leaves
together with the sticky silk.

And so by attaching these leaves
together they build a nest.

They live all above ground which
they can do in a tropical climate.

So they have their own set of self
organizing challenges in and solutions.

Absolutely fascinating to watch you can
see how come I have written papers in

biology once you start thinking about
the social insects as an engineer.

To find out what their survive just
amazing stories to be understood and

and then try to make sense of from
an engineering point of view and

then possibly bring back into
the world of human activities.

OK.

So let's go inside our warehouse
now that we see sort of what we'd

like to do with what we'd
like to do is have a.

Coordinate these order pictures.

These are Order pickers who are picking
recorded music in this case.

And this is about half of
them along this one line this

is an aisle of the warehouse so that comes
out towards me and so we can see six or

eight order pictures here it comes out
further towards me with another six or

eight order pickers so
there were about twelve or

sixteen order pickers here and
each one has got a tray which is.

Product being picked for a particular
store in this case the customer is

a Wal-Mart store in fact this distribution
center chips only to Wal-Mart.

And so this tray is product recorded
music that's going to Wal-Mart and

so you can think of this is
being like an assembly line.

But the product being assembled
here is not a Model T. Ford.

Each one identical one after the other but
instead is a customer order of

items requested by a particular Wal-Mart
store and each one wants different things.

Because each one faces slightly
different customer demand.

And so while there might be some
statistical similarity between what

is needed in one wall much store in
another in general it's going to vary so

you can think of this is being like
an assembly line in which one product is

a Dell computer and the next product
is a Macintosh computer and the next

product is a pair of running shoes and
the next product is an iphone or.

So they're all sort of.

You know they can vary a lot so that
not identical from one to the other and

their problem is how do we.

Make this efficient.

How do we get maximum use of
the productive capability of the workers

in a way that respects their.

Integrity.

Is fair to them but also moves that
product as quickly as possible

because we don't want to have
products sitting still and

the time products sitting
still to sit still.

It costs us money.

You may have noticed during
two thousand and six the price

of Dell Computers dropped seventeen
percent Dell laptops up seventeen percent.

What that means is their product
was basically evaporating money.

At that rate and so.

Understandably they don't want
that product ever to sit still.

They wanted always to be moving and
moving quickly to the customer and

it's the same here even with recorded
music with any set of goods.

It's evaporating money and so

you want to accelerate its
movement to the to the customer.

OK So this is our challenge.

So I've taken at of the whole supply
chain I've taken this one tiny piece

these twelve or
sixteen workers right here.

How can I make the product go faster here.

The reason for focusing on this is
that this is one of those key spots

in the supply chain the product previously
was sitting still sitting on the shelf.

A customer has called and
ordered the product and

now it's up to us to get it out and
get it to the customers quickly

as possible because we can't build
a customer until we get it on the truck.

OK so.

This determines all the customer
experience downstream in the supply chain

how quickly can we get this out.

So let's choose this tiny little
piece of the supply chain and

try to make it efficient.

I'm going to tell you first of all
about what the people do now mostly.

And here's two standard solutions.

One is cold walk and pick.

And in that one.

What would happen is each
individual worker here.

Here's my tiny warehouse.

This is just like the aisle
that you see up above.

And here I've got a stack of customer
orders and in walk in pick system.

I would look at this customer order and
it says I should start at location three

one two three and pick the quantity
required which is one here.

So I'd pick.

I need something to put my product
in I'd pick one item to go in here.

And then next I want to go to location for
him I'd pick one item at location five and

six and then nine under lock and
pick system.

Each worker would carry this
public all the way to the end so

they'd assemble this order.

In assembly line terms
they've built the product.

It's now ready to be shipped.

Fortunately our shipping
department is right here.

So we can put it here and shipping will
pack it up load it on the truck or

put their wood on the back of a U.P.S.
fan in a way it goes.

And we've gotten the product out.

OK So that's the challenge with walk in.

Pick every workers on their own.

So that's one standard method

another standard method is called
Zone picking or pick and pass.

I find it easier to think of as being
like a traditional assembly line.

If you think of an assembly
line such as the classic.

Model T. assembly line of Henry Ford.

What he did was.

It was too much to have a worker assemble
a Model T. entirely by themselves.

You'd have to train a worker to be
able to build a car from scratch.

OK and what his genius was was to sort
of think of this idea of breaking down

the work into smaller and smaller pieces
and getting specialization of labor.

And so zone picking is an attempt
to sort of bring that kind of

logic into the supply chain.

And so what we might do is have
one worker who's responsible.

Just for these five locations.

So the.

Build only part of the product.

Wherever this customer order touches
one of these five locations.

So in this case I would have
picked from three four and five.

And at this point I would stop.

I'm not allowed to go into
the other workers zone.

And what I would do is leave the product
here for the other worker to continue.

I pass it off like that then I go back and
start a new order.

OK So that is the behavior of others
own picking which is the assembly line.

These are the two standard ways and
what I thought I'd do is.

Let's run away or house here.

And so will actually pick
some custom orders here.

See how these two work and
then I'll suggest another method

based on the amps based on.

The idea of self organization and try it.

We'll compare it.

Now when I do this in class and
you'll see this in warehousing class.

Typically what I do will spend
a whole ninety minutes or

sometimes even longer on this and will not
only try these methods but I'll challenge

the class just think of any method you
can and will try everything and anything.

And so it's an interesting experiment.

I'm going to leave.

Three volunteers to be order pickers.

Anybody want to ruin your afternoon
with a little please come forward.

We have one don't make me call on you
because you know I'll call in my students.

And the volunteers.

Barbara.

Yes.

OK.

Please.

Let me you'll need a pencil.

If you can find one easily let me show
you what a customer order looks like.

On to.

You just so you can see.

Now since I'm in charge of this
warehouse I'm going to ask you.

To follow a certain protocol.

Which is what it says line here
that simply means what intrigues

on the shopping list and so they just
read the entry in the shopping list and

you'll visit these locations in sequence
you can see they're already ordered for

you in the sequence in which
you'll encounter them here.

And you'll pick.

The items and once you've picked it.

Ask you to put a little check there now.

Don't wait till you get to
the very end to put your checks.

As you pick it stop it.

Put a little check.

OK We're going to run this.

We'll start by doing walk and pick.

In under walk and

pick that means each one of you is going
to be sort of an independent agent.

Now we need a product.

To be picked.

And so.

This is it.

Paperclips.

Thank

you.
You know.

Barbara Dimon disturbing these among

the five here in this from
Goldman the five here.

OK now.

There's one other thing I need to do in
order to make this a realistic simulation

remember when I described for Amazon.

During the busy season.

They'll go from five hundred full time
workers to adding one thousand temporary.

OK.

But you can imagine if you're in charge
of a workforce like that it's going to be

hugely disparate you'll find some people
are highly motivated and learn quickly.

You'll find some people
are borderline employable.

And you've got to integrate
these people somehow

into an operation that is
focused on speed and efficiency.

So that's one of the challenges and
so therefore I have to.

In order to make this more realistic.

I don't want to I mean your
typical worker in a warehouse

is not your typical well educated
person on a college campus.

And so what I need to do is slow down
some of my workers a little bit and

so Barbara I'm going to ask you to
pick out the product only with a pair

of pliers here so it's going to be
a little bit harder than doing it by hand.

OK So this is going to
unfairly slow you down but

it's necessary to make
the simulation realistic and.

To me I know your name.

J J OK And remind me the.

With N L.

D D P OK OK Jay show or Buddy what you're
picking with so that they won't blame you.

Afterwards.
OK OK So that's going to make it unless

a difficult for him but this is within
the realm of real worker speeds.

I mean one of the things and finding
a solution is it's got to work with real

workers right so one of the things
we had to do was find out.

Well what do workers look like in terms of
their velocities how fast they work and

we found in taking surveys in distribution

centers a factor of three can separate
the very fast this from the very slowest

workers in fact more than three but it's
very typical to find a factor of three.

So these will be the our workers
everybody else is in management.

And I need somebody in management with
a watch that perhaps can't give us

a count here.

What we'd like to do is run this
warehouse for five minutes.

And so we'll need from you is
a five four three two one GO.

And then you'll time it are were
ordered pickers will jump

to work immediately remember you're
being paid based on your productivity.

So you're going to want to.

Finish as quickly as possible and

put your orders here for
shipping put paper here and order here.

OK everybody got it.

We're ready when you're ready.

Order takers and go to work.

OK Now while they're working.

I'm going to be talking because I'm going
to talk a little bit about what's good and

what's bad about each
form of organization.

Five minutes five minutes
right we'll get we'll try.

Three different methods.

And we'll give each one five minutes so
this won't take us.

What's this.

OK.

Yes you've got to put the tick
marks right afterwards.

Thank you.
Management.

OK we'll have to keep an eye.

Sometimes it takes a few moments for
them to learn the process as we go.

Now you can see some predictable things
are fastest workers moving out ahead.

Right now.

That's exactly what we would expect.

Good good good barber good.

Good diva.

OK But let's get some let's go
as fast as you can remember.

Are you going fast.

You need to have a sense of urgency.

Yes I am the replenish or

I'm going to be recycled
unbeknownst to them.

I'm not actually going
to ship this by U.P.S.

I'm going to take it around and
put it back on the shelf.

OK.

Good put it down and
run back and get some more.

Urgency urgency.

OK good.

OK now.

Like any sense.

This is an engineering problem how to
make the system work more efficiently.

There's no final right and

wrong answer is just that every
solution has some things that are.

Good about it and
some that are not so good and so

one of the roles of
the engineer is to figure out.

Well what's good about this and
what's not so good about it.

So I know when it makes sense to use it.

One thing that's good about it
is if I've got a lot of turnover

in my workers particularly among
the temporaries you know who

knows who's going to be here tomorrow
if I got a thousand workers.

You know probably nine hundred are going
to show up in you know the other hundred

maybe home sick who are their
child has to go to the dentist or

they've lost their enthusiasm for
working in the warehouse so

they found a better job or something but
that's always going to be going on and

it'll be a different hundred home each
day and this method works pretty well

with those a lot of turnover because
the workers don't have to coordinate with

each other right you just come you're
working almost as an independent agent.

You arrive at work you.

Start picking the orders and
at the end of the day you go home.

So there's a lot of flexibility if
somebody were to take a bathroom break.

It doesn't interfere with anybody else
they can continue working exactly as they

were previously.

So that's a good thing it's flexible.

It doesn't require much management
thought you don't have to set up anything

for the workers.

So it's a.

Typically the first kind of order picking
scheme that you'll see a warehouse a dot.

OK we can see Deepa has run ahead again.

Again because she's faster now everybody
else she's sort of past the other workers

for the second time.

There are on the other hand
some possible disadvantages.

If we want to be as
productive as she can be

we need to make sure
that she's not impeded.

And we need to provide additional room for
her to be able to pass the other workers.

OK In a real warehouse space is expensive.

You would generally if you have
extra space you want to use it to

hold more product or.

Are bringing in different products but

here we would need extra
space just to allow passing.

So that the product can can move and not
be held up we do the last thing the world

we want is for deep have to
stand there with nothing to do.

Because she's our best worker.

Another possible problem is that.

We lose the learning effect in a sense
where asking each worker to build

a model T. entirely by themselves.

Now this doesn't seem
like a serious impediment

with such a tiny simulated warehouse.

But in a real warehouse.

There's likely to be.

Let's stop right there.

Thank you.

That's why workers want to stand back here
in the break room for one minute and we'll

give you a chance to catch your breath
before we go on in a real warehouse the.

There may be thousands of locations and

so what happens when you get is basically
when you get a temporary worker in and

they're picking from
thousands of locations.

They've got the shopping list and
they do a lot of this.

It's like me trying to find
a particular product in the Kroger

you know I'll get closer
to the right aisle and

then I do a lot of sort
of local searching.

Well that.

We're paying for that if you're management

youth see that's unproductive
time it's not value added time.

It's not moving the product and so what
we'd like to do is get rid of that and

one way to get rid of it is help the
workers learn where the product this hard

to do when they have to learn the entire
thousands of locations so that is why many

places as soon as they get some support
they move to this zone picking where

we try to take and create an environment
in which the workers can learn better.

So we give the workers a subset.

Let me point out one other disadvantage
here and that is I'm going to read to you.

The sequence in which
the orders were completed.

Now I have numbered these orders and
that stack their lists the.

Orders in sequence one
two three four five six.

So here the sequence in
which they're finished.

It's.

Number one number

two and number three is that right.

OK So

mostly in sequence.

One two three The Because diva
started last in the past.

Everybody.

Did you start first or last.

You started first.

I see what happened.

You just had to.

OK you got started with the very
last of the orders that I should

have taken him away from you.

And so
they're not really in sequence here.

Let me move up to here's a better
example order to order for

order three Order six with
the last several to be completed.

So what happened was I started
with the orders in sequence but

they've came out in some other sequence.

Now why is that a problem.

Well I might be loading these onto
a truck in reverse sequence of delivery.

So I might have had my my
software system might have

planned the truck route out and release
the orders in just the right sequence and

now they come off in a different
sequence because I don't know

who's the sequence in which
the orders are going to be completed.

Because it depends on who the workers
are and how fast they are so

I lose control of the sequence of workflow
and that can make more work downstream.

If this work were going to
be flowed onto a trailer.

I would have to have additional workers
here to disentangle the orders.

So I would have slowed things down.

So another disadvantage can
be a serious disadvantage.

OK well let's also let's see how
productive we were one two three

four five six seven eight they
finished eight orders in five minutes.

Mario would you be the keeper
of the productivity figures here

eight eight orders completed.

And now let's do the pick and
pass operation.

And.

Deep I'm going to ask you to work zone
one which is locations one through five.

J..
You're in the middle six or

ten Barbara here at this end.

OK.

Now we're going to work
as before except that.

Only Depot will start orders.

Everybody else has to stand
still until people get you work.

When Depot gets.

Finishes an order just place it here.

OK and try to keep them in
sequence as much as possible.

And I think we're ready for our countdown.

Yeah.

All right.

Let's get active active here.

OK.

That's right.

You're the only one who completes orders.

OK.

So you can see one of
the problems already with this

is starting up the products starting
up the at the beginning of the day.

If you had like that distribution center
we just of which was just saw a picture.

If you had sixteen workers
at the beginning of the day

that first worker is going to
be busy in the other fifteen and

be standing there waiting until
the work gets out of there.

So you know which maybe five ten
minutes or so there's a potential for

a lot of waste.

And in fact places that use this

frequently have more
complicated management issues.

Because what they have to do is bring in
the workers early some of the workers come

in early for example we might ask people
to come to work thirty minutes early.

Jay to come in fifteen minutes early and
then Barbara at the regular time and

then we might dismiss them at the end of
the day in that sequence the who goes home

first fifteen minutes later done and
fifty minutes after that barber goes home.

And so what this does this means that the
engineer or management has got to stay.

You know come to work an hour early and

stay an hour longer at the end of the day
management tends not to like that.

We can see a mother possible problem here.

Work in process.

This is work that was moving towards
the customer and has stopped.

OK In general we don't like to see
that because for several reasons

one is it's got to be double handled
because deep it had to put it down.

Diva had to put it down and then Jay
will have to pick it up and double hand.

Thing is waste and waste costs us money.

Now it may be five seconds three seconds.

But when it happens in a distribution
center like Amazon with one thousand five

hundred workers you add up three
seconds here five seconds there and

pretty soon you're talking
significant person hours and

then it's a cost that just
continues all the way.

In addition whenever the product
stops like this it's at risk.

Somebody could bump it over
somebody could put the wrong item

in this they could get out of sequence
the orders could get out of sequence.

So.

This is where quality problems come in.

In addition this style of operating as
an assembly line has the problem that

it's never possible to get it right.

There's always a bottleneck
so-called bottleneck.

Does everybody see
the bottleneck here and.

You can tell a bottleneck because there
is work in process right in front

of the bottleneck.

OK so that this is like
a giant arrow pointing to J..

Now this creates social problems.

In the warehouse because Jay is
sitting here with this giant pulsing

arrow pointing at him thinking
God I'm being made to look like

an idiot here I wish
people would slow down.

And you know it.

Maybe create tension in the workplace
between these two workers.

In the same way.

Barbara may not be so happy.

If.

She's gets ahead of Jay and
has nothing to do.

You can frequently
downstream of the bottleneck

the workers are generally what they
call starved is the technical word for

it which means they've got nothing to do.

And so Barbara may be thinking.

Come on.

J M You know you're not give me
enough work to do this is a big sign

to management that maybe
they don't need me.

OK So again workplace tension.

And if you are the manager of this.

These are all problems that
you've got to deal with.

Now in a real warehouse you might
not actually see this sort of thing.

And the reason is this work
tace place tension like this.

The social conflicts that
you create by the imbalance.

Thank You are so
severe that what they give

the workers take care of it themselves
in a real warehouse what will happen.

Is.

Everybody will slow down
to match chase pace.

Thank you.

Everybody will slow down to match J.C.

pace and you'll never get as manager
will never see the difference.

All you'll know is that.

How come by.
Workers aren't getting being

more productive.

That end up working at the speed
of the slowest worker.

OK so that's not good.

That's not engineering efficiency.

OK Thank you.

Workers want to return to the break room.

And have a cup of tea and

we'll resume in one minute while
we assess the productivity here.

Notice that there's a lot of productivity.

That doesn't get that accounted for.

Here the work in process.

This is nothing you can
ship to a customer.

So you don't get credit for it.

It's not anything that is viewed
it's nothing that gets measured

in the important statistics
which is is the customer happy.

This is just work in process.

And so we will ignore it and we will
count simply merely the completed orders.

And do we really finish only five orders.

OK.

Five orders.

And partly.

You can see that depot was our best

worker was sort of impeded
from being more productive.

All she did was produce a lot
of work in process which created

its own problems we had to
find a place to put it.

You know it may gather dust it.

It's just sitting there.

OK We're can process now.

In real warehouse what we can do you
probably thought of this already.

What we can do is try to be a little
better about assigning the work here.

Now all I did in assigning the work
is divided into equal pieces.

And statistically speaking since I
generated the orders I know how they

are structured about equal amounts of
work comes from each of these locations.

But if I really want to be better
smarter about this what I have to do is

give a little extra work.

OK now and J. A little less work.

Now how do you think that's
going to make people feel.

OK yeah yeah I mean the natural human.

If you start off the day by saying
this is your assignment here and

your workers doing great and
then you come back later and

say I want you to take somebody
else's some of Jay's work.

The natural human reaction is why
are you asking me to do their work when

you make them do their work.

That's what the way anybody would feel.

And so
again this is creates workplace problems.

So all

of the what typically gets done is to try
to be smarter about dividing up the work.

But now we're talking
about significant expense.

Because in order to divide up the work.

First we'll have to measure the work.

That means I have to do
time motion studies.

So I need it now I need a trained engineer
out there with a stopwatch watching

a worker pick this task.

I pick this item.

Shampoo off the third shelf.

One hundred times and then I want to watch
another worker do one hundred times so

that I want to do not
just the pill shampoo.

I want to do the white shoulder
shampoo and watch them pick that.

And so you can spend months or
years building a work content model.

How much work does it take in order
to pull product of different kind.

It's off different locations.

Now what you get from that is
of course just a very good guess

you get a distribution.

Sometimes people will pick it faster
sometimes they'll pick it slower but

you get a distribution and
you kind of look in the middle and

say you know what that
should be my work standard.

That's what I should expect
people to do but guess what.

When you have real
people in the warehouse.

They're not statistically
average workers there may be

clumsy as you just saw clumsy and slow and

other workers who are fast and actual And
so you don't get a statistical

statistically mean worker you
get a particular worker and

so what that means is there's
always imbalance in the warehouse.

Even if you have paid one
hundred thousand dollars for

the labor management module which enables
you to track the productivity of each of

the jewel worker and your engineer
does the time motion studies and

you divide the work up to match.

If there still can be a balance.

It's unavoidable.

OK let's try another method now this
other method is inspired by the.

The bees the ants in this case.

And here's how it works.

So if I can invite my workers back.

Here for the third and last.

Last order picking.

It's going to work like this.

I'm going to ask people to go first.

She'll be followed by Barbara.

And Jay last deep when we say go
you're going to pick up an order.

Not yet but soon as we say go in there.

You're going to pick up an order and
get to work.

And you're going to carry that
warder all the way to the end.

So when you're done.

Put it down here exactly as before.

And now here's the distinctive thing.

Soon as you put this order
down you come back and

take over the work Barber is doing.

So whatever barbers doing she's
going to pass it to you and

then Barbara you will go back and
take over the order that Jay is picking.

You want to get it from Don't
stand there and wait for

him to let you know you wanted to go ahead
and take of Robin take it over because

the object here is to get Jay back on
the new orders quickly as possible.

That's how it works OK Now
in order to use this method.

I don't need to know
the speeds of the workers.

I don't need to know a word content model.

All I need to know is who is faster and
who's slower.

OK And in every place I've ever been
where there are workers the workers no

management may not know but us the workers
they know who's faster and who slower.

OK so
if you want to give us the ready set go.

OK Off goes deep.

OK.
Yesterday professor you start right now.

That's right.
As soon as you can.

No no you just follow her.

Let's go.

About this.

Look if you one extra credit point.

OK

now watch people when she finishes because
this will be the distinctive thing here.

Finish.

OK good.

How come take Barbara's work.

Great perfect Barbara you take over here.

Perfect.
Jay and you start a new one.

That's it that's all
the workers need to know.

Notice how much easier it is for
management in that I don't need.

I don't care about.

We're content.

We're content is sort of the classical
industrial engineering creation of

the twentieth century right Frederick
Taylor and anti motion studies and so on.

It's very time

consuming to build a work content
model and it has to be maintained and

it's also grounds for
a lot of labor management dispute.

In this case I can avoid it entirely.

I don't care.

I simply still sequence my
workers from slowest to fastest.

OK.

Now people will go back take over.

Barbara's work.

OK good.

OK And now.

Take over J's.

And one thing you might notice here is the
deepest sort of carved out a region for

herself and it's a larger region than
that carved out by the other workers.

And that's because under this method
workers tend to expand to fill territory.

Which is a measure of their productivity
compared to their colleagues.

I'll have more to say about that.

But deep our best worker is not

in impeded in any way she can cover
as much ground as she's capable of.

Yes QUESTION.

I've given it the name bucket brigade.

Because I think of it in terms of
each workers got a bucket of water.

They're trying to get it to
the fire as quickly as possible.

Deepa throws the bucket of water on
the fire and then runs back and gets.

A bucket from the preceding person.

So I call it a bucket brigade.

QUESTION Yes.

Very good question.

In fact.

When we first invented this method and
tried to get people to use it.

There was some resistance
because of concern for

quality control because I don't know
who's going to pick a product and

here it could be any one
of the workers and so

in some cases you can lose the ability to
track errors back to individual workers.

But it turns out that.

This method smooth.

The flow of work so that in all
cases that have been reported to me.

The quality has improved.

Even if they lost the ability to track
it back to the individual worker.

It's loose the flow of work and therefore.

When you have big chunks of work coming
through it tends to put huge pressure

on whoever's got that chunk of work
in there more prone to make errors

in this case because it's moves the flow
of work errors have always gone down.

Now since we invented this
some of the providers of I.T.

support have figured out
how to track in fact.

Exactly who it was that picked each order.

So that's less of an issue now.

But good question Are we almost.

OK.

Yes.

What a great question.

But a great question.

I'll tell you a little story about.

About our first implementation of this
where the workers had exactly the same

question that we really need say.

It was a different Jake they're.

OK.

Thank you.
Workers.

I'm going to you're you're done for
the day.

Thank you thank you.

Let's just.

Yes.

That.

We've we've had that problem.

Generally what we do when we go in is we
ask the workers well who's your fastest

worker and.

At first we were not certain.

We'd get a clear answer.

In fact we did a lot of surveys we'd give
people secret ballots and ask them to rank

their team members or their fellow workers
according who they think is fastest and

you know we found that there
was remarkable unanimity.

All the way down from Festus to.

Next to the last.

And the the only point
of extreme disagreement.

Thank you.

The only one of extreme
disagreement was who was lowest.

And the slowest worker almost always
rank themselves much too high.

So we found that now occasionally we
even the workers themselves were off but

not by much but what we would
observe is that a slower worker or

somebody that we thought was slower.

We keep catching up to the worker in
front of them and if that happened.

We figured out that you know what
this other work is probably faster.

It's what positions.

Let's put it to be here line two three

four five six

seven eight nine ten.

So that's twenty five percent more
productive than the next best one.

And that's pretty close to what we see in
industry twenty five percent improvement.

And so across a range of implementations
was there another question up.

Yeah.

You should rank them from
fastest next fastest and.

You know.

Yeah but if they're almost as fast.

It tends not to matter.

The main thing is you want to have your
festers workers at the end of the line

because you don't want it you want to you
don't want to impede your best workers.

It depends on the industry but

generally sixteen is an unusually large
number so that was one of the larger once.

Let me.

I've got a few more slides to show you.

OK sorry one second here.

The log again.

Here.

How many.

Product yet points each one of them.

That.

Typically what they'll do is

well let me show you some
pictures here from a real one.

He so
here's the brigades just like as in ants.

Each worker is required to know only
one very small rule which is just

continue working until somebody takes
your work and then go back and get more

from the designated person perceiving you
so you go back upstream and get some more.

So that's all they have to remember.

No they didn't know.

And then this is what I do in my career
as a professor publishing in academic

journals what we did was basically
build a mathematical model of this.

And you can prove from this that
the balance of merges spontaneously so.

In other words.

It's almost as if you set up
a force like that due to gravity

that constantly pull the workers
to the right place and

I have a little simulation to show you
sort of keep stepping over you here.

And this simulation the rig worker is the
fastest worker in there walking back and

taking work over from the next fastest
worker which is the Green Worker and

now this line is perfectly
balanced there is no bottleneck.

And I know this because this is
a simulation that I wrote so

I can get to say how much work
there is and where it is.

I've indicated the point of
perfect balance by the way.

Yes I'm twisting my laptop and
it doesn't seem to have any effect.

Maria.

There K..

Yes.

There are so.

OK.

So you can may see the little tick

marks there that's the mathematical point
of perfect balance in the real world.

That's the point we'll never know.

We can only try to guess at.

But now I'm going to reassign

move these workers to random positions
I'm just going to click here.

And so the line is very unbalanced now.

And you will see that just within the
space of a few minutes the movements of

the workers will reposition them in such a
way as the line balance is reestablished.

So watch where the handoffs are now and
you can see.

Initially there.

Quite far from the ideal but

each successive set of hand-offs
is going to be closer and

closer to the ideal and this is with
the workers having no consciousness of it.

They don't even know they don't even have
to care about smoothness of flow and

balance of work and bottlenecks.

Or even where the and I don't even
have to know where the bottleneck is

I don't have to know where
the point of perfect balance is but

they are in fact computing the point
of perfect balance right there.

So you can see.

Was it taken three or
four orders for the workers

to complete before they've reestablished
the sort of mathematical optimum.

OK So they they in effect what
we've done is make the workers

themselves into the computational device
that computes the answer to our problem.

The engineer doesn't have to do it at all.

We simply co-opt the workers.

Into doing the computation for us.

OK Here are some companies
that use bucket brigades.

And these are the numbers they've
reported to me the changes in their

productivity The three eight of throughput
the rate of order completions and

you can see these are pretty significant
numbers twenty five percent fifty percent.

And these are you can see
this idea is very simple.

It doesn't require any I.T. support it
doesn't require any special equipment

in fact the biggest challenge to its use.

Has typically been the fact that
the warehouse may have invested

heavily in I.T. support that is now
no longer necessary and people don't

like you don't like to go to your manager
and say you know the hundred thousand.

I spent six months ago for
the new I.T. system.

We don't need it and

that's it's a little embarrassing position
to be so these are all numbers that

were reported to me by the engineering
departments in these places and

for most of them the evil mentation
time was on the order of an hour.

Yes.

Great question and in fact that's
a frequently asked question.

And now I'm no expert in compensation or
incentive pay or anything like that.

I can tell you that the range of
companies here have implemented

a range of compensation schemes
the one from the picture

that I showed you with the sixteen
twelve to sixteen workers.

Picking for Wal-Mart.

This was not a Wal-Mart
distribution center but

up company who single customers Wal-Mart.

They didn't do anything.

All they said was OK we're no
longer picking by the zone system.

We're now picking by Booker brigades and
they still got a They're

interested merchandisers they got a twenty
five percent improvement in throughput.

The same week and
they did nothing different.

So my ex experience is that
the very fast its workers basically

are the fastest workers because they can't
help themselves they're competitive.

They're have pride in their work and
they're going to work fast.

No matter what.

Unless you.

Beat them with a stick or something that.

It just matters to them to be good.

Yes.

Because here.

Well let me let me show
you here in the picture.

Here's our very first implementation this
is the national distribution center for

C.V.S. drugstores in Knoxville Tennessee.

This is before back up again
they're doing zone picking and

their philosophy was you can't get
product off the line and to the.

Summer until you get started.

So they had their fastest worker
at the beginning of the line.

Opening these red coats the way we call
them or boxes and throwing money line.

In with the predictable result there
were a hundred red boxes on the line.

You can see all the problems first of all
how can you tell if when workers unusually

slow can't tell.

Everybody's got you know it's
a zillion red coats in front of them.

You can't tell if a worker
is unusually fast.

This is the exactly environment in which
everybody slows down to the pace at

the slowest pace.

There are other problems as well.

This order picker in the white
shirt here when she picks

a tooth brush to be sent
to a particular store.

She's going to turn around
to put it in a red tote But

there's one hundred toes very
easy to put in the wrong one.

Furthermore somebody upstream.

May have pushed the train of totes
in order to bring more work in and

so when she turns around.

It's very easy to put
the wrong toe because

the wrong might have gone down here.

If you're lucky she's going to
have to walk to the red zone and

that's unproductive.

When we use what we
implemented bucker brigades.

This is what it looked like afterwards.

You can see all that work
in process has disappeared.

Now we can see the slow worker.

Now we can see the fast worker.

Well this is the fastest worker
in the now in a black shirt.

She's just pushed off this tote onto
the powered conveyor it's going to

run down the shipping department in
this case there were nine picket lines.

Yes.

We just ask them.

We just ask them a way to explain
what we were doing they said

OK well that sounds good but the workers
like this because it made them more

as a team felt they felt more as a team
rather than they were doing zone picking

before they felt like their whole world
was between these two ends of the zone.

And they had nothing to do
with each other and now.

I like the way C.B.S. did it they were the
opposite extreme for Andersen what C.B.S.

did was tell the workers
you're part of a team.

The Festus worker is the team leader and

we're going to give her
a special increase in salary.

And you can see.

I mean everybody sees the hierarchy.

If you're the second fastest worker you
want to be catching the fastest worker

because you get to be charge you
get the little bump in salary

you get the I would be essentially that
person sort of became part of management.

If you're a new worker you're
at the very beginning and

basically every time they pick an order.

It's a peer review you want to catch
the person in front of you and

you want to avoid being caught for
the person behind you and

it's sort of a benign peer review
because everybody's in it together.

It's C.V.S. what they did was say kill
the workers for every dollar you save this

will give the team sixty cents and so
they could share that additional benefit.

I like that but.

They got a thirty four percent increase
in throughput Anderson merchandisers got

a twenty five percent increase in
throughput they didn't do anything so

I don't know what they're as I said
I'm no expert on incentives that pay.

I don't know what the right
thing to do in this case is.

Can't we do just as well without j.

In fact.

This team had nine workers and
one of them was a relatively new worker

who only you know traveled about that
far before people took over her work and

in fact during lunchtime
the workers in fact came to us and

said exactly the same thing we can
pick just this fast without that

slowest worker and divide
the incentive in mind this one ways.

Can you find something else for her to do
so now the workers are basically thinking

like management about how to allocate
the workers most efficiently to the.

Tasks.

This is their own measurements reported to
us over time so the horizontal axis you

see time here one point zero is
there some work standard that's and

it's disguised that because they didn't
like it to be known but this is how many

pics per person hour they expected the
worker to achieve and in the twelve weeks

before they were picking at about ninety
five percent of work standard which in

my experience is universal in the business
world because if your workers are picking

one hundred five percent of work standard
your work standard isn't high enough.

If they're picking at eighty percent
of work standard your work standard

discreetly unrealistic and so there's some
complex negotiation that goes on there and

so that everybody in the world.

This is like Lake Woebegone where the.

All the children are above average.

Well in this case of all the workers or
ninety five percent of work standard.

It's on the twelfth week that we
implemented bucker brigades and

they measure their own predictive ity
their industrial engineering group measure

the productivity here for
several more months and

the last measurement they got it
was thirty four percent above.

The work standard.

So what this means is a mansion if you're
having like the Amazon are these people

similar way hired hundreds

of additional temporary workers during
this time of year and had overtime.

If you can get be thirty four percent more
productive with your current workers.

You can avoid a heck of
a lot of overtime and

you can avoid reliance on
temporary workers as well.

So that her turned up be
very important for them.

Now this you now know the idea.

You are all if you raise your hands
you are now fully empowered to be

bucket brigade consultants you
know as much about this as I do.

That's all there is to it.

Now of course I wish I could say that I
discovered it in such simple way but in

fact I did did it by with my colleague Don
Eisenstein from the University of Chicago.

He was like a.

Student at the time we pushed
around a lot of equations and

it sort of came out of the equations and
we kept thinking that will work now and

then finally we decide well let's
try it and then afterwards.

It became obvious.

So I guess this is the way
the research goes in general.

Here's some other places
that use bucket brigade's.

You may recognise Subway sandwiches.

They've got a similar problem they're
assembling different products.

Each one is slightly different and then of
the cashiers at the very end of the line.

When the last customer has paid and
left you know have

idle work pass to the here but all the
customers are at the front of the line.

HOW DO YOU GET THAT WERE capacity
back to the beginning of the line

they have experimented with doing
a sort of round robin where

that worker at the end comes back
to the beginning of the line but

you can see what happens then everybody
ends up following the slowest worker.

They had also experimented with balancing
a traditional assembly line they had

called in a consultant to balance in

the same way a lot he did time motion
studies on spreading mayonnaise on bread.

OK Now when I was in college I
worked in the fast food industry.

I spent my time at McDonald's and I can
tell you this makes perfect sense to me

they couldn't get a work standard because
when they looked at the distribution of

task times in the fast food industry.

It looked like this.

The distribution was all variants and
scarcely any defined to mean at all.

So if there was no work standard they
couldn't balance in the center line.

The nice thing about Becker Brigades is
they don't care they move the workers to

where the work is based on how long did
it take you make that last sandwich.

We did this Mitsubishi plant.

Again we were just walking
around with the plant manager.

This is a little sub assembly
line that ran off here and

the lady in Brown is attaching
a subcomponent to a Tele.

Said here.

The biggest simply line is
moving from right to left

in this process was the bottleneck.

And just like we teach him in us real
engineering they had taken the work to do

and divided up into tasks and measured the
standard allowable minutes for each one.

They're very good industrial engineer had
posted a little sheet here that described

each of those tasks and
assigned subtask to each of the workers

all we did was pick that very nicely
prepared sheet pull it off the wall and

say OK here's how you do
bucket brigade's do it.

This was no longer the bottom line.

Bottleneck.

In fact the work that had been spread out
all along this green conveyor out here.

Shrank up just to this tiny little area.

This you don't see in the televisions in
the front of the line here because they

finished all of this work in
process got moved downstream.

You could see all the other workers in
the factory looking back to see what

what's happening there.

How come with all this work.

It's flooding down on us.

In the you point manager gave
instructions for the plants and

grounds people to tear out this
conveyor they didn't need it anymore.

And all of that happened when
the space of two or three hours.

So the essential things to make this
work are basically your workers have

a single skill.

So it's got to be simple assembly.

You probably don't want your dentist or

a brain surgeon to be using
this sort of technique.

It's got to be a single skill you know
I hand coordination simple assembly.

Yeah.

Now let's get back to ants.

But.

Yes.

No No In fact you need cross training
here because you can't be sure.

Now see when you have credit when you
talk about cross training you're implying

there's more than one skill.

Yeah so you've got to be careful.

It's either got to be a single scale
like order picking in a warehouse is

a perfect environment because every.

Thing is just the hand coordination
even this simple assembly.

There is a little bit of skill or
practice involved in this but

it's still simple enough the workers
can pick it up pretty fast.

When you talk about cross training it
suggests there's more skill involved and

then you have to be a little more careful.

You've got to get some
cross training then.

Either by running the bucket brigade for

a while or giving the workers cross
training before you put them on the line.

Yeah.

Yeah some people.

Here's a group that could not use
that Coach leather wear makes those

nice five hundred dollar yuppie handbags.

If their factory they wanted to use
brigades in the sewing but they could for

most of their product but they had one.

There he expensive product that
required a very special seam and

there was only one guy who could do it.

And so they didn't want just
anybody sewing that seam

because you didn't want to ruin
a five hundred dollar handbag.

So they had this guy parked on that
machine and he's the only one who could do

that they could run bucket brigades
downstream of him and they could run buck

or Gates upstream of him but
he was the only one who could do that.

Let's get back to ants.

A colleague of mine sent me a paper after
we had published our national work on

bucker brigades and it included this
picture and this was describing

the behavior of a seed harvesting ant
Messer barbarous which is found in Spain.

And you can see something that looks
remarkably like bucket brigade here.

This is a seed harvesting and they're
describing the behavior of ants carrying

seeds from a pile back towards the nest.

In what happened was they discovered
the slower worker would pick up the seed

here carry it for a short distance he
would be taken over by another worker and

in fact a faster worker.

Carried still further and
then taken over by.

Another still faster worker who
carried it all the way to the nest.

Now sometimes it would be
three ants four five six ants.

So the size of the bucket brigade if you
will sort of varied from time to time but

it was the Clearly the distinctive.

Pattern of it starting work
by the slowest workers.

Ants in this case passing it to
a series of ever faster workers

until it's delivered to the nest.

Come on with a reading my papers.

I don't think so.

So what we did was write a paper with
the biologists who found it and.

They didn't know what to make of this
they simply describe this behavior and

said Isn't this curious.

Well we figured out what was going on.

It turns out that there's two
levels of self work in the sation

happening in this case.

Now.

In our warehouse what we did we have
to have the industrial engineer or

management or
somebody get set up the system.

You got to sequence the work or

some slowest the fastest and then self
organization will balance the line.

But the ants to it even more
fundamentally what they do is each ant we

hypothesize follows
an even simpler rule and

that rule is grab the first seed
you can carry it back to the nest.

That's it.

Grab the first you know and the neuron
brain has only about one hundred fifteen

your arm so they're not very smart they
can't remember slowest a fast this or

anything like that but they can
remember the impulse grab the first see

that you can carry back to the nest.

The trick is the bio when the biologists
measured these they discovered that

the Not surprisingly the bigger
ants are the faster ants

the smaller and so
the slower the slower ants.

OK So now look what happens.

Imagine you're the biggest ant.

You walk out of the nest and
you walk up the forging trail.

Here comes in and carrying a seed.

You grab it.

You can do it because
you're the biggest ant.

You rested from them and
you carry it back to the nest.

You walk out again along
the forging trail.

You can grab any seed you see
because you're the biggest to hit

the biggest ant which
is also the fastest and

is going to spend all its time
here at the end of the line.

Just like the.

OK Now imagine you're the littlest ant and
also the slowest and.

When you walk out of the nest and
travel up the forging trail.

You see a lot of seeds but

you can't grab any of them because
only bigger ants have them.

You're the littlest.

You travel all the way to the end of the
foraging trail until you find the pile of

seeds and only then can you pick
up a seed and you pick it up and

start to carry it back to the nest in the
soon as you do the first chance you meet

is going to take it from you.

And you go back out here and
pick up another one.

So what happens is there's this
sort of two face in a station.

The first one in which the ants
sort themselves from smallest to

biggest which is the same as slowest the
fastest and the second one in which they

balance themselves to maximise
the rate of flow of seeds to the nest.

And in fact when you do a little
simulation of this it's pretty amazing

to watch because at first.

It's just going every which way and
you think what's going on and

then all of a sudden from at of that
Murk there's sort of crystallizes this

pattern in which all the big
answer on one side and

all the little answer on the other side
and all of a sudden you see that and

then here's the seed starting here and
accelerating to the nest and it

turns out in the biology context there's a
lot of other reasons economic reasons why

you want the seed to accelerate to the
best because in a sense as it gets closer

to the nest it becomes more valuable
just like a supply chain for a product.

You've had more invested in it.

You know the products been moved that I.

I Pod is a lot more valuable
in a Circuit City in

Los Angeles that it is back on
the factory floor in Shanghai.

OK it's a lot closer to being able
to realize revenue for it and

in the same way that seed
is a lot more valuable

close to the nest where it can
be converted to calories and so

it makes sense to accelerate
the product as it gets gets closer.

OK I think that's all I
want to say about them.

I'm going to close here with
probably one more slide after this.

I want to point out that there
are bad forms of self organization.

What you see here is called a circular
mill and this is a particular affliction

of the Army and we're looking at
an African army and killed the driver and

here and they have the feature
that they are blind nearly blind.

They're so blind.

It's not even worthwhile for
them to wear glasses.

They follow the scent trail of their
colleagues at the so-called pheromone

which is a chemical deposit that the
answer respond to and so as they travel

along they don't see where they're going
they just smell where they're going now if

you were ever under attack by army ants
what you can do is take a stick and just

sort of tease the flow of the ants into
a little loop and when you do they will

continue following the smell of the ants
in front of them exactly like this and

they will continue until they
die of thirst or starvation.

They're not the other only
creature susceptible to this.

Incidentally is a certain kind of
caterpillar in South Africa that

has the same property but

that's a bad form of self organization and
exactly as the ants have this.

Weakness organization
this is a weakness of

self organization in general this is where
the engineers really needed the design or

when you try to pick an idea
of self organizing and

you realize that in a human context you've
got to be careful about bad things that

can happen to self organization and
in fact bad things can happen.

We already saw with Subway
sandwiches having everybody

circle around eventually makes
everybody follow the slowest worker.

So that's a bad form of self organization.

But even better brigades are susceptible
to bad forms of self organization.

If you get the workers
in the wrong sequence.

We started them slow it's the fastest.

If you get them in a bad sequence
strange things can happen.

Counter-intuitive things here for example
are the results of two simulations.

This is the assembly line
from the start to the finish.

And here I flooded the point
of handoff over time and

you can see those first few handoffs
kind of jump back and forth and

but gradually came to this
point of perfect balance.

And here the line is in perfect balance.

With I put the workers in the bad
sequence a really pathological sequence.

If I plot the points of hand-offs
you can see they look random.

And in fact in a very
strong sense they are.

Almost indistinguishable from randomness.

Imagine that you are at
the shipping department from a.

In a distribution center and
you're waiting for orders to reach you.

All of these orders are perfectly
identical all of the workers are.

Absolutely reliable but you've got
the workers in the wrong sequence

what you'll see is work
coming off the line.

In chaotic manner.

In fact this is true mathematical
chaos indistinguishable in effect from

randomness and so it's as if
the shipping department experiences.

A production department that has no

there's flipping coins to
decide what to do next.

Indistinguishable.

Whenever you have that
kind of variability.

It is the enemy of efficiency.

That's what industrial engineer struggled
to eliminate in business process.

This is variability.

OK I've gone on too long.

Let me just point out that if you want
to know more about Beck or brigades.

Here's my web page and I've got links on.

That bunker brigades You can also send
me e-mail or if I don't know if anybody

even uses the post office anymore these
days but if you do write me a letter.

OK Thank you all very much.

You've been very patient.