How to Think about Batches

Tuesday, September 13, 2011

You have a stack of, say, a hundred letters and envelopes to mail. You must process these by hand. Is it faster to do each step of the process for all the letters before progressing to the next step (again, for all the letters), or is it faster to everything for each letter, one letter at a time?

The answer may surprise you:

Why does stuffing one envelope at a time get the job done faster even though it seems like it would be slower? Because our intuition doesn't take into account the extra time required to sort, stack, and move around the large piles of half- complete envelopes when it's done the other way. It seems more efficient to repeat the same task over and over, in part because we expect that we will get better at this simple task the more we do it. Unfortunately, in process-oriented work like this, individual  performance is not nearly as important as the overall performance of the system.
Another writer elaborates further, in response to people who were skeptical of this claim:
The shorter stuff and seal times, though, are due to the fact that you are already holding the item from the previous step. You gain 1 second each time from not having to find and pick it up...

You lose between 2 and 5 seconds every time you move the pile around between steps. Also, you have to manage the pile several times during a task, something you don't have to do nearly as much with [one piece flow]...
Returning to the first post, there are other advantages that have nothing to do with the efficiency of the process:
[I]magine that the letters didn't fit in the envelopes. With the large- batch approach, we wouldn't find that out until nearly the end. With small batches, we'd know almost immediately.
There are other advantages to doing work in small batches that apply even for processes that are, or can be, automated:
All these issues are visible in a process as simple as stuffing envelopes, but they are of real and much greater consequence in the work of every company, large or small. What if it turns out that the customers have decided they don't want the product? Which process would allow a company to find this out sooner?

Lean manufacturers such as Toyota discovered the benefits of small batches decades ago. When I teach entrepreneurs this method, I often begin with stories about manufacturing. Before long, I can see the questioning looks: what does this have to do with my startup?

But the theory that is the foundation of Toyota's success can be used to dramatically improve the speed at which startups find validated learning.
This last point is impressive, and it reminds me of how an engineer once solved a challenging problem by learning how to "fail faster".

I have to admit that I was highly skeptical that stuffing one envelope at a time could outpace batch processing, but I suspect that it was my passing acquaintance with the great advantages automated manufacturing can offer in  terms of time savings. It is interesting to learn that batch processing not only doesn't always save time, but has other disadvantages.

This post is a reminder that our thinking about even simple things like stuffing an envelope can be limited by implicitly-held premises or assumptions. Checking against reality can very easily refute one's "wisdom," and discovering why one was wrong can both correct and lead to new  knowledge, including about matters that are not obviously related to one's original query. 

-- CAV


Jennifer Snow said...

I discovered this for myself years ago when I used to work at Marshalls as a clothing-hanger. They wanted us to take a stack of clothes, put hangers in all of them, then go back and put size nubs on all of them, then hang the entire stack at once.

Not only was this NOT faster than doing it the other way (hanger and size each item, then hang it on the rack), the difficulty involved in getting a hanger inside a shirt or on a pair of pants that was lodged somewhere halfway down a stack meant that the clothes often got stretched or distended--or had to be re-hung, anyway.

Not to mention what happened when all the clothing in a given stack wasn't identical, which happened quite a lot as well.

From what I've seen, batch processing works best when you have different people doing different steps--and even then you're not handing off an enormous stack of partially-completed whatever, you're handing them off one at a time. It is more of an assembly-line process in that case, and it is more efficient in terms of overall time even though it's less efficient in terms of how much each piece is handled.

Gus Van Horn said...

Thanks for the illustrative example.

Employers like that both tempt clock-watchers to robotically follow orders and end up with work forces who have low morale.

I think you have a point that batch processing can be better than single-piece-flow, although this article teaches me to be wary of not looking at any such application with a jaundiced eye.