If you've been reading this blog for awhile, you know that many of my ideas about efficient work come from lean manufacturing. This is the framework I use in thinking about how to reduce waste.

One of the ideals of lean manufacturing is single-piece flow: building one item at a time to precisely meet customer pull. In this scenario, there's no buildup of needless inventory, because everything is built to a specific customer's demand. (Chrysler offers a cautionary tale about the consequences of building a giant pile of unsold inventory.) This one-piece flow stands in stark contrast to "batch processing," in which orders are built up to a certain level before production starts, in order to reduce average production cost.

Knowledge workers like you, who spend their days handling text files, spreadsheets, and email, can't operate in this ideal world of single-piece flow. For one thing, their work is too complex. You've got multiple value streams flowing through you, with each stream flowing at a different rate. That makes it impossible to determine takt time (basically, the average pace of work, for you non-Lean folks).

More significantly, the volume, speed, and variability of customer demand makes it ludicrous to work on each item as it arrives in order to precisely meet customer pull. You'd have to switch tasks every two or three minutes. But paradoxically, this is precisely what you do. Everytime an email comes in, you check it. Everytime the phone rings, you answer it. Everytime a coworker has a question, you stop your work to answer her.

For your own sanity (and efficiency), there has to be some sort of batching of work. But how do you reconcile that with the ideal of single-piece flow and customer responsiveness?

The answer, I think, lies in viewing knowledge workers as monument machines (large, complex machines through which multiple value streams flow). These machines inevitably take a certain amount of time to set up. Batch production makes sense for these machines, where the time and cost of setup is amortized over the entire production run. (That's why simpler machines are better, whenever possible.)

You're like that, too. It takes you awhile to set up -- to "get into" an engineering problem, a spreadsheet, or a performance evaluation -- so it makes sense for you to work in batches. (If you've ever been interrupted in the middle of one of these tasks and lost focus, you know the effort required to get back into it.) These longer "production runs" will reduce the waste and inefficiency caused by switching between tasks too often.

Of course, this means that inventory will build up -- unanswered voice mails and emails in particular -- so the flow of the value stream won't be perfectly smooth. But I believe that the waste of waiting borne by the downstream person (or the ultimate customer) is less than the waste caused by you switching between tasks several hundred times per day. The best you can do is to minimize the waste of waiting by batching your email and voice mail -- dealing with those inputs regularly, two to four times per day, for example. And in all likelihood, this kind of methodical approach to handling your paper and electronic inboxes will result in fewer lost or unacted-upon tasks.

So take a few minutes and look at your work as a value stream. Apply the same critical thinking to your workflow as you would to a physical production line. And see if you can't justify intelligent and mindful batching, instead of the mindless urge for one-piece flow.