Business time is lumpy, not continuous

For most of a decade my calendar had the same shape every month, and I could have drawn it before the month started. A stretch of ordinary days at the end of one month, then the first working days of the next taking almost everything, then a fortnight where I could think.

That pattern held across a public practice, a corporate finance team and a firm running books for a portfolio of clients. Different employers, different software, same shape. Nobody designed it. It is what happens when the obligations that matter have dates on them and the dates are shared.

The rhetoric does not match the distribution

Every organisation I have worked in talks about work as a flow. Capacity, run rate, throughput, all of it borrowed from language about things moving steadily past a point. Plans get built on that language. Somebody works out roughly how much there is to do in a month and roughly how many people there are, divides, and concludes the team is adequately sized.

Then the month arrives and four days of it hold most of the pressure, and the team is not adequately sized on those four days, and is idle enough on other days to feel guilty about it.

The mistake is not the arithmetic. It is that the average has nothing to do with what anybody experiences. Nobody works an average day. You work the day you are in, and the days are not interchangeable, because the ones with a deadline attached cannot borrow capacity from the ones without. Time only moves one direction, so a quiet Tuesday cannot be spent on the following Wednesday.

Once I saw it I could not stop seeing it. Payroll is lumpy. Invoicing is lumpy. A software release is lumpy in exactly the same way, which is the thing I found most surprising when I moved into delivery, because software talks about itself as continuous more than any field I have worked in.

I did smooth some of it, and it moved

At the corporate finance job I spent a long time standardising the entries and reconciliations that made the close slow, and the close got meaningfully faster. I thought I had flattened the curve.

What actually happened is that the lump moved and got sharper. The work that had been spread over a week of grinding was now concentrated in the two days where judgment was required, because those were the only days left. The total came down. The peak did not come down proportionally, and the peak is the thing that determines whether the month is survivable.

I would still do it again. But I stopped describing it as smoothing, because that word promised something the work did not deliver.

What the lump is doing there

The other correction took longer. I used to treat the concentration as pure waste, a scheduling failure that a better organisation would not have.

It is not only waste. A shared date is a coordination device, and a very cheap one. Everybody knows when the answer is needed, so nobody has to negotiate it, and the thing that is finished on that date is finished for everyone at once. Work that has no such date tends to have no such moment, and I have watched plenty of it drift for months in a way the close never drifted.

So the deadline buys you something real, and it charges for it in peaks. That is a trade, not a defect. What I object to is pretending the trade was not made, sizing the team for the average, and then treating the predictable overload as a character problem in whoever is standing there in the first week.

You can tell a lot about somebody’s job from which days they will not agree to meet on. They rarely explain why. They just know.