Note

Your cost-per-case benchmark is probably wrong.

Not because the arithmetic is bad. Because of what went into it, and what quietly did not.

Most firms that price fixed fees do it against a single average. Somebody built that number once, it was correct at the time, and it has been load-bearing ever since. It is usually the most consequential figure in the business and almost never the most audited.

Here are the three ways we have seen it go wrong, in order of how much damage they do.

1. The population includes things that were never the work

Every billing system accumulates records that are not representation. Administrative files. Balance-forward entries from a migration. Test matters somebody created in 2019 and never deleted. Matters opened and closed the same week because the client went elsewhere.

None of these are wrong to have. All of them drag an average toward zero. In one rebuild, removing them alone moved the median materially — and the firm had been quoting against the lower figure for two years.

The fix is not a minimum-dollar filter. A floor that drops everything under three thousand also drops genuinely small matters that represent real work, and it inflates the result far more than the noise it removes. Exclude by reason, name the reason in the data, and count how many each reason removed.

2. The filter silently misses records

Averages are computed over whatever matched the filter, and a filter that matches nothing reports no error. In one case the case-type filter relied on a field that was blank on roughly seven hundred genuine matters. They were not excluded — they were never seen.

A query that returns fewer rows than it should looks exactly like a query that returns the right rows. Nothing in the result says otherwise.

Before trusting any aggregate, count the population it was drawn from and reconcile that count against the total by hand, once. If the two do not agree, the difference is the finding.

3. A lifetime average is used to price a fixed term

This is the subtle one, and it is the one that costs the most.

A fixed fee covers a period — four months, six months, until a defined milestone. But almost every report measures a matter across its whole life. Those are different questions, and a number that answers the second cannot answer the first.

Measure in engagement months instead: what a matter bills in its own first month, its own second, its own third. When we did that for one firm, the picture inverted. Very few matters exceeded the quoted fee inside the fixed period. Most were still billing well after it ended. The firm was not underpricing the work. It was underpricing the duration.

One more trap in that calculation. When you bucket by engagement month, month four can only include matters that have actually reached month four. Counting a two-month-old matter as a zero in month four drags the figure down exactly where a fee term is being decided. Carry the sample size next to every column.

How to check yours this week