What a retainer actually costs you

A retainer is the most attractive line on an agency’s P&L. It is predictable, it is recurring, and it makes the forecast look like a product business rather than a series of arguments about scope. It is also the line most likely to be quietly losing money, because the thing that erodes it — an hour nobody wrote down — leaves no trace anywhere you are looking.

Here is the arithmetic, and the three places it tends to go wrong.

The arithmetic

Take a £3,000 a month retainer, sold as “about twenty hours”, delivered by someone on £52,000 a year. The headline rate looks like £150 an hour, which is comfortable. The real one is not.

Seventy-two percent. Now run it again at the hours actually worked, rather than the hours sold. In our own numbers, and in every agency conversation we have had about this, the gap between the two is between fifteen and forty percent — and it is almost never in your favour.

Seven points of margin, on one account, from five hours a month that nobody decided to give away. Across eight retainers that is a salary.

The problem is not that the hours were over. It is that nobody knew until the quarter was over.

Where it goes wrong, one

The timesheet is filled in on Friday. Retrospective logging is not really measurement — it is reconstruction, and people reconstruct generously in both directions. The half hour on the phone on Tuesday is gone. So is the ninety minutes lost to a file that would not export.

The fix is not discipline, because discipline is what everybody has already tried. It is reducing the logging to one action, in the place the work is already happening: a timer on the task, which already knows the client, the project, and whether it is billable.

Where it goes wrong, two

Scope creep arrives as favours. Nobody sends an email saying “please do an extra four hours unpaid”. They ask for one more round, or a version for the deck, or a quick look at something a different supplier built. Each is twenty minutes and each is obviously worth saying yes to.

What makes this survivable is not saying no. It is being able to say, in week three rather than month four: we are at eighteen of twenty hours, here is what is left, which would you like it spent on. That is a pleasant conversation. The one in month four is not.

Where it goes wrong, three

Unused hours are treated as profit. The month a client goes quiet feels like a good month. It is the clearest early warning you will ever get that the retainer is about to be cancelled — a client who is not using you is a client building a case for not paying you.

Roll-over rules help, and so does a cap on them, but the thing that actually helps is noticing. Two consecutive months under 60% of the bucket is worth a phone call, and it is the kind of pattern that is invisible in a spreadsheet and obvious on a chart.

What to actually do about it

• Track hours against the retainer, not against the month. The question is “how much of this bucket is left”, not “how busy were we”.
• Make logging a single click from the task. Anything that requires opening a second tool is a thing that happens on Friday, badly.
• Set the alert at 80% of the bucket, not 100%. At 100% the money is already spent; at 80% you still have a choice.
• Review under-use as seriously as over-use. It is the cheaper problem right up until it is the expensive one.
• Put the rate you are actually achieving — retainer divided by hours worked — somewhere you see it monthly. It is the only number in this post that matters, and almost nobody calculates it.

Nucle.OS does the first four because they are the same four things we needed. Retainer buckets with roll-over and a cap, a timer that starts from the task, and hours that count against both the project budget and the month’s bucket as they are logged — that is the Deliver it, and get paid half of the product. The fifth one is a decision, and no software makes it for you.