Musicians talk about fees constantly and payment terms almost never. A hundred euro gig paid in cash on the night and a hundred euro gig paid in seven weeks are not the same gig, but only one of those numbers ever gets discussed.

I have logged every gig I have played and every payment I have received since 2010. Three hundred and twenty five payments. Here is what sixteen years of receipts say.

The headline

Fifty six percent of my gigs are paid on the night. Cash at the end, envelope in the case, done before I pack up.

For the rest, the median wait is at least 48 days. The worst single case on record is 768.

Both numbers matter, and they answer different questions. The first tells you how much of your income arrives when the work does. The second tells you how long the remainder sits out in the future.

How the wait breaks down by year

Three things stand out.

2020 sits at 332 days. Five gigs, all in the wrong year. Venues that shut in March paid what they owed when they reopened, and some took the best part of a year. It is the only bar on the chart that is about something other than administration.

2018 has no bar at all. Twelve gigs, twelve payments, every one on the night. That was a year of small clubs, where the format is simply that you get paid before you leave.

The middle is flat. From 2012 to 2025 the median sits between 23 and 84 days with no trend in either direction. Fourteen years of not getting measurably better at this.

The share paid on the night

The median only describes gigs that entered a queue. This is how many never entered one.

This is the number that actually decides whether a month feels comfortable. A year at 76 percent, like 2019, means the money broadly arrives when the work does. A year at 30 percent means most of your income is deferred, and the chart above tells you roughly how far.

The two charts do not move together, which is worth noticing. A year can have a high same-night share and still hold a long tail, or the reverse. They are independent facts about how you get paid.

The number no average will show you

A median is a summary, and summaries hide queues. So here is the queue: every gig I have played that has an agreed fee and no payment recorded against it.

Eight gigs. One thousand and forty five euros. The oldest has been outstanding for 343 days.

That is about a third of a decent month's income, sitting in other people's accounts, and no summary statistic was ever going to put it in front of me. The median said 48 days and felt survivable. The queue says there is a three hundred euro gig from last September that I have quietly stopped chasing.

Two rules that make the number trustworthy

Anyone keeping a similar record will hit the same two decisions, and both change the answer materially.

Exclude same-night payments. Leave them in and the mean comes out around 36 days, which describes nothing that ever happened: more than half the book settles the same evening and the rest takes six weeks or more. Averaging those two populations produces a number that belongs to neither.

Do not exclude gigs that have not been paid. This one looks like the same kind of tidying and is the opposite. A gig with no payment has no payment date, so it silently vanishes from the arithmetic, and what survives is the payers who paid you. That is survivorship bias, and it hits recent years hardest, because old years have resolved while this year's slow payers have not shown up yet. Count an unpaid gig at the days it has waited so far, and read the result as "at least".

For me that second rule is currently worth 22 days on this year alone.

What I take from it

Sixteen years in, the shape of it is simple. Over half of the work pays immediately. The rest takes about seven weeks. A small number of venues take a year or never pay at all, and those are invisible in every average you might compute.

None of this needed special software. It needed writing down two dates per gig, for sixteen years, and then being willing to look at them properly.

The data

One note on the record. A 2017 payment carried the date 1 January 1970, an empty field that some software filled with the beginning of computer time; on its own it dragged the all time average 47 years negative. It is now stored as an unknown date, which is what it always was. Sixteen years of hand kept data has exactly the number of dents you would expect.