How the figures are built

Fee basis and the commission haircut

Every observation carries a fee_basis field. Gross-of-commission sources — industry rate cards and agent-rostered artist profiles — are converted to artist take-home before percentile computation using a 20% default agency commission haircut (net = gross ÷ 1.20); agent-brokered bookings use the booking’s recorded commission. Already-net sources (MU/Equity rates, venue budgets, verified submissions) pass through unchanged. Every published figure is therefore artist take-home — client-facing gross prices typically run ~20% higher.

Re-sighting, deduplication — and what n really counts

A listing that survives week after week is information — the price is stable — but it must never count twice. When a weekly scan meets a listing it already knows, it bumps that observation’s sighting counter instead of inserting a copy.

Reading note: the n shown against a cell counts weighted samples, not unique gigs — an observation counts at every geographic level it informs and once per re-sighting, so cell-level n runs far higher than the pool of unique observations. The distinct pool size is published in every issue (4,431 approved observations in the August 2026 issue).

Cell thresholds and geographic fallback

Publication happens at three levels: city (major UK metros, with satellite towns folded into their hub — Hove reads as Brighton, Oldbury as Birmingham), region (the twelve official UK regions), and the UK baseline. Every cell in the grid prints. Nothing is held back. What a cell has to earn is not its place in print but its label.

The rule, in one line. A city cell is measured when it holds at least three observations collected in that city — three separate gigs there, not three sightings of one rate. A city cell that holds fewer still publishes, computed from exactly the same evidence, but it is labelled imputed: its figure leans on the pooled market rows rather than on local collection, and it says so in the API, in the open dataset and on this site. Region and UK cells are measured by definition — they are roll-ups of everything applicable to them, which is what a national baseline is.

Why the distinction exists. Since 12 August a city cell has been computed from its own observations plus every observation in the pool that carries no location at all — market rate cards, union minimums, published rate schedules — on the reasoning that a national price list is a price charged in that city too. That is right for the arithmetic and it stays. But it also means the raw count of rows behind a city cell can be large while the locally-collected count is zero, so the count alone cannot tell you whether you are reading a measurement or an estimate. The local floor is what tells you. Every city cell publishes both counts — local_sample_size and blanket_sample_size — in the open dataset (the is_imputed flag itself is on the live tables and the calculator; in the CSV, measured means local_sample_size of three or more), so you can always see which kind of cell you are reading and reconstruct the judgement yourself. Region and UK rows leave both columns empty: the split only means something where a pooled layer is being laid over local evidence. When you look up somewhere with no cell of its own, the live calculator walks the chain — city, then region, then the national figure with a regional adjustment — and tells you which level answered.

The floor only ever rises. Three is deliberately low while the Index is young: a cell published with its evidence attached and honestly labelled is more useful than no cell at all, and every figure carries its n so a reader can weigh it. The floor is a value in the database rather than a promise in prose, and the database refuses any attempt to lower it — so no figure can ever be re-labelled measured by relaxing the bar, and raising it costs no coverage at all: the same cells print, more of them carry the honest label. At present more than half of all measured cells rest on fewer than ten observations, so single-city figures are indicative; the national baselines carry the weight. § 11 covers the other kind of estimate — a cell a city has no observations for at all — and how it is flagged.

Regional adjustments

The adjustment layer blends ONS regional household income and family-spending data with public industry price benchmarks, tuned per gig type: corporate fees track regional wealth closely, wedding spend barely does, and pub fees hardly flex at all because venue economics set the ceiling everywhere. Adjustments only ever fill gaps — wherever real observations exist, they win outright.

Derived cells, and how they are marked

Coverage is uneven: a city may have plenty of wedding observations and none at all for a four-piece in a pub. Rather than leave a hole that makes cities incomparable, the Index fills that gap with a derived cell — and marks it as derived everywhere it appears, in the API, the dataset and on the page. The value is the national figure for that same gig type and band size, scaled by how that city prices against the national baseline on the cells it does have. We use the city’s own observed pay level rather than its neighbours’, because geography is not what sets gig fees — Brighton’s nearest large market is London, and borrowing London’s rates would simply make Brighton look dearer than it is.

Every city carries the same grid. An index is only an index if you can look up any town and find the same data points, so the grid is every gig type and band size the UK baseline can anchor — 23 cells, identical in all thirteen cities (seven gig types across four band sizes, minus the six combinations the UK baseline cannot anchor, plus an any-size wedding cell). It is not narrowed to whatever a city happens to have collected; that would make towns incomparable, which is the fault derived cells exist to fix.

Two rules keep it honest, and they are deliberately the ones that stop nonsense rather than the ones that limit coverage. A city needs three measured cells of its own before we will estimate how it prices at all — below that it gets no derived cells. And a measured cell more than three times, or less than a third of, its national counterpart is treated as unusable and replaced by the derived figure, with the original retained in the record. Those thresholds, and the local floor in § 9, only ever ratchet upward as coverage grows; the database refuses to loosen them to make a figure publishable.

Derived cells are never evidence. They are never counted in a sample size, never used to calculate another cell, and never written back into the observation record — the raw data stays purely observed. They are rebuilt from scratch on every refresh, so the moment observations arrive for that city and gig type, the derived figure disappears and the measured one takes its place.

Two kinds of estimate, one flag. A city cell is flagged is_imputed either because the city had no observations at all for that gig type and band — the derived cell described above, method ratio — or because it holds fewer than three locally-collected rows and therefore leans on the pooled market rows, method blanket_pool (§ 9). Both print; both are labelled; the method column tells you which you are reading. Anyone wanting measured figures only can filter the CSV on local_sample_size of three or more (the flag itself is exposed on the live tables and the calculator), so the judgement can be reconstructed from the raw dataset rather than taken on trust.

Refresh cadence

The cells rebuild nightly at 05:00 UTC. Industry rate cards re-scan on Mondays; the artist panel snapshots on the 28th; approved submissions flow daily. Every rebuild is archived rather than overwritten — that is what makes issue-on-issue comparisons possible. The full report ships monthly.