Analysis

How we estimate last season's commercial revenue

The squad cost rule counts a club's commercial revenue for last season, and nineteen of the twenty clubs have not filed those accounts yet. Until now each of our figures was a judgement. This is the method that replaces it: what it is fitted on, how it was tested against two alternatives, how wide its range is and why, where we overrode our own rule, and what it cannot know.

By Arthur van de Laak9 October 2026Updated 25 October 2026 · View updates21 min read

Figures as of 9 October 2026, Orange Engine release V2BV · Latest figures · What the engine does not capture

The squad cost ratio compares what a squad costs with the revenue the rule counts. One line of that revenue is commercial: sponsorship, advertising, shirts, catering. The rule takes it from last season’s accounts, and for the test of 1 March 2027 last season is 2025/26. Those accounts are filed between late 2026 and spring 2027. One club’s figure, Manchester United’s, is filed. For the other nineteen the figure in our model is an estimate.

Until 9 October 2026 each of those nineteen estimates was a judgement of ours. This article sets out the method that replaces them, the evidence it rests on, the test it had to pass, how wide its range is, and what it cannot know.

What the rule counts

A club’s commercial revenue under the rule is the higher of two figures: its commercial revenue for last season, and £15m. For a promoted club there is a third: its average over the last three seasons, where it has been in the Premier League in at least one of them. The highest counts. (Appendix 2 of the Premier League Handbook, Rule A.1.29(c).) “Last season” means the twelve-month accounting period that covers it. (Rule A.1.72(a).)

The rule also says what commercial revenue is, and it is not one line of the accounts. Sponsorship and advertising, food and drink sold at the men’s first-team matches, direct-to-consumer business and the women’s team count at gross revenue. Merchandise counts at gross profit: sales less the direct cost of the goods. Stadium hire on non-match days, hotels, conference centres, other football teams and other operating income count at net profit. (Rule A.1.19.) A club that has outsourced its shop is treated as if it had not. (Rule A.1.24(b).)

So the estimate has two parts. This article is about the first, the commercial figure itself. The cost of merchandise that comes off it is the subject of an earlier article.

What we did before

For each club we took the last filed commercial figure and raised it by our own view of the club’s season: a new shirt sponsor, a promotion, a year in Europe. Most of those figures had no recorded parts, so nobody could check how we had reached them. Each was then allowed to vary by 7% either way, the same margin for every club.

Neither the judgement nor the 7% rested on evidence. Two of the figures, Chelsea’s and Manchester City’s, came with a story about the season that no document supported. Those were judgements of ours and are withdrawn.

The method

The estimate is the club’s last filed commercial revenue, grown by a rate that depends only on how large that revenue is.

The rate falls as the base grows: the larger a club’s commercial revenue, the more slowly it has grown in percentage terms.

The formula is: rate = 62.52% less 4.366 percentage points for each unit of the natural logarithm of the base in thousands of pounds. It is fitted by a method that one unusual year cannot pull about (the median of the slopes between every pair of observations), and it is frozen. It is not chosen or altered on the ratios it produces.

The base is the commercial lines printed in the club’s 2024/25 accounts, read on the page. Where sponsorship and other commercial income are printed apart, they are added. A merchandise, retail or catering line printed apart is part of the base at its sales value, and the cost of the goods comes off afterwards. Aston Villa’s accounts put player loan fees inside a commercial line, and they are taken out. A line named for matchday, hospitality, broadcasting, property, player trading or grants is never in the base.

The method knows nothing about contracts. A sponsor gained or lost in 2025/26 is outside it. A press report of what a sponsorship is worth is never an input, and neither is a league position, a trophy or a new stadium.

The evidence

The rate is fitted on 81 year-on-year changes in commercial revenue, at 17 clubs that stayed in the Premier League, over eight seasons: 40 changes into the seasons 2014/15 to 2018/19, and 41 into 2022/23, 2023/24 and 2024/25. Commercial revenue fell in 15 of the 81.

Every figure was read on the page of a filed set of accounts. A figure read only by machine is never used.

• The seasons since 2019/20. The 55 figures we had previously read by machine were read on their page images by two readers working independently. The two agree with each other on all 55. One figure was corrected, by £1,000. Both readers also found three accounting periods that were not twelve months long, which we had recorded as twelve.
• The earlier seasons come from 147 sets of accounts. Every season was read on two different filed pages by different readers: once as the current year of its own accounts, and once as the comparative printed in the following year’s. Of 557 printed lines read twice, 539 agree. The 18 that differ are cases where the later accounts re-presented the earlier year.
• Each of those earlier changes is taken inside one document: this year’s column against last year’s column of the same accounts, so that both figures share a definition and a reporting entity. The limitation is that each such change rests on one document read by one reader, with each of its two figures cross-read in a neighbouring document by another.

What was left out, and why:

• The pandemic. Changes into 2019/20, 2020/21 and 2021/22 are out: 36 of them at clubs that stayed in the Premier League.
• Periods that are not twelve months. A thirteen-month year against an eleven-month one is not a growth rate. A comparative headed only by its year is used only where both periods are shown to be twelve months, by the accounts themselves or by the register’s record of the dates the accounts were made up to.
• Manchester United before 2019/20. The club company’s commercial line is £4.3m for 2016/17 and £267.9m for 2017/18. That is a change in which company booked the income, not in the business. United’s earlier years are out, on both companies.
• Aston Villa before 2019/20 (another reporting entity, and loan fees inside the commercial line) and Ipswich Town’s club company.
• Manchester United’s 2025/26, the one figure filed since. It is kept out of the fit and used as a test.

Three methods, tested before one was chosen

We compared three ways of making the estimate:

• no growth: last year’s figure;
• size only: the rate above;
• half and half: half the club’s own last change, half the size rate.

Each was tested on seasons it had not seen. The error is the forecast over the filed figure, less one.

(“Typical” is the median.) No growth is the weakest of the three overall, and it is too low by a median of 12%. It was the best of the three in two flat seasons, 2015/16 and 2018/19. It was also the best on the one figure filed since: Manchester United’s commercial revenue fell 4.8% in 2025/26, so it came in 4.8% below the no-growth forecast, 10.1% below size only and 11.8% below half and half.

Size only and half and half are within half a percentage point of each other on both tests, each ahead on one. We took size only: it performs about as well, it has one fewer moving part, and it does not carry one club’s unusual year forward into the next.

One caveat on the test itself. It was run on 78 changes, before we settled how to treat comparatives headed only by a year. The rate is fitted on 81. The test has not been rerun on the 81. With every such comparative added, 86 changes, the typical errors are 13.6%, 8.3% and 9.2%.

The rule we held ourselves to. The method was chosen on one thing: how well it would have forecast commercial revenue. It was not chosen on the squad cost ratios it produces, or on which side of 85% it leaves any club. The rate, the range and every club’s figures were fixed before the model was run on them.

How wide the range is

Set the method against the record and the misses are larger than 7%. Across 82 outcomes (the 81 changes and United’s 2025/26), the typical miss either way is 8.7%. One filed figure in ten came in more than 17.1% below the forecast, and one in ten more than 19.0% above. The lowest was 32.5% below and the highest 60.7% above. These are observed misses of past forecasts, 82 outcomes from 2013/14 to 2025/26. They set the size of the range. They do not show that future filed figures will fall inside it at any stated rate.

The misses are not the same for every club. They narrow as the base grows, and for small bases they run further on the upper side. So the range is a scale that moves with the base, fitted separately on each side.

Our model draws the commercial figure many times between a low end and a high end, most often near the estimate, and publishes the range between the 10th and 90th percentile of the results. The two ends are set to follow the one-in-ten marks in the table. It is an assumption range: it shows how far the figure moves between the assumed bounds, not a probability that the filed figure falls inside it. The two ends are the ends of a draw. They are not the lowest and highest values we think possible, and nothing in them is a judgement about a club.

Where we overrode our own rule

We compared two ways of setting the range: the continuous scale above, and three fixed bands (under £50m, £50m to £150m, and £150m and over). Each was tested on seasons it had not seen. The test asked how often past filed figures fell inside the interval, against an intended 80%. It is a back-test on the same 82 outcomes described above: it records how each scheme would have performed on that history, and does not establish that future figures will fall inside the range at the same rate.

Before running the comparison we wrote a rule into the tool, so that the choice could not follow the results. The continuous scale would be used only if its coverage was at least as close to 80% as the bands’, its interval score was no worse, and its average interval was no more than 5% wider. Otherwise the bands would be used.

The continuous scale passed the first two tests and failed the third. Its interval is 7.4% wider than the bands’. By our own rule the three bands had won, and the tool’s record of the comparison says so.

We overrode the rule on 8 October 2026 and use the continuous scale. The reasons: its coverage of past filed figures, 82.9%, is closer to the intended 80% than the bands’ 74.4%; the extra width is a fair price for that; and it removes two arbitrary steps, at £50m and £150m, where a club a pound either side would have been given a different range. The 5% limit was a number we had picked in advance, not a finding. The override was made on those measures alone, before the model was run on any club. The three-band result is kept as the benchmark the scale beat, and the choice will not be reopened on any club’s outcome.

One thing neither scheme does: United’s 2025/26 came in 11.0% below the published method’s forecast (the comparison above, fitted before the method was final, gives 10.1%), and the scale’s one-in-ten mark at United’s size is 9.3% below. It is outside the published range and inside the ends of the draw.

Clubs that changed division

The size rate is fitted on clubs that stayed in the Premier League. It says nothing about a club that has just come up or gone down, where commercial revenue moves by far more.

Promoted clubs. We have nine promotions read from filed accounts, outside the pandemic: commercial revenue rose by between 73.1% and 196.2% in the first Premier League season, and the middle of the nine is 127.4%. Sunderland’s last filed Championship figure, £14.4m, is inside the range of bases those nine started from, so Sunderland’s estimate is that figure raised by 127.4%: £32.7m, plus £0.9m for the women’s team. Its range runs from the smallest of the nine rises to the largest. It is a provisional promoted-class estimate, resting on nine observations, until Sunderland’s 2025/26 accounts are filed. Two promotions fell in pandemic seasons and are left out.

Leeds are the exception. Leeds came up in the same summer, but their Championship commercial revenue, £58.1m, is nearly four times the largest base among the nine. A rate observed on small bases is not applied to a base outside them. Leeds take the size rate for their base, 14.6%.

Relegated clubs. We have six relegations read the same way: commercial revenue fell by between 10.1% and 65.1%, and the middle of the six is a fall of 56.7%. Ipswich Town went down in 2025 and came straight back, so their last season was a Championship one. Their estimate is the £34.7m filed for the Premier League season of 2024/25, lowered by 56.7%: £15.0m. As a check, their last filed Championship year, 2023/24, was £15.3m. It is a provisional relegated-class estimate, resting on six observations, one of which (a fall of 10.1%) is far from the other five. For Ipswich the rule counts the three-year average, which is higher: £18.5m.

Coventry City and Hull City were in the Championship in 2025/26 and were not relegated into it, so neither rate applies. We hold their last filed figure with no growth, £9.8m and £6.1m, until their accounts are filed. Both are below £15m. The rule counts £15m for each, and that is what the model uses. The £15m is the rule’s minimum. It is not a forecast of either club’s commercial revenue, and we do not present it as one.

What is added after the growth

A few clubs have commercial revenue the rule counts that is not in the accounts we read for the base. Each is added after the base is grown, at its last filed value, and is not grown itself.

Newcastle’s fan zone rests on a reading of the rule. The central figure counts it at operating profit, £2.8m. The range runs from nothing to its gross revenue, £11.2m, across the readings of the rule. That makes Newcastle’s published range slightly different from the scale’s own range.

Leeds’s merchandise. Leeds handed their shop to a partner from 1 July 2025. The rule counts merchandise at gross profit either way, so the base keeps merchandising income at its sales value and the cost comes off, as at every other club. Leeds publish no cost figure. Ours is an estimate, not a filed figure: merchandising income of £24.6m at the 55% margin of the clubs that do publish one, which gives a cost of £11.1m, in a range of £8.6m to £16.0m. Until this release we had taken nothing off.

Manchester United have filed 2025/26: £317.3m for the group. There is nothing to estimate. The figure in our September forecast, £310.8m, stays as it was published and is scored against the filed one.

Each club’s estimate

Figures are in pounds, millions, rounded once from the thousands the model holds. “Drawn between” gives the two ends of the draw, not the published range.

What it did to each club’s ratio

Seventeen clubs’ figures move: ten down, six up, and Bournemouth’s by less than the precision shown. Coventry, Hull and Manchester United do not. No club changes side of 85% or 115% on its central figure. Most ranges are wider than they were, because the range of the commercial figure is wider than 7%.

Two figures before player trading now sit a fraction of a percentage point from a line: Manchester City’s is 84.9%, only 0.1 percentage points below 85%, where it was 87.1%, and Chelsea’s is 114.4%, 0.6 percentage points below 115%, where it was 115.5%. Both margins are smaller than the uncertainty in the commercial estimate, so neither tells reliably which side the figure is on. These figures have since crossed their lines: Manchester City’s is 85.6% and Chelsea’s 115.1% (release V2CB, 23 October 2026); see the Manchester City rebuild and the Chelsea rebuild.

Four clubs’ assumption ranges span the 85% line. For each, the estimate does not reliably tell which side of the line the club is on:

• Bournemouth: central figure 85.5%, assumption range 79.5 to 87.8%.
• Brentford: central figure 83.8%, assumption range 79.5 to 87.5%.
• Newcastle: central figure 84.7%, assumption range 78.9 to 88.6%.
• Tottenham: central figure 80.4%, assumption range 76.4 to 85.6%.

Next season is not covered

This method estimates one figure: commercial revenue for 2025/26, the season the March 2027 test counts. It is not a forecast of 2026/27.

Several club articles carry tables that look ahead to the tests of 2028 and later. Those tables hold this season’s revenue where it is and roll only the profit on player sales forward. They are arithmetic on this season’s figures, and they are now labelled that way: next season’s commercial revenue is not forecast in them. Where one of those tables lands within a percentage point or two of a threshold, the estimate cannot tell which side, and we no longer say it can.

The model also carries its own projection of next season for each club. That projection still rests on an earlier commercial assumption, made club by club, and this release does not change it. Moving the forward layer onto a published method is an open item for the next release.

What the method cannot know

• It uses pounds of the day. The rate is fitted on figures from 2013/14 to 2024/25 with no adjustment for prices.
• Growth has been faster since the pandemic. The typical change is 12.5% in the 40 earlier changes and 16.0% in the 41 since. The fitted rate sits between the two periods: the filed figures of the earlier years came in a median 4.0% below its forecasts, and those of the later years 3.6% above. Which of the two 2025/26 resembles is not known.
• Companies and definitions change. United before 2019/20, Villa before 2019/20, Ipswich (the present group was dormant in 2021), Coventry (no breakdown of turnover before its 2020 accounts) and Chelsea in 2024/25 (the women’s company was sold) are breaks we excluded or worked round. Definitions also drift inside the period: Arsenal’s retail was folded into commercial, Brighton’s nine lines became four, Palace re-presented hospitality. Each change is taken inside one document so that its two figures share a definition.
• United’s series changes company. After 2019/20 it holds the club company to 2023/24 and the group for 2024/25 and 2025/26. No change is taken across the two.
• The one figure filed since fell outside the published range, as described above.
• One reader for each earlier document, as described above.
• The page images are not published. The readings name the document, its checksum and the page.
• The three-method test was run on 78 changes; the rate is fitted on 81.
• The estimates for promoted and relegated clubs rest on nine and six observations and are provisional.
• Other operating income is counted in full where the rule counts only the profit on it, because no club discloses that profit.
• Merchandise cost is at 2024/25 values while the commercial figure is grown to 2025/26.
• It knows nothing of contracts, as described above.
• The stress case does not test revenue. The stress case printed beside each club’s ratio is the worst case on costs alone: every uncertain cost input is set against the club at once, with revenue held at the central estimate. It is not an upper bound of what the simulation can produce, and with the wider commercial range a few simulated results at some clubs are now above it. A joint downside on revenue and costs is a later proposal.
• The additions are at their last filed values, and where they move the two ends of the range differently (Newcastle) the published range is no longer exactly the scale’s own range.

Documented does not mean right. Every estimate here can be rebuilt from filed figures and the published rate. That makes it checkable. It does not make it correct.

The forecasts already published

The forecasts of record on our track record page are the figures we froze and had independently timestamped on 30 September 2026. This release does not replace them. They will be scored against the 2025/26 accounts as they were published, with the commercial estimates they were built on.

The figures of this release are a later estimate, held in the release’s own timestamped record. The track record page shows each club’s forecast of record and its current estimate side by side; it does not yet score a later set of figures separately. When a club’s accounts are filed we will report how far out its commercial estimate was, for the figure published in September and for the figure this method gives.

When this stops mattering

Each club’s 2025/26 accounts replace its estimate with the filed figure, and we will report how far each estimate was out. When the Premier League clubs’ accounts are all in, in spring 2027, the rate and the range will be refitted on the longer record as a new version of the method. They are refitted for a release, not between releases, and never on a club’s ratio.