Benchmarking your metrics against sector norms

The slide comes back from strategy with one line on it: "monthly churn 4.4% vs sector norm 2.9%." Forty minutes of the leadership meeting then go on why the product is twice as leaky as the market. Nobody asks which services sit inside that 2.9%, whether it counts trialists, whether it is a monthly rate at all, or who paid for the study. The comparison layer is where most benchmarking effort should go, and it is mostly provenance and arithmetic rather than data gathering.

Which datasets are worth citing

Rough order of how much weight a number can carry in front of a board:

• Company disclosures. Netflix's quarterly letters give revenue per membership by region; Disney's filings give Disney+ Core subscribers and average monthly revenue per paid subscriber, with domestic and international shown apart. Both are defined in footnotes and tied to audited accounts. Neither publishes churn, and neither publishes acquisition costacquisition costCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition →.
• Regulators. Ofcom's Media Nations, published annually, pairs household survey work with industry data for the UK. The method is public, nobody is selling anything, and it describes the market rather than one company. It lags by months and stops at national aggregates.
• Measurement panels. Nielsen's The Gauge reports monthly share of US TV viewing time by platform, with streaming above 40% of that time in recent years. It measures attention, not subscriptions: good context for demand, worthless as a churn comparator.
• Commercial subscription trackers such as Antenna or Ampere Analysis. Both sell the data under discussion, so the charts they release free are the ones they most want quoted. They are also the only practical source for the split that matters most in streaming: churn on a "primary" service people organise their week around, commonly estimated in the low single digits monthly, versus a "flanker" bought for one show, often mid single digits or worse.
• Trade-press "industry averages" with no linked method. Fine in a pitch, not in a plan.

A four-part test before any external figure enters a deck: name the population, the period, the denominator, and who funded it. If one of the four is missing, the number can inform a discussion but must not become a target anyone is paid against.

Normalising before you compare

Almost every published media benchmark needs restating before it touches your own numbers.

Period. Monthly and annual churn do not scale linearly. 3% monthly compounds to 1 - 0.97^12 = 30.6% a year, not 36%. Someone dividing an annual 30% by twelve gets 2.5% and has quietly moved the goalpost by half a point.

Denominator. Your internal base may include people on a free trial; a tracker built on card or receipt data usually sees only paying accounts. Strip trialists out before comparing, or your churn looks worse than a peer measured on paid accounts alone.

Realised revenue against shelf price. Subscribers who sign up inside an app store arrive net of a platform fee of 15 to 30%, and UK and EU prices are quoted with VAT while US prices are not. If 40% of your base pays through in-app purchase at 15%, the shelf price overstates realised revenue per user by roughly six points before any promotional discount.

Margin basis. Value-per-subscriber comparisons only hold if both sides used margin, as the lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → lesson builds it, rather than gross revenue. The same discipline applies to acquisition cost: comparable only once platform fees and owned-channel costs are allocated the same way on both sides, which is the allocation the CAC lesson works through.

# Restating a published benchmark onto your own definitions: illustrative only
def normalise(monthly_churn, iap_share=0.0, iap_fee=0.15):
    annual_churn = 1 - (1 - monthly_churn) ** 12
    arpu_haircut = iap_share * iap_fee
    return round(annual_churn, 3), round(arpu_haircut, 3)

normalise(0.03, iap_share=0.40)   # -> (0.306, 0.06)

Comparability across markets and models

Netflix reports revenue per membership by region because the US and Canada figure runs far above Asia-Pacific. Disney reports Hotstar apart from Disney+ Core because Hotstar's monthly revenue per subscriber is under a dollar against roughly eight dollars in the domestic market. A single global "streaming ARPU norm" built by averaging those together describes no business that exists.

Two further splits break naive comparison. Ad-supported tiers mix subscription and advertising revenue and behave differently on retention, so a blended norm hides which side of the mix is moving. Wholesale subscribers acquired through a telco or device bundle carry close to zero direct marketing cost and renew on someone else's billing relationship. If a third of your base arrives through a bundle and the benchmark was built on direct-to-consumer signups, the two acquisition-cost figures are not measuring the same activity at all.

Where sector norms mislead a leadership team

• Survivorship. Services with poor retention shut down or fold into a parent, and then drop out of the panel. The published average improves without a single operator improving.
• Composition. Netflix passed 300 million paid memberships at the end of 2024; its marginal acquisition cost and word-of-mouth base resemble nothing in year two of a service. An average that includes it flatters no one and instructs nobody.
• Broken time series. Netflix stopped reporting quarterly membership counts from the start of 2025 and pushed attention to revenue and margin instead. Any benchmark keyed to subscriber numbers lost its main reference point, and multi-year charts that look continuous are usually stitched across changed definitions.
• Benchmark as target. Make the sector churn norm a bonus metric and you will hit it, by cutting the cheap high-churn channels that were also feeding volume. Churn improves, net adds fall, and next year's plan sits on a smaller base.
• Norms as permission. "We are at the sector average" ends the analysis exactly where it should begin, since the average contains the businesses currently being outgrown.

A quick provenance check

When there is no usable norm, build one

For most media businesses the better comparator sits in the house. Your own trailing cohorts, held to one definition for eight or twelve quarters, answer the question a sector average cannot: are we improving, and in which market and which channel. That series is yours, consistent, and immune to a vendor changing its panel.

Around it, three layers. Ofcom and Nielsen for market direction, which tells you whether a churn rise is yours or everyone's. A matched pair or trio of services with genuinely similar structure (same region, same price band, same mix of ads and subscription), read from their own disclosures line by line rather than through a blended index. And an explicit confidence label on every external figure in the deck: source, period, and estimate or actual. A CFO who sees "Antenna panel estimate, US, Q3, paid accounts only" argues about the right thing.

🎬 [VIDEO: "Netflix's Business Model Explained" - https://www.youtube.com/results?search_query=netflix+business+model+explained - a good primer on how a scaled streamer's subscriber economics and churn dynamics work in practice, useful context before benchmarking your own numbers]

Key Takeaways

• Rank sources before you quote them: audited disclosures from Netflix or Disney, then regulator work such as Ofcom's Media Nations, then panels like Nielsen or the commercial trackers who sell the data, then trade-press averages with no method attached.
• Normalise on four axes before comparing anything: period (3% monthly is 30.6% annually, not 36%), denominator, realised revenue after VAT and platform fees, and margin basis.
• Regional and model mix make blended norms meaningless. Netflix's US and Canada revenue per membership against Asia-Pacific, or Hotstar's sub-dollar figure against a domestic Disney+ subscriber, are the reason to compare like with like.
• Survivorship and definition changes move published averages on their own. Netflix ending quarterly membership disclosure in 2025 broke the market's most-cited series.
• A benchmark turned into a bonus target changes behaviour, not performance. The cheapest way to hit a churn norm is to stop buying the volume that was carrying growth.