Measurement
Part of Fitness creator measurement in England: defining health metrics from the CMO activity guideline
Fitness creator benchmarks only mean something once you record what they compare
Research fitness creator benchmarks without false comparisons by recording geography, populations, definitions, methods, uncertainty and breaks.
What to take away
- Record the decision first, then the metric definition, then the sources. A benchmark you cannot re-derive is not evidence.
- Build the reference set from the creator's own matched releases: same format, length band, distribution, publishing window and disclosure status.
- External averages travel badly. Geography, surface, audience and metric definition all differ, and small loyal audiences inflate rates.
- Sport England's Active Lives Adult Survey covers population activity in England. It cannot benchmark views, creator income or campaign conversion.
- Where comparable evidence is absent, record "no defensible benchmark identified" and treat the gap as part of the decision.
Define the benchmark unit before you search
Write the decision first, and name its owner. A production team might compare approval completion. An editor might compare viewing behaviour for one format.
Then fix the metric fields, using an illustrative monthly view-duration benchmark.
Define the benchmark unit
- Population
- Eligible organic views of named long-form videos, England
- Numerator
- Watch minutes on those views
- Denominator
- Views counted in the same versioned export
- Unit and period
- Minutes per view, calendar month
- Source and owner
- Versioned analytics export, audience analyst
- Inclusions and exclusions
- Named organic format and England filter in; paid ads and test uploads out
Also record the account timezone, source uncertainty and the update trigger that ends comparability. A platform definition change or a filter change breaks the series, so mark the date.
YouTube's Content analytics record defines average view duration for watched content and gives no universal target for it.
Example: a comparability card and a five-release reference set
Complete a card for every candidate source, in this order.
Comparability card
- Record the publisher and the publication date.
- Note the geography and the population covered.
- Note the sampling frame and the method used.
- Add weighting, field dates and stated uncertainty.
- Recover the denominator, or reject the source.
A press release from a fitness brand claims 4.2 minutes per view, an illustrative figure. The card fails it: no population, no denominator, no field dates. Reject it and write down the reason.
Check any agency that supplies the comparison, as our provider desk reviews do for England fitness briefs.
Now build the reference set from matched releases.
Sorted, the five values run 2.6, 2.9, 3.0, 3.4 and 4.1. The median is 3.0 minutes per view. The spread is 1.5 minutes. Report the count of five beside the median.
Why cross-channel league tables mislead
YouTube's impressions and click-through-rate FAQ warns that smaller, loyal audiences can produce higher rates. A rate lifted from another channel therefore travels badly.
The Code of Practice for Statistics asks official producers to explain quality, limitations and uncertainty. A private creator report does not become an official statistic by copying that practice, but the same questions expose weak comparisons.
Campaign tagging matters too. A taxonomy change is an update trigger, not a reason to merge two unlike series. Our guide to creator attribution methods keeps one tagging rule intact across a full year.
Sport England activity data in context
Sport England's Active Lives Adult Survey methodology draws a random sample of English households, runs continuously and weights results to population estimates by geography and demographics. The Active Lives technical note defines activity through moderate-intensity-equivalent minutes.
Those records suit a matched England population question. They cannot benchmark views, creator income, campaign conversion or clinical efficacy. A shift in a population survey cannot be credited to one publisher. State the age band, collection window, activity definition and survey uncertainty.
Keep health outcomes out of performance work
NICE public health guidance PH49 recommends setting objectives and valid process and outcome measures before behaviour-change work starts. The Magenta Book says evaluation design should follow the question and explain limitations.
Interpret health results within the approved study population, comparator, measure and follow-up. Never pool testimonials or before-and-after posts into a success rate. Our guide to health outcomes outside performance benchmarking explains why outcome measures sit outside the return calculation.
Before you act
- Write the decision before you search for a benchmark.
- Record the metric definition with numerator and denominator.
- Complete a comparability card for each candidate source.
- Freeze the inclusion list before calculating the reference set.
- Say "no defensible benchmark identified" when comparable evidence is absent.
Common questions
What fields must a comparability card carry?
Publisher, date, geography and population come first. Sampling frame, method, weighting and field dates follow. Numerator, denominator, unit and uncertainty close the record. If the denominator cannot be recovered, reject the source.
Why is an external average a weak comparison for a fitness creator?
Geography, surface, audience and metric definition rarely match, and YouTube warns that smaller loyal audiences can produce higher rates. Start with the creator's own matched releases.
What can Sport England's Active Lives data legitimately be used for?
They give external context for a carefully matched England population question, such as adult activity levels by age band and region. They cannot benchmark views, creator income, conversion or clinical efficacy.
How should health outcomes be handled in benchmarking?
Keep them outside performance benchmarking and interpret them within the approved study population, comparator, measure and follow-up.
