Measuring gaming creator work with a metric dictionary and attribution kept apart from cause

Start a report with the decision it must support. A gaming channel may need to decide whether to repeat a stream format, renew a sponsor, change a thumbnail, commission an editor or stop a costly series. Each question needs its own measure, comparison and evidence. A page of unrelated totals cannot answer all five.

This guide is for creator businesses based in England. Platform documentation is global, Ofcom research is UK-wide and privacy rules cited here apply across the UK. Sources were reopened on 5 September 2026. The method is desk research and does not claim access to a working analytics account.

What to take away

Start with the decision the report must support and write a measurable question with action, population and time window.
Build a metric dictionary that records plain name, platform label, definition, unit, scope, date range, extraction date, source owner and known limitation.
Keep attribution separate from causation and name the attribution model and event rather than presenting credited actions as observed causes.
Control data capture with an extraction log, a content key that survives title edits and a campaign naming convention before launch.
Build a dashboard that shows the reporting question, period, included content and data freshness, with a separate data-quality panel.

Write the measurement question first

Use a sentence with an action, population and time window. "Did the first three ranked-match analysis videos hold returning viewers better than the previous three during their first seven days?" is measurable. "How did gaming content perform?" leaves the format, audience, comparison and decision undefined.

Choose one primary result and a few diagnostic measures. For a live sponsor activation, the result might be verified visits to an agreed landing page during a stated attribution window. Peak concurrent viewers, watch time near the segment and disclosure evidence can explain delivery, but they are not interchangeable with visits or sales.

Record the decision rule before seeing the result. State what would lead to repeat, revise, investigate or stop. A threshold may come from a contract, a budget model or the channel's matched history. Do not invent an industry average to make the decision appear objective.

Build a metric dictionary

Every field needs a plain name, platform label, definition, unit, scope, date range, extraction date, source owner and known limitation. Keep the platform wording. If a label changes, preserve the old definition beside historical data rather than silently joining unlike series.

Distribution and discovery

Thumbnail impressions record registered appearances of a thumbnail on YouTube, while impressions click-through rate describes how often an impression led to a view. YouTube's Reach documentation also separates internal traffic sources, external sources, direct or unknown traffic and search terms. An impression is not every possible exposure across the internet.

Use reach measures to diagnose packaging and discovery. Compare impressions and click-through rate together, alongside source mix and the video's age. A lower rate after distribution broadens can coexist with more viewers, so a single percentage should not trigger an automatic thumbnail change.

Consumption and attention

Views describe starts that meet the platform's rules. Watch time sums viewing duration, while average view duration divides viewing time across relevant views. Audience retention shows where viewers stayed, skipped or rewatched within one video.

YouTube says the retention report is available at video level and can take one to two days to process. Mark the extraction delay. A sponsor report captured immediately after upload may be incomplete even when the figure has a precise decimal.

For a gaming guide, examine retention around the promised answer, setup, match start and sponsor passage. A dip is a clue for editorial review, not proof that a particular sentence caused someone to leave.

Active audience

Subscribers are an account total, not the number of people currently watching. YouTube defines monthly audience as unique viewers over a rolling 28-day window and splits that audience into new, casual and regular viewers in its Audience tab guide. The page also warns that some geography and demographic data may be limited.

Use unique or returning viewers when the question concerns current audience breadth or loyalty. Preserve the selected period because a rolling monthly audience does not become a calendar-month measure merely because a report is labelled August.

Live participation and delivery

Live reporting needs both audience and technical fields. YouTube lists concurrent viewers, peak concurrents, chat rate, views, duration and average view duration in its live metrics record. Stream health is a separate operational signal.

A busy chat may reflect community activity, spam or a contentious moment. Pair the rate with moderation incidents and the run sheet. Peak concurrents describe one maximum point; they do not say how many individuals watched or how long they remained.

Money and commercial contribution

Separate platform estimates, final platform earnings, sponsor fees, affiliate receipts, sales, refunds and non-cash benefits. Use pounds consistently, identify whether VAT is included and distinguish invoiced amounts from cash received. A campaign report should state which components it includes.

YouTube's revenue analytics explanation distinguishes creator RPM from advertiser CPM and says RPM excludes most sponsorships and indirect revenue. It also notes that the rate can fall when unmonetised views rise even if revenue remains unchanged.

HMRC's online platform income guidance covers money, gifts and services received through online activity. Reporting a campaign to a client does not replace business records or professional tax advice.

Control the data capture

Create an extraction log with the article or stream ID, platform, report name, filters, dates, time zone, export time, file location and operator. Save raw exports separately from calculations. Never paste a revised number over the source value without a note.

Use a content key that survives title edits. Add game, format, live or recorded status, duration, sponsor, publication time and planned audience. These descriptive fields let the analyst compare similar work without inferring categories later from filenames.

For links leaving a stream or description, adopt a campaign naming convention before launch. Google Analytics documents manual parameters for source, medium, campaign, content and related dimensions in its traffic-source guidance. Mixed spellings and recycled campaign names make later grouping unreliable.

Keep collection proportionate. The ICO's final storage and access technologies guidance covers cookies, tags, link decoration, consent, exceptions and advertising measurement. Have a qualified reviewer assess the actual implementation, especially when combining platform, site, shop or sponsor data.

Build a dashboard that exposes scope

The first screen should show the reporting question, period, included content and data freshness. Put the primary outcome next, followed by only the diagnostics needed to act. A separate data-quality panel should show missing exports, changed definitions, late adjustments and unresolved joins.

Do not mix channel totals with video rows or live snapshots with archive performance. Use separate views for a single release, a matched content group, a sponsor campaign and the business month. Where the same measure appears twice, keep its unit and window identical.

Give every chart a source and extraction note. A line without an origin looks authoritative but cannot be checked. A small team can meet this standard in a spreadsheet; visual complexity is not evidence quality.

Choose attribution evidence deliberately

Platform analytics can show that viewers watched and where traffic appeared to originate inside that service. Tagged links can connect a visit to campaign dimensions. Codes can connect an order to an entered value. A post-purchase question can capture remembered influence. Each sees a different part of the path.

Google Analytics defines attribution as assigning credit to touchpoints before an important action and documents data-driven, paid and organic last-click, and Google paid-channels last-click models in its attribution guide. Name the model and event rather than presenting attributed credit as a directly observed cause.

Reconcile results instead of selecting the highest. For each method, state eligible actions, lookback window, duplicates, returns, missing consent, cross-device gaps and any modelling. The final report can show a range or separate columns when methods answer different questions.

Keep attribution separate from causation

A sale credited to a link occurred after a recorded touchpoint under a chosen rule. It does not prove the sale would not have occurred without the creator. Seasonality, a game launch, another campaign, stock availability or existing demand may also move the result.

The UK government's 2026 Magenta Book explains that stronger causal analysis uses a counterfactual, such as a suitable comparison or control, subject to design assumptions. Most routine creator reports cannot meet that standard. Use "associated with" or "attributed under this rule" when that is what the evidence supports.

If a decision warrants a test, choose comparable content or audiences in advance and change one material feature. Protect viewers and commercial commitments. Small samples, spillover between groups and simultaneous changes should appear in the limitations rather than disappearing from the summary.

Use benchmarks in the right order

Start with the same video's planned target, then its performance at a fixed age, then a matched group from the same channel. YouTube's Advanced Mode tips support comparisons across videos, groups and equal 24-hour, seven-day or 28-day lifespans. Equal age matters for games tied to launch news or patches.

Platform typical retention is useful for a narrow attention comparison. YouTube says it uses the channel's ten latest videos of similar length for typical retention. It is not an England gaming-creator revenue standard.

Use regulator research only for the population it measures. Ofcom's Media Nations 2026 describes UK audience behaviour across television, online video, radio and audio. It provides market context, not a target for one channel's click-through rate or sponsor return.

Write the report for a decision

Open with the decision, finding and confidence. Follow with the defined measure, comparison, material diagnostics and what changed. Put source notes and detailed calculations where the reader can inspect them without interrupting the conclusion.

Label estimates, modelled fields, missing values and provisional revenue. Show absolute numbers beside rates when sample size changes the interpretation. If a sponsor supplied sales data, identify that source and the matching rules rather than implying the creator independently verified it.

End with a specific action and review date. A result may justify repeating the format, testing a different opening, correcting tracking, requesting missing merchant data or making no change until the sample matures. "Keep monitoring" is not enough unless the report says what will be checked and when.

A repeatable monthly close

Freeze the period, export raw sources and reconcile content IDs. Refresh the dashboard from controlled files, investigate breaks, then write the decision summary. A second person should check the metric definitions, arithmetic, evidence links and strength of language.

Archive the report with its extracts, code or formulas, campaign dictionary and approval notes. On the next close, changes should be visible rather than overwritten. The result is a smaller report that a creator, sponsor or adviser can reproduce and challenge.

Before you act

Write the measurement question with action, population and time window.
Record the decision rule before seeing the result.
Create an extraction log for every data pull.
Use a content key that survives title edits.
Adopt a campaign naming convention before launch.
Name the attribution model and event in reports.

Common questions

What should a metric dictionary include for each field?

How should live participation metrics be interpreted?

Live reporting needs both audience and technical fields. A busy chat may reflect community activity, spam or a contentious moment. Pair the rate with moderation incidents and the run sheet. Peak concurrents describe one maximum point; they do not say how many individuals watched or how long they remained.

Why keep attribution separate from causation?

A sale credited to a link occurred after a recorded touchpoint, but that does not prove the link caused the sale. Name the model and event rather than presenting attributed credit as a directly observed cause. Reconcile results instead of selecting the highest.