iGaming Metric Dictionary

The problem it solves

Every operator eventually holds the same meeting: two teams present the same metric with different values, both correct by their own definition, neither definition written down. The meeting resolves the number and not the cause, so it repeats monthly. The dictionary is the artefact that makes the cause fixable: a single place where each metric's computation is stated precisely enough that two independent analysts produce the same value.

This page describes how to build one. The reader-facing companion — what each term means and why it matters — lives in the iGaming glossary; the two layers should share names and definitions, with the dictionary as the source of truth.

Anatomy of a definition

A dictionary entry that ends arguments carries more than a formula:

Field	What it pins down	The dispute it prevents
Formula	Numerator and denominator, stated on named datasets	"My GGR includes voided bets, yours doesn't"
Grain	Per player, per brand, per market, per day — and the aggregation rule upward	Averages of averages presented as the metric
Time base	Timezone, cut-off, and whether the metric is event-time or settlement-time	The daily number that changes depending on who runs it
Population filters	Test accounts, bots, fraud, self-excluded players — in or out, explicitly	Funnel rates that move when the exclusion list does
Restatement rule	How late-arriving data changes a closed period, and whether history is frozen or re-stated	"Last month's number changed and nobody knows why"
Owner and effective date	Who answers for the definition, since when	Archaeology instead of accountability
Known variants	The regulatory base, the finance base, the product base — with the bridge between them	The filed number versus the board number

The metrics that earn their place first

Build outward from money and disputes, not inward from ambition. The core set for most operations:

The revenue bridge — turnover, GGR, NGR and every deduction between them. This is the family where two coexisting definitions cost the most; the decomposition discipline is covered in NGR drivers.
The acquisition funnel — click-to-registration, reg-to-deposit, FTD — where attribution windows and event definitions decide the value more than behaviour does.
Retention and value — cohort retention, churn, LTV — the metrics with the most silent assumptions (windows, censoring, revenue base), unpacked further in cohort analysis.
Payment health — approval rate, payout timing — where the definition must name the attempt-deduplication rule or the metric is fiction.
Bonus economics — granted versus consumed versus cost, the distinction that bonus economics treats as the first fix.

Governance: the part that keeps it alive

Dictionaries die two deaths: they drift from the systems that compute the metrics, or they freeze while the business moves. Both are governance failures with structural fixes:

Definitions live next to the code that implements them — generated into the readable document, never hand-maintained in parallel. The same principle as the configuration matrix: the document is a view, not a source.
Changes are versioned with effective dates. A definition change is an event in the data platform's history; every chart crossing that date should be able to say so.
Disputes route to the dictionary, not around it. The operational habit that matters most: when two numbers disagree, the first question is "which definition", and the answer gets written back into the entry — the dispute becomes an improvement instead of a recurring meeting.
New metrics need a sponsor, a formula and a reason. The bar is low but real; it keeps the dictionary a contract rather than a wish list.

The test

Hand the dictionary entry and access to the warehouse to an analyst who has never computed the metric, and compare their result to production. If they match, the entry is real. If they need a conversation to match, the missing content of that conversation is exactly what belongs in the entry. Run this test on the revenue bridge first — it is the family most likely to fail it and the most expensive one to leave ambiguous.

Continue reading: The event schema — the layer every definition ultimately resolves to. The iGaming glossary — the reader-facing companion of this dictionary.

Metric dictionary — common questions

What is a metric dictionary and how is it different from a glossary?

A glossary explains what a term means to a reader; a dictionary tells a system and an analyst how to compute it — formula, grain, filters, source dataset, owner, effective date. Our iGaming glossary is the reader-facing layer; this page is about the operational layer underneath it. An operation needs both, and it needs them to agree.

Why do two dashboards show different GGR for the same day?

Almost always one of four causes: a different timezone or accounting cut-off, a different treatment of open or voided bets, late-arriving settlements re-stating the day, or one dashboard silently including a product or brand the other excludes. None of these is a bug in the dashboard — each is a missing line in the definition. Fix it in the dictionary and both dashboards inherit the fix; fix it in a dashboard and you have created a third definition.

Who should own the metric dictionary?

One named owner for the dictionary as an artefact — typically the head of data or BI — and one named owner per metric, who is accountable for its definition being current and correct. Ownership by committee produces the exact ambiguity the dictionary exists to remove. Changes go through review with an effective date, like code, because a silent definition change re-states every historical chart that uses it.

How many metrics belong in the dictionary?

Fewer than the dashboards suggest. Start with the ones that appear in money conversations — the revenue bridge, the acquisition funnel, retention and cohort measures, payment health — and add a metric only when two people have actually disagreed about it. A dictionary of forty contested definitions gets used; a dictionary of four hundred aspirational ones decays.

Should regulatory reporting use the same definitions?

The same dictionary, not necessarily the same values — the regulator's base may legitimately differ from the management view (tax bases and jurisdictional rules differ). The dictionary should record both variants explicitly, with the bridge between them, so 'why does the number we filed differ from the board deck' is answered by a documented mapping rather than an investigation. See regulatory reporting.

The metric dictionary.