Place commercial KPIs by action horizon
Choose CRM, data platform or warehouse from latency and actionability—not from one universal source-of-truth claim.
Independently created. Contains no employer or client implementation detail, internal names or figures.
01Context
Sellers need signals while they work an account, managers need daily priorities across accounts, and finance needs month-end figures that reproduce exactly. All three come from the same orders, invoices and targets. When each team computes its own version wherever it happens to work, one KPI name carries three time windows and nobody can say which number is right.
02Decision drivers
- D01
How soon after the event someone must act on the signal
- D02
Whether the KPI needs data from more than one system
- D03
Whether a past value must reproduce exactly
- D04
One owner per definition, wherever it is computed
03Options considered
| Option | Right when | Cost / trade-off | Verdict |
|---|---|---|---|
| Compute every KPI in the warehouse and surface it in the CRM | Reporting-led organizations where nobody acts on a signal within the day | Signals arrive after the moment to act; the CRM shows yesterday’s number as if it were current | Rejected |
| Compute every KPI in the CRM | One system, record-level signals and little history | Cross-system and historical measures become fragile formulas; past values cannot be reproduced | Rejected |
| Place each KPI by the horizon of the action it drives, under one semantic contract | Signals, management priorities and certified history drawn from the same facts | Three places of computation, a definition contract and an owner per KPI | Selected |
Every option is right under some premises. The verdict applies to the context above.
04Decision
Classify each KPI by the horizon of the action it drives. Interaction-local signals (for example, days since the last contact) are computed in the CRM; near-real-time priorities across accounts in the data platform; certified historical measures in the warehouse. Every KPI is published through one semantic contract—name, definition, grain, time window and owner—and the CRM receives results with their explanation, not the calculation.
05Consequences
What this choice changes elsewhere in the system.
- Semantics
A KPI name cannot silently stand for different time windows: the contract states which one.
- CRM design
The CRM stores results and reasons, not analytical logic it cannot test.
- Reporting
Finance-grade reconstruction stays in the warehouse, where past values reproduce.
- Ownership
Each definition has one owner, even when the KPI is computed in more than one place.
- Operations
Prioritization depends on the data platform’s freshness, which becomes a monitored service level.
06Revisit when
The decision holds while its context holds. Challenge it when:
One operational store can compute cross-system signals at interaction speed, with versioned history.
Sellers start needing cross-system history inside the interaction—for example, margin or credit at quote time.
Definition ownership moves to one analytics team that also owns the CRM signals.
07Where this decision is applied
The decision says why this path was chosen; each case shows how it takes part in a larger architecture.
- Data case / 01Designing a governed commercial performance data modelA governed commercial performance model: orders, invoices, targets and customer context conformed at a declared grain, with explicit time, version, correction and currency semantics—so every metric is computed once and reproduces the same business meaning everywhere.
- Data case / 02Designing embedded commercial analytics around management decisionsAn embedded decision layer for commercial managers: organised by management question rather than by system object, with one filter semantics, governed metric definitions, a path from overview to the record where action happens, measured usage—and no analytical truth of its own.
- Data case / 03Operationalizing analytical signals back into CRMAn activation contract for analytical signals: a small set of governed, read-only values and segments at account grain, with as-of date, version, provenance and an explicit exit—published as the full expected state, applied idempotently and reconciled against what the CRM actually holds.
- RevOps case / 05Designing forecast governance and pipeline coverage that trigger actionForecast categories from stage semantics, overrides recorded at the forecast level, one owner of the final number, weekly snapshots—and a coverage contract whose threshold starts pipeline generation.
- RevOps case / 06Running commercial management cadence from the CRMFour forums—daily, weekly, monthly, quarterly—each with participants, evidence, decisions and outputs defined, run from system views instead of exported spreadsheets, and measured by adoption signals that show whether the routine is real.
- AI case / 01Designing evidence-grounded AI preparation for commercial decisionsOne architecture used before a customer visit and before an account review: governed context assembled deterministically, a context contract per source, every statement typed and cited, bounded read-only tools, explicit behaviour when context is stale, conflicting or missing—and the decision left with a person.
- Operating Model & Process Design / 04Designing account management as a performance, development and recovery systemOne management system for existing customers: understand the account, measure, detect deviation or opportunity, diagnose, decide, commit, act, follow up and reassess—with performance states that change who must do what.
08Related decisions
Decisions this one depends on, or constrains.