Before relying on a critical figure in a recommendation—regardless of whether it originates from a colleague's spreadsheet, an analytics dashboard, or an AI tool's summary—authors must evaluate it against four core analytical questions:

1. **Provenance:** Where did it come from? Identify the specific system, report, or person that generated the figure, along with the underlying raw data.
2. **Time Window:** Are the evaluated timeframes comparable (e.g., covering the same season and the same number of weekends)?
3. **Inclusions and Exclusions:** What is in and what is out? Determine exactly which stores, products, or customer segments are counted.
4. **Like-for-Like Comparison:** Are both sides of the comparison counted and evaluated using the exact same standard?

Four Questions for Key Numbers

When a supporting number in a recommendation is discovered to be inaccurate, the author should state the error directly and fix it where stakeholders will openly see it, explaining what changed and confirming whether the core proposal still holds. Proactively disclosing revisions and presenting like-for-like figures preserves professional credibility and trust, whereas allowing decision-makers to discover errors on their own severely compromises credibility.

Transparent Error Correction

Spot-checking is the verification practice of directly recalculating the single most important number in a recommendation using raw, primary source data rather than relying on secondary reports. While fluent summaries generated by dashboards, colleagues, or AI assistants may appear trustworthy, independently recomputing the underlying data marks the critical difference between merely assuming a number is accurate and definitively knowing it.

Spot-Checking Raw Data

Learn Before

Related

Learn After