AI Prompts for CRM Data Analysis

Most CRM analysis fails before it starts, because the data is not trustworthy: stages advanced in bulk before quarter end, close dates that always fall on the last day of the month, activity logged inconsistently, fields left blank. The first job is a data quality audit that says which analyses the data can support. After that, the useful questions are about relationships — which activities, at which stages, for which segments, actually correlate with winning.

These prompts audit the data, analyze the relationships, and design reporting. They work from exports you provide; the model should show its method and state where the data cannot support a conclusion.

Before you use these

Have these ready to replace the highlighted [variables]:

• Opportunity export: stage history, dates, amount, owner, segment, source, outcome
• Activity export: type, date, owner, associated opportunity or account
• Field completion rates and known data practices (who updates what, when)

The prompts

• 1. Audit CRM data quality
• 2. Analyze activity-to-outcome relationships
• 3. Design reports that answer sales questions

1. Audit CRM data quality

Act as a revenue operations analyst auditing CRM data quality.

Opportunity extract: [paste: id, created date, stage, stage change dates, close date, amount, owner, segment, source, outcome]
Activity extract: [paste: type, date, owner, opportunity/account]
Practices: [how and when reps update records; any bulk updates; what is mandatory]

1. Completeness: completion rate per field, and whether missing values are random or concentrated (by rep, segment, stage).
2. Corruption tests: close dates clustered at period ends; stages advanced in bulk on the same day; opportunities created and closed on the same day; amounts unchanged from creation to close; activity logged only at stage changes; stage regressions. Report each test with the evidence.
3. Reliability rating per field (trust / use with caution / do not use) with the reason.
4. Analyses the data can support today, and the ones it cannot (e.g. stage conversion is unreliable if stages are bulk-advanced).
5. Fixes: field validation, required fields by stage, process changes — ranked by the analytical value they unlock.
6. The minimum data hygiene needed before forecasting from this CRM is credible.

Do not analyze past the data's reliability. State clearly when a field should not be used.

2. Analyze activity-to-outcome relationships

You are analyzing the relationship between sales activity and outcomes.

Data: [opportunities with outcome, stage history, amount, segment, owner; activities with type, date, linked opportunity]
Reliability notes: [from the audit]

1. For won vs lost opportunities, compare: number and type of activities per stage; time in each stage; number of distinct contacts engaged; presence of specific milestones (demo, executive meeting, proposal, security review); multi-threading; days from creation to first meeting.
2. Report each comparison with sample size and the size of the difference. Flag comparisons where the sample is too small or the field unreliable.
3. Segment the findings (deal size band, segment, source, rep tenure) where sample allows; note where a pattern reverses by segment.
4. Identify the three patterns most strongly associated with winning and the three most associated with losing. State plainly that these are associations, and the alternative explanations (e.g. good deals get more activity because reps sense they are good).
5. Convert the actionable patterns into stage exit criteria or coaching points.
6. What to measure prospectively to test whether changing the behavior changes the outcome.

Present as tables with a short narrative. Do not report a pattern without its sample size.

3. Design reports that answer sales questions

Act as a sales analytics lead designing CRM reports for [team].

Decisions made: [weekly: e.g. which deals to inspect, where to add pipeline; monthly: forecast, coaching focus, territory issues; quarterly: process changes]
Reliable fields: [from the audit]
Audience: [reps, managers, leadership]

1. For each decision: the question the report must answer, the metric(s) with exact definitions, dimensions and filters, the visualization or table layout, cadence and owner.
2. Metric definitions to standardize across all reports: pipeline coverage, stage conversion, sales velocity, average sales cycle, win rate (by count and by value), slippage, activity per opportunity. State the formula and the exclusions for each.
3. Prerequisites: fields or validations that must exist for each report to be trustworthy; the reports to defer until then.
4. The 'vanity' reports currently produced that serve no decision, and the recommendation to retire them.
5. A one-page weekly manager view: the five numbers, the deals flagged, the actions.
6. Governance: who owns definitions, how changes are versioned.

Design for decisions. If a report does not change what someone does, do not include it.

Related prompts

• Pipeline Review and Analysis
• Sales Forecasting
• Deal Reviews
• Win/Loss Analysis
• Buying Signals and Trigger Event Analysis
• Data Analysis prompts

Logical next step

After this, most sales teams move on to Pipeline Review and Analysis.

All Sales prompts · Search the full library