Why your reports disagree—and how to reconcile them
Start by agreeing on what is being counted, which date determines the reporting period, and which source owns the metric. Then compare the same records over the same period. Automating reports before those decisions are settled can make conflicting numbers arrive faster.
This checklist is for owners and operations leaders who receive different answers from spreadsheets, CRM, finance or operational systems. The aim is a dependable answer to one management question, with the remaining uncertainty visible.
What should you gather first?
Choose one consequential decision, one metric that supports it and one recent reporting period. Bring the source exports, the report showing the disputed number and someone who understands the underlying process. Preserve the original exports so another person can repeat the comparison.
1. Define the decision before the dashboard
“We need better reporting” is too broad to validate. A useful starting question is “Which location needs attention this month?” or “Where are qualified leads failing to become customers?” Write what the owner will do differently once the answer is available. This keeps the reconciliation focused on numbers that matter.
2. Write the metric’s counting rules
For every disputed KPI, record the counted entity, numerator, denominator when applicable, timeframe and source. Include exclusions, date basis and who approves changes. A CRM’s booked sales and accounting revenue may describe different stages of the same business activity. They should not be expected to match until the definitions and timing agree.
3. Match the reporting grain
Decide what one row represents: one order, one customer, one location per day or another explicit unit. Check that joining two sources does not repeat a customer or invoice across several rows. Compare totals before and after each join. A plausible-looking dashboard can conceal double counting.
4. Reconcile one stable period
Compare the same date range, time zone, status filters and included entities. Start from source totals, then trace the difference to individual records. Classify exceptions rather than adjusting the report until it looks right:
- Timing: an update arrived later, or the sources use different dates.
- Scope: a location, account or transaction status is excluded.
- Definition: “lead,” “sale” or “customer” means something different.
- Data quality: a key is missing, a record is duplicated or a connection failed.
5. Assign a source of truth and an owner
Choose the authoritative source for each domain, not one system for everything. Record where leads, revenue and operational activity are maintained. Keep raw extracts separate from cleaned or modeled data. Agree who resolves each exception and how a changed definition will be communicated.
6. Make incomplete data obvious
Show the last successful update and flag missing or late data. Pair a headline metric with a quality check, such as unmatched records, source coverage or the distribution across locations. Missing data should not silently appear as zero. Keep dated snapshots so corrections do not erase the history used in earlier decisions.
7. Automate after the comparison is repeatable
Validate the same rules against another period, document known limitations and then automate the extraction and reporting. Your acceptance check should include agreed source totals, exceptions, refresh behavior and a handoff another person can follow.
What does this look like in practice?
The Dru Bru case study describes connected production, financial and operating reporting. The useful lesson is to build around the work and decisions being supported. Its reported outcomes are specific to that engagement, not a forecast for another business.
Your next step
Use this checklist on one recurring report. If the definitions are clear but the reporting still depends on repeated manual work, explore business reporting and automation. If you are still identifying the bottleneck, the free reporting assessment offers an initial snapshot before the optional email step.