All products

Reconcile two reports with AI

Work out why two credible-looking totals differ. Compare definitions, time boundaries and the records counted before declaring one source wrong.

Reconcile two reports with AI when a meeting grinds to a halt over whose number to use. The metric-reconciliation skill follows the calculation on both sides, recording the first place the figures diverge. It can also document an agreed tolerance when the difference is immaterial.

For example, one booking export may count a party once and another may count every place reserved. Neither file has to be faulty. Show which question each answers, give the totals meaningful names and leave a record for the person preparing next week’s figures.

What it does

  • Compare figures for the same date range and time zone.
  • Check what one row represents before adding values together.
  • Find different filters, joins, treatment of nulls or refresh times.
  • Calculate an absolute gap and percentage from numeric inputs or local CSV extracts.
  • Record a correction, a designated source or a documented difference in meaning.

Install and use

  1. Install the skill in a compatible agent and use Python 3 for the local comparator. Supply authorised totals or CSV exports, metric definitions and the period.
  2. Ask, “Trace the weekly booking discrepancy. Check whether both reports count cancellations before looking at the remaining rows.”
  3. Review the first evidenced difference and agree the treatment with the owners of both reports.

FAQ

Source

Made by Nimrod Fisher. Original skill: metric-reconciliation, published on GitHub under the MIT licence.

View on GitHub