A Practical Guide to Reconciling Ad Platform Data With Your Reporting Dashboards

Why reconciliation gets skipped

Every reporting pipeline has the same weak point: the moment data moves from an ad platform into a dashboard. Currency conversions, attribution windows, timezone boundaries, API sampling — any one of these can quietly shift a number without anyone touching a formula. There's a simple check that catches it: compare your reporting dataset against a direct platform export. Easy in theory. Almost always skipped in practice, because it's manual, it's tedious, and it's the first thing to get cut when a deadline is close.

Here's the catch, though. The cost of skipping it doesn't show up right away. It shows up later — usually in front of a client, or in a QBR — when someone asks why the dashboard says one number and Google Ads says another. By then it's not really a data problem anymore. It's a trust problem.

What "source of truth" actually means

A source of truth isn't inherently more accurate than any other export. It's just the dataset you, your client, or your stakeholders have agreed to treat as the reference point. In practice, that's almost always a direct export from wherever the activity happened: Google Ads for search spend, Meta Ads Manager for paid social, your billing system for revenue. Everything downstream — your warehouse, your BI tool, your client-facing dashboard — should agree with it. Or you should be able to say exactly why it doesn't.

That's really what reconciliation is for. Not proving your dashboard is right. Giving you a fixed reference point, so when numbers disagree, you already know which side to check first.

A simple reconciliation workflow

You don't need a data engineering team to do this well. A lightweight, repeatable process beats an occasional deep audit:

1. Export the source-of-truth report directly from the platform for the period you're reporting on. Not a copy someone forwarded you.
2. Export the same period from your reporting dataset (warehouse, spreadsheet, dashboard export, whatever you've got).
3. Match up the columns that should mean the same thing — spend, impressions, conversions — even if they're named differently.
4. Sum each matched column in both files and compare the totals. Don't just sample a handful of rows.
5. Flag anything outside a reasonable tolerance for investigation. We treat under 1% as a clean match and over 5% as a real discrepancy.

What to do when you find a mismatch

Not every mismatch is an error. Attribution model differences, in-flight conversion lag, currency rounding — all legitimate reasons two totals won't match exactly. You're not chasing zero variance. You're chasing explainable variance. Before you touch the dashboard, check the obvious stuff first: date range boundaries, timezone settings, whether both exports cover the same campaigns, whether one side's been filtered.

If you can't explain a discrepancy in one sentence, it's not reconciled yet — it's just noted.

Once you've found and actually understood a real discrepancy, write it down somewhere your team and client can see. Not a Slack thread that scrolls away and disappears. That's the difference between a one-time fire drill and an audit trail you can point to next time someone asks.

Automating the check

This workflow holds up fine for one dataset checked occasionally. It stops holding up the moment you're running it across multiple clients, multiple platforms, every single reporting cycle. That's the exact gap the Data Integrity Checker in KPI Compass closes: designate one uploaded file as your source of truth, and every other dataset in your workspace gets compared against it automatically. Column-level sums, match/drift/discrepancy status, the specific numbers behind any flagged variance — no spreadsheet built by hand.

Still worth knowing the manual version, though. It's the same logic running under the hood, and understanding it makes an automated check a lot easier to interpret.