What is a data discrepancy?
A data discrepancy is a difference between two reports that appear to measure the same activity. A store may record 1,000 paid orders while an ad platform reports 780 attributed purchases. The numbers can both be valid if they answer different questions.
Where do differences come from?
- Definitions: One report counts paid orders; another counts checkout completions or credited conversions.
- Attribution: Platforms use different click and view windows and assign credit under different models.
- Coverage: Consent, blocked requests, app permissions, and missing identifiers affect which events each tool sees.
- Timing: Time zones, delayed uploads, refunds, and processing schedules shift totals.
- Implementation: Duplicate tags, failed API calls, inconsistent order IDs, or mismatched event names create real errors.
Do not dismiss a gap as normal before checking the implementation. A consistent difference can reveal a broken event or a missing destination.
How do you investigate?
Start with one metric and one date range. Write down the event definition, time zone, attribution window, and source for each report. For purchases, compare a sample of order IDs and timestamps against the store's order records. Check rejected API events and duplicate IDs. Then separate explained differences from records that still need investigation.
For channel reports, compare like with like: an ad platform's attributed conversions are not the same metric as all orders in the store. Keep a short reconciliation note so the next analyst can repeat the check.
How Hardal handles it
Hardal collects each event once on your own first-party domain, then sends the same record, with the same event ID, to GA4, Meta, TikTok and your warehouse. Every destination starts from one source, so a remaining gap is easier to trace to an attribution rule or a definition. See how server-side tracking sets this up.
Related terms: Event deduplication, Attribution, Server-side measurement, First-party data