Concept Everyone

Why our numbers and yours differ

Updated 30 October 2026

Two sets of numbers that track the same users rarely match to the unit, and the gap almost always has one of a few causes: our days are UTC days, our Conversions count only an offer’s first step, reversals lower the day a conversion was earned and postbacks that failed, repeated or were tests count differently on your side than on ours.

The usual causes

Cause	What we count	What you might be counting
Time zones	UTC days. A Texas evening lands partly on tomorrow.	Your local days
Conversions and events	Conversions is the first step of an offer only. Later steps are Events.	Every postback as a conversion
Reversals	Taken off the day the conversion was earned, so a closed week can drop	Taken off the day the reversal arrived, or not at all
Postbacks that never reached you	Every conversion, sent or not	Only what your server received
Repeats	One conversion, even when the postback was sent twice	Each copy, if you dedupe on `action_id` alone or not at all
Test postbacks	Nothing: a test creates no conversion	Tests your code credited
Rejected or in-review conversions	Not in totals	Nothing, since no postback was sent
Money fields	Your earnings	`payout` on the Hosted Offerwall, which is the player’s reward there
Clicks	Taps that reached our tracking link	Taps on your own button, including ones that never loaded the link

Example

Example, made up: Pebble Rewards’ own database shows 12 Harbor Blocks installs for 29 October, Texas time. Our Dashboard shows 10 conversions for 29 October.

One install came in at 7:30 PM in Texas, which is 12:30 AM on 30 October in UTC. It’s in our 30 October.
One “install” in their database is the player reaching level 10. Their code booked every postback as an install. On our side it’s an event.

12, less one on the next UTC day, less one later step, is 10. Nothing is missing.

Reconcile in four steps

Get our rows for the same UTC window. Publishers: Conversions, then Download CSV, or GET /publisher/conversions with your API key; Pull conversions instead has the call. Networks: Reporting, then Flight Recorder lists each postback with its Action ID.
Match on action ID and status. Our actionId is the postback’s action_id. A reversal repeats the action_id with status=reversed, so match on both. For conversions first sent before 10 September 2026, match on the end of the ID: the postback’s can be longer.
Rows we have and you don’t: open Reporting, then Flight Recorder, and look at Delivery. failed means your server didn’t answer with a 2xx; skipped_no_url means no postback URL matched. Publishers: with no postback secret, nothing is sent at all. How to answer a postback covers retries.
Rows you have and we don’t: look for test=true, and for the same action_id and status stored twice.

Advertisers: our numbers and your attribution tool’s

We charge and count an event only when it carries a click ID that matches a tap on your campaign. These never count:

Events without a click ID, logged as missing_attribution.
The same eventId sent again: we answer duplicate and count only the first.
Events whose timestamp is more than 48 hours from when we received them, logged as outside_window.
Test events.

Your attribution tool also has its own windows and its own way of picking which tap gets the credit, so its installs from AffiliateOS can be higher or lower than ours. Event logs on each campaign shows each event we logged and its status; a repeated eventId isn’t logged again.

Send us this

If the gap is still there: the email you sign in with, the UTC window, your count and ours, and a few action_id values (advertisers: event IDs) that you have and we don’t, or the other way round.

Reversal. A conversion taken back after it was reported, by the advertiser or after our traffic checks. We take back its payout and send the publisher a reversal postback.What is a reversal?