Discrepancies between Ad Platforms and Rockerbox
Spend Discrepancies
When comparing spend numbers reported by ad platforms to Rockerbox data, you may see slight discrepancies and don’t worry, that’s normal!
We expect to see 5-10% discrepancy between the two data sets. There main cause of this is:
- Timezone differences: Rockerbox reporting is standardized to UTC time
Conversion Discrepancies
There are multiple reasons that platform-reported performance will difference from Rockerbox's First Party Attributed reporting.
- De-Duplication: If a user has > 1 marketing touchpoint on their path to conversion, credit from various attribution models in Rockerbox will differ from the "full-credit" view the platform will report. Read more on deduplication here.
For customers leveraging Rockerbox's multi-touch attribution model, fractional credit will be assigned to the channel. For example
Path to Conversion | Facebook Credit | Rockerbox Credit (MTA - example credit) |
Facebook View | 1 | 0.45 |
Non-Brand Search Click | 0.3 | |
SMS Click | 0.25 |
- Attribution windows: Rockerbox maximizes the attribution window that is applied to each channel. In many cases, this means Rockerbox will continue crediting a channel for driving a conversion for longer than a platform's default (typically 1 day view 7 day click).
- Data Access Differences: differences in the type of data each platform may be able to access may differ from Rockerbox to the platform. While these cases vary by vendor, common examples include:
- Platforms attribute views and engagements on their site and mobile app, but do not share this with third parties.
- Impression pixels that can't be fired on in-app inventory
- Log level reporting received by a vendor off-cadence that's not provided to Rockerbox
- Leveraging third-party user identity platforms
- Methodological Differences: all platforms have their own methodology for tracking and reporting on the impact of their channel. While specifics depend on the vendor (and how closely their methodologies align to Rockerbox's) common example include:
- Proprietary Modeling: each platform's own method of modeling the impact of their channel when the entire impact cannot be tracked deterministically