To balance stability & predictability with innovation, Podscribe (as of mid-2026) buckets its attribution updates into pre-announced releases, listed below.
Tomahawk (Sept 1, 2026)
This release improves accuracy, with an increase in determinism and reduction in modeling.
Advertisers should expect:
- more deterministic matches, and a reduced modeling factor
- less than a 25% impact to modeled outputs, in most cases.
- greater accuracy and stability
The main changes are:
- New Podscribe Internal IPv4 <> 6 Graph
This new self-built graph enables Podscribe to directly attribute IPv4 impressions to IPv6 conversions (or vice versa), leading to an increase in unmodeled (deterministic) conversions and reduced modeling. These matches can now be made because Podscribe stores both IP versions for most households in this graph, so if an IPv6 address is observed, the corresponding IPv4 IP is also now known. Since most advertisers will be able to deterministically match over 95% of their impressions now with the same IP address version, we are also removing IPv6 modeling for all advertisers, which was our interim solution as we built this graph, to model for such cases. We believe this simplification will also increase stability.
This update will most benefit advertisers not already sending conversion events through the Podscribe JS web tag or Shopify app, because those integrations typically already send both the IPv4 & v6 address. Most impacted are server-to-server and MMP integrations, as those in most cases do not send both IP types for each event.
- Improved Noisy IP Modeling
- Remove modeling of noisy IPs impression that are linked to household IPs through a device graph (primarily for noisy IPs that come with MAIDs, since MAIDs can be linked to a HH IP). Previously we did not reduce noisy IP modeling for these, because the device graph impact was < 20% for most campaigns.
- Use per-day instead of monthly noisy IP coefficients. This coefficient being more dynamic should result in more accurate attribution for campaigns whose inventory type changes, ie significant fluctuations in the number of household IPs.
- Configurable Device Graph Options Historically, Podscribe has linked MAIDs (from noisy IPs) and HEMs (from purchases) to household IPs only when multiple device graphs can confirm the linkage. This gives the most confident matches that are independently validated by different data partners. However, many advertisers, often those that pass deterministic matches (the individual conversions report) to MTAs, can greatly benefit from more deterministic matches, even if some are below peak confidence. Advertisers, with this new toggle, can now decide on their own between using only matches agreed up on by multiple device graphs (most confident), or matches found by any device graph (most scale).
Podscribe, in most cases, still recommends the conservative option of only using matches agreed upon by multiple providers. Besides the matches having greater confidence, using any graph matches will increase an advertiser’s reliability on 3rd party device graph providers, which may result in less stable results.
- Truncated IP Improvements
- More complete truncated IPv6 IP detection. Different ad servers truncate IPv6 IPs in different ways (IPv4 typically just is truncated by setting the last octet to
.0), so this update makes our check more complete. - (Early) measurement for truncated IPs through SmartServe holdout groups. For the first time ever, Podscribe can support truncated IP attribution. With SmartServe, Podscribe can separate truncated IPs that receive ads into an exposed and randomized user-level holdout group either through “ghost ads” (economical) or PSA ads (more accurate). Then for each group we can how many total conversions occur on all masked IPs behind the truncated IP. Eg if the truncated IP is 10.1.2.0, then all conversions on IPs starting with 10.1.2 would be counted. This may not work for all campaigns, so expectations should be kept low, but if a campaign is SmartServed it may work.
This may not work for a variety of reasons, most specifically noise, or if the campaign cannot be SmartServed. We advise keeping expectations low, and always still requesting non-truncated IPs, as that will always be more accurate.
Since this solution requires attribution to be done on a significantly higher amount of IPs, there likely is an additional fee for truncated IP attribution. Please speak with partnerships@podscribe.com.
- SmartServe Randomized User-Level Holdout Group (RUHG) Incrementality Integration
With SmartServe, Podscribe can compute incrementality via randomized user-level holdout groups — either through “ghost ads” (economical) or PSA ads (more accurate). These results generally we believe are more accurate, although not necessarily as flexible as synthetic control incrementality, the default for the past several years. Podscribe now will update its default to automatically prefer SmartServe holdout incrementality over synthetic incrementality, for any campaigns that have holdout SmartServe incrementality. This option is configurable in the Overview Settings.
Powered by