Topics · Metric
Ad conversions
On Instagram, ad conversions are actions, such as purchases, credited to an ad after someone sees it.1 Advertisers define the period in which actions count; Meta’s research explains that shorter periods provide fresher training data but miss later conversions, while longer periods capture more outcomes with less-fresh data.1 Meta’s engineering team reported in August 2026 that behavior-based modeling and broader improvements increased Instagram ad conversions by a cumulative 6%.2 In October 2026, Meta’s engineering team described GEM’s shift to one central model learning across Meta’s advertising platforms, including Instagram, for advertising goals that include conversions.3
Attribution windows and delayed conversions
Meta research and patents. These may not describe how Instagram works today.
Meta’s research says shorter attribution windows give prediction models more up-to-date training examples but miss later conversions, while longer windows capture more outcomes at the cost of less-fresh data.1 Meta’s research describes traditional approaches that combine models trained on different windows or update models as those windows close.1 Meta’s research says these approaches increase training costs and can destabilize learning when the same ad view receives conflicting labels about whether it led to a conversion.1
Conversions as an advertising objective and performance measure
Meta research and patents. These may not describe how Instagram works today.
Where Meta’s attribution research concerns which ad conversions enter training data, Meta’s engineering team’s modeling research concerns improving conversion results from the ads it recommends.123 For advertising goals that include conversions, Meta’s engineering team describes GEM’s 2026 shift from separate improvements for Facebook, Instagram and Reels to one central model learning across Meta’s advertising platforms.3 Meta’s engineering team also describes using thousands of past events, including clicks, views and purchases, to understand people’s behavior more closely and improve conversion rates.2 Meta’s engineering team reports that these behavior-based improvements, together with broader modeling changes, produced a cumulative 6% increase in Instagram conversions.2