Adjust Insight: Incrementality testing in practice | Adjust

Summary

Incrementality testing has become essential for app marketers navigating multi-channel user journeys. Adjust's InSight, launched in 2024, applies machine learning and AI to evaluate the incremental effect of budget changes, delivering results with 95% confidence intervals. This statistical rigor distinguishes genuine growth from natural variation or demand redistribution.

A benchmark of U.S. tests across iOS and Android reveals distinct patterns: on iOS, budget increases produced statistically significant lift in roughly one-third of cases for major platforms and about one-fifth for others. On Android, the dynamic shifts—up to a quarter of budget increases led to statistically significant organic cannibalization, where paid spend merely replaced organic conversions without net growth.

These findings underscore the value of causal modeling: lift percentages indicate magnitude, but significance determines reliability. InSight translates this statistical output into action through its Results and Recommendations view, which highlights incrementality percentage, significance, and model accuracy (MAPE) at a glance, and a data table for granular analysis of budget shifts, bid strategies, and partner mix. For ad ops decision-makers, the actionable takeaway is clear: statistically validated lift supports scaling, while cannibalization flags the need for reallocation.

InSight does not replace attribution but complements it, linking channel-level visibility to causal impact. By integrating with Adjust's ecosystem, teams can centralize reporting and make data-driven decisions with greater confidence.

Analyst Note

What's notable here is not just that Adjust's InSight offers incrementality testing, but that the benchmark data explicitly quantifies organic cannibalization as a measurable risk—especially on Android. For UA teams, this moves the conversation from 'did we get a lift?' to 'did we shift existing demand or create new?'. In a post-IDFA world where attribution fidelity erodes, incrementality becomes the backstop for budget confidence.

The fact that statistical significance is reported with 95% confidence intervals and MAPE metrics signals a maturation from directional 'lift' toward action-ready thresholds. The practical impact is clear: teams can now identify which partners genuinely scale addressable audiences versus recycling organic users. This is particularly timely as privacy-driven fragmentation pushes marketers toward modeled measurement.

The competitive angle is that Adjust positions InSight as a complement to attribution rather than a replacement, which may appeal to teams hesitant to fully abandon last-click frameworks. The industry signal is that statistical rigor, not just algorithmic black boxes, is becoming table stakes for measurement partners.