← Blog / MEASUREMENT
Measurement in 2026 is triangulation
28 July 2026 · 2 min read · by Obscale
LOG / measurement-triangulation
operating notes
For years the measurement debate was framed as a contest: attribution versus media mix modeling versus incrementality testing, pick your church. By 2026 the serious teams quietly stopped picking. The three methods specialized, and the craft is knowing which question each one answers.
Attribution answers: what happened, right now
User-level attribution, built on first-party events and an identity graph, is the only method that works at operational speed. It tells you this morning that yesterday’s creative swap moved revenue, that a landing page broke checkout on Android, that email flows are quietly outperforming paid social.
Its weakness is honesty about causality. Attribution observes correlation in the journeys it can see. Post-ATT, that view is partial by construction, which is why treating attribution as courtroom truth ended so many careers.
Speed is its job. Use it daily, trust it directionally.
Incrementality answers: did it cause anything
Geo holdouts, audience splits, budget step tests. Slow, expensive, occasionally painful, and the only method that measures causation instead of assuming it. Every platform’s reported numbers, and your own attribution, should periodically stand trial against a holdout.
The 2026 refinement is cadence. Testing everything constantly is a fantasy; testing nothing is negligence. We run structural tests on our biggest line items a few times a year and use the results to calibrate the daily instruments in between.
MMM answers: where should the next million go
Media mix modeling came back from the dead when signal loss made user-level data incomplete, and modern MMM runs weekly on top-line data no privacy policy can take away. It is the board-level view: diminishing returns curves, channel saturation, the value of the spend you cannot track.
Its weakness is resolution. MMM will not tell you which creative works. It will tell you that your search budget passed the point of diminishing returns two markets ago.
The discipline is reconciliation
Three instruments produce three different numbers for the same quarter, and that is normal. The work is reconciling them: attribution calibrated by incrementality, MMM sanity-checked against both, and one person accountable for the synthesis rather than three teams defending three dashboards.
We built ScaleTrack to be the first instrument, the fast one, and to be honest about what it is. One source of truth was always a slogan. Three sources, reconciled, is a measurement stack.