Why your ROAS is wrong and how cross-device identity fixes it
A shopper who sees you on Meta on the web, taps a TikTok ad on iOS and buys on Android looks like three people to most tools. Here is how an identity graph collapses them into one, and what it does to your reported return on ad spend.
Read the articleLatest articles
Click, impression, view-through: the three windows every MMP argues about
28-day click, 1-day impression, 7-day view-through. Where these industry-standard windows come from and how changing them reshapes which channel gets the credit.
One model, every vertical: adding banking without touching the frontend
How a vertical-agnostic catalog lets a commerce analytics platform absorb loan approvals and card activations with three SQL inserts and zero React changes.
Forecasting revenue in a market with Ramadan
Backtesting OLS against Holt-Winters, excluding in-progress months, and layering a KSA Ramadan regressor so seasonal peaks stop breaking the model.
Stop inventing your margin: real-cost profit vs. the 40% placeholder
Why a fixed COGS assumption quietly lies to your board, and what it takes to compute contribution and gross profit from actual product costs.
Answers, not dashboards: asking your data a question in plain language
The design philosophy behind letting anyone on the team get a defensible answer, complete with the query behind it, without waiting on a data team.
Get new posts by email
Occasional, practical, no spam. Attribution and data-modelling notes as we publish them.