Data & Intelligence
Attribution
We help you estimate what your marketing spend caused. You see which evidence to trust for which budget decision, and how uncertain it is.
The problem
Every ad platform claims the sale, and added together they claim more sales than you made.
Each platform counts the touchpoints it can see and credits itself under its own rules, so one order can appear in Google, Meta and email reports at once. GA4's data-driven model sees more, yet only on journeys it could observe. Declined cookies, phone calls and referrals fall outside it.
Attribution tells you who touched the sale. A budget decision needs a harder answer: would the sale have happened anyway? Branded search and remarketing often look strong because they reach people already close to buying.
You may recognize this
- Platform-reported conversions add up to more than your actual orders or deals.
- Branded search shows your best ROAS, and nobody has tested cutting it.
- Your sales cycle runs longer than the platforms' lookback windows.
How we work
Set the evidence ladder
We match methods to decisions: platform signals for daily optimization, mix modeling for quarterly allocation, controlled tests for the biggest calls.
Close the CRM loop
Closed revenue flows to the warehouse and, where consent allows, back to platforms as offline conversions. We then compare data-driven, last-click and CRM views to see where they disagree.
Run incrementality tests
Geo experiments, holdouts or platform lift studies get a written hypothesis, power check and stopping rule. Small budgets may only detect large effects, and we'll say so before you spend.
Calibrate with MMM
Given enough history and spend variation, we fit a mix model in open-source tools such as Meridian or Robyn and calibrate it against test results. Outputs are ranges.
What we deliver
- Measurement blind spot review
- Offline conversion pipeline from CRM
- Incrementality test designs and readouts
- Budget scenarios with uncertainty ranges
How we measure
- Cost per incremental conversion: what a channel adds beyond the baseline, read with its confidence range.
- Gap between platform-claimed and CRM revenue: tracks how far platforms overstate themselves.
- Share of spend backed by tests: how much budget rests on more than a model.
A B2B supplier's long sales cycle
Illustrative scenario: Say a B2B equipment supplier runs search and LinkedIn campaigns against a sales cycle of several months. Reports credit most deals to branded search at the last click. Before cutting LinkedIn, we would pull closed revenue from the CRM, hold out matched regions, and compare pipeline after a full cycle.
Common questions
Which attribution model is the right one?
None of them is true. Each spreads credit under its own assumptions. We watch where they disagree and settle what matters with tests.
What does marketing mix modeling require?
Usually a few years of weekly spend and sales data, with real variation across channels. On flat or small budgets, simple tests will serve you better.
How do consent rules affect this?
People who decline can't be followed, so user-level attribution always has gaps under KVKK and GDPR. Geo tests and mix models work on aggregates, though they bring their own uncertainty.
Before we work together, let's think together.
Requests are answered by the founder.