Data & Intelligence
Analytics
We answer the questions your business has been waiting on: who stays, where you lose people, whether a change worked. Each answer states its confidence and limits.
The problem
Your data sits in several systems, and a simple question takes weeks to answer.
Customers live in the CRM, payments in accounting, behavior in GA4, product usage in a database. IDs don't match, so each question turns into a round of exports and VLOOKUPs.
The second trap is interpretation. Averages hide groups moving in opposite directions, correlation gets read as cause, and a falling number sends people looking for the cut that fits their theory. Good analysis starts by writing down the question and the rival explanation before anyone opens the data.
You may recognize this
- Answers arrive after the decision has been made.
- Key analyses live in personal spreadsheets and can't be rerun.
- The effect of a change is judged by comparing before and after.
How we work
Run a question backlog
Open questions are ranked by what they'd change. Each gets a hypothesis, a rival explanation and an evidence bar agreed up front.
Model the warehouse
We join GA4's BigQuery export with CRM, payment and product data, resolving customer identity where your legal basis allows. Transformations are versioned code, often dbt.
Set data contracts
Source owners agree on fields, types and change notices. Automated tests catch nulls, duplicates and broken joins before they reach an analysis.
Segment, then write it up
Cohort and funnel work shows what blended figures hide. Experiments are sized in advance, anything else is labeled observational, and each analysis ends in a short decision memo.
What we deliver
- Prioritized analysis backlog
- Version-controlled warehouse models
- Data contracts for key sources
- Cohort, funnel and segment analyses
- Library of decision memos
How we measure
- Time to answer: shows whether analysis keeps pace with the business.
- Decisions changed: the honest test of whether a piece of analysis mattered.
- Reproducibility: when a second analyst can rerun the work, knowledge stays in the company.
A subscription business with flat churn
Illustrative scenario: Imagine a subscription meal-kit business whose churn looks flat, which leadership reads as stable. Splitting customers by signup month and acquisition offer might show promo-driven cohorts leaving early while older ones stay. The conversation would then shift from retention emails to what those offers bring in.
Common questions
How is this different from your GA4 and GTM work?
That work fixes how data gets collected. This work joins it with your other systems and answers business questions. If collection is broken we fix it first, since analysis on bad inputs only looks rigorous.
We have a data team. Would this help?
It helps when the team is buried in reporting requests. We can take on modeling, test coverage or one priority question, and the code stays in your environment.
Can we join customer records under KVKK?
Possibly, depending on your purpose and what your privacy notice covers. We minimize fields, pseudonymize identifiers where we can and restrict access. The legal call belongs to your counsel.
Before we work together, let's think together.
Requests are answered by the founder.