I make your data stack AI-ready.
So your team can ask your data a question and trust the answer.
For SaaS and DTC teams scaling spend across Europe.
What AI-ready is built on.
It comes down to one thing: data the business can trust. Collected clean, defined once, checked against what drove growth.
- 01 Collect
Tracking that stays correct
When product and marketing ship and iterate, your tracking keeps reporting the truth. A broken event is caught the day it happens.
Your tracking plan becomes a data contract. Automated tests check every live event against it, verifying the data layer and every analytics and marketing request.
- page_view dataLayer · Google Analytics 2/2
- add_to_cart dataLayer · Google Analytics 2/2
- begin_checkout dataLayer · Google Analytics 2/2
- purchase dataLayer · Google Analytics · Google Ads · Facebook 4/4
10/10 checks passed · no drift against the contract
- page_view dataLayer · Google Analytics 2/2
- add_to_cart dataLayer · Google Analytics 2/2
- begin_checkout dataLayer · Google Analytics 2/2
- purchase dataLayer · Google Analytics · Google Ads · Facebook 3/4
└ purchase: revenue missing from the Google Ads conversion
9/10 · break caught at build, blocked before release
Every event checked in the dataLayer and each tool it fires toclean events - 02 Model
One metric everyone trusts
Marketing, finance, and product stop bringing three different numbers for the same metric. Every dashboard gives the same answer, every time.
dbt models turn raw events into clean, tested tables; a metrics layer defines each KPI once, in code, so 'active customer' means one thing company-wide.
- staging 14 models · 31 tests
- marts 9 models · 42 tests
- metrics layer active_customer · net_revenue · cac
active_customer· one metric, every surface
dashboard 12,480finance report Active customers 12,480board deck 12,480 Active customers1 definition · 3 surfaces · 0 mismatches
Models and tests run on every merge, like the rest of the codebasetested metrics - 03 Prove impact
What your spend actually caused
You learn which channels drove new revenue and which just took credit for demand you already had, reported as honest ranges.
Geo holdouts measure lift directly and calibrate a Bayesian marketing-mix model, so every estimate carries credible intervals. Plus cohort LTV and true CAC.
paid search ✓ confirmedpaid social ✗ overstated 16×decision: rebalance toward search
claimed = platform attribution · measured = lift vs holdout regionsAI-ready
The specific projects.
Most combine two or three. Not sure where the gap is? Start with the assessment.
AI-Readiness Assessment
Assessing your data on AI readiness. You leave with a score, recommendations, and a roadmap to keep.
Readiness · 5 dimensions
- signal
- quality
- consistency
- access
- governance
- Collect CI · 3–6 wk
Tracking & data quality
Your data collection setup and tracking integration, brought to a standard that holds release after release, verified by automated tests instead of manual QA.
- Model dbt · 4–6 wk
Analytics engineering
Every marketing and product source, clean and verified in your own warehouse, whether BigQuery, Snowflake, or Redshift. Modelled with tests, docs, and data quality monitoring.
- Prove impact experiments · 4–8 wk
Incrementality & marketing mix
Get a clear answer on which channel is earning its budget. Geo holdouts, marketing-mix models, and causal inference to settle the allocation debates.
Visable
B2B commerce platformTracking engineered like software
Several teams changing tracking in parallel during a platform migration, with reliable data needed from day one.
What I built
- 01 Tracking as data contracts, written as an AI-readable semantic layer.
- 02 Data-quality tests in CI, so a broken event fails the build before it reaches a report.
- 03 A platform that makes tracking data and test results accessible across the organization.
Result
Tracking the team runs like code: breaks caught before release, one place to see what's tracked and whether it holds.
Where I built the craft.
Ten years in-house before going independent. The experience behind the work above, not a roster of freelance clients.
-
Pipedrive
Sales CRM
Geo experiments and causal analysis that isolated true incrementality, contributing to six-figure savings on spend that wasn't driving activations. Built the demand-gen dashboards that steered an eight-figure budget.
-
Immowelt
Property marketplace
Took a marketing data lake from concept to production on AWS and Google Cloud, with Databricks for the ETL pipeline, and introduced Tableau to the wider marketing organization for self-service analytics. Marketers and executives could make better, faster decisions.
-
air up
DTC consumer brand
Built the go-to-market analytics foundation that made performance comparable across markets during international expansion.
When no one trusts the data, the analytics stops mattering.
I'm Henning, a data analyst. For more than ten years I've built the data and analytics that teams base their decisions on, inside Pipedrive, Immowelt, and air up, now independently. Long enough to learn what gets me hired: a team that has stopped trusting its numbers and quietly gone back to gut feeling. My job is to give them data they can stand behind again.
I work the whole chain myself, in your repo and your warehouse: the tracking, the dbt models, and the causal experiments that prove which spend moved something. When the engagement ends, the work stays: code and tests, with best practices from software engineering keeping your tracking, models, and experiments at the highest standard.
My strength is understanding your business and supporting the decisions that make sense for you, backed by years of hands-on technical work.
Tracking
Google Analytics
Google Tag Manager server-side
Consent Mode v2
Tracking tests
Engineering
dbt
Semantic layer
BigQuery · Snowflake
Redshift · Postgres
Causal
Bayesian MMM
Geo experiments
LTV · CAC
Tableau · Power BI
Things people ask before we start.
- 01
What do we walk away with when you leave?
Working code, tested and well documented, in your own repos: your tracking, your models, your pipelines. Your team can run, change, and extend it without me. No black box, no lock-in.
- 02
How do I know the engagement will work out?
I take on few clients at a time and keep each engagement tightly focused. Every step has a clear goal, so you see working results as we go.
- 03
Why hire you instead of an agency?
You get one specialist who learns your business and works embedded in your team, personally accountable for the outcome. You always know exactly who's doing the work. Agencies fit better when you need broad execution across many channels.
- 04
Can we start small?
Yes. Start with the AI-Readiness Assessment: two weeks, fixed fee. I score your data foundation and hand you a prioritized roadmap you keep, whether or not we work together further.
- 05
Will the tracking you build pass our privacy review?
Privacy is built into how I work: consent-respecting, data-minimizing, and designed to pass your data-protection review on the merits. The right approach depends on your stack, and I build to it.
- 06
Do you work with our stack?
Yes, any modern warehouse, from BigQuery to Snowflake to Postgres. The specific tools matter less than the experience I bring; I get productive in an unfamiliar stack fast and adapt to how you already work.
- 07
Where are you based, and how do you work?
Remote-first. I'm based in Hamburg and work with teams across Europe; I'm available on-site for kickoffs and key reviews anywhere in the EU.
Let's start with a conversation.
No pitch, no upsell. We talk through what you're working on, and whether I'm the right person to help.