Talk to your marketing data.
We build one data model out of every marketing source you have: spend, campaigns, conversions, margin and the blended model that tells you what actually works. Then you simply ask it questions, in Claude, in ChatGPT, in Slack or in the app. No dashboard hunting, no warehouse to build, no data team to hire.
Teams already steering with Datacompany.
You spend on marketing. You can't tell what actually worked.
Five compounding reasons every marketing manager is flying half blind, and why hiring your way out costs more than it saves.
We build your marketing data model. You talk to it.
Your sources do not arrive as raw API dumps. We connect them, align the channel names, strip the double counting, normalise currencies and regions, attach margin to every conversion, and put the blended model on top. What you end up with is one coherent model of your own marketing, and a question in plain language gets a real answer out of it.
- 01ConnectGoogle Ads, Meta, LinkedIn, Microsoft Ads, Search Console, your analytics and your CRM over OAuth. We keep those connections alive when a platform changes its API, so nobody on your side ever gets a ticket about it.
- 02ModelChannel names aligned, double counting removed, currencies and countries normalised, margin per goal attached. Then the blended model on top: MMM, attribution and holdouts side by side into one number per channel, with a confidence band.
- 03AskPut your question where you already work. The answer comes out of your own model, not out of a platform grading its own channel.
It is your model, not our dashboard. You get read-only database access with stable v1 views and a nightly push to BigQuery, Snowflake or your own Postgres.
Ask in plain language, right where you already work.
Connect Datacompany to Claude or ChatGPT and you ask your question in the chat you already have open. Everything an assistant can reach is read-only and scoped to your organisation alone. And if the AI answer doesn't satisfy you, a real data analyst takes over. No ticket system, no extra invoice. This is the part nobody else does.
- Change or delete anything. Every tool is read-only.
- See another organisation's data, on any request.
- Read your chats, your files or anything else in the assistant.
- 01Step 1You askType your question in English or Dutch, like you would in ChatGPT. No SQL, no metrics setup, no dashboard hunting.
- 02Step 2AI answersOur LLM writes the query against your live data, runs it, and answers with the chart and the assumptions it made, all auditable.
- 03Step 3One click to a humanNot convinced? Click 'Escalate'. A senior data analyst picks it up within a working day, with full context of your question and the AI's answer.
- Which channels made the most profit last month, and how sure is the model?
- If my budget stays flat next month, where should I move spend?
- Conversions dropped last week. Is that real, or is a data source behind?
Last-click misses what actually moves the needle.
Same business, same revenue, same month. Two views. Last-click can only see channels with a trackable click on the path. Our blended model is built on holdouts and incrementality, so the channels that genuinely drove revenue show up at the share they earned.
We model podcasts, billboards, bus shelter ads, sponsorships and local activations into the same view as your digital spend. No more guessing whether the radio buy paid back. It either lifted sales in the treatment region or it did not.
One independent measurement of what actually works.
Datacompany combines Bayesian MMM, attribution and incrementality experiments into one source of truth, refreshed weekly, validated against real-world holdouts, and explained by AI with a senior data scientist one click away.
From connected to confident in weeks.
See exactly what moved the needle.
Channel-level incrementality, holdout-validated attribution and campaign-level causal measurement. Built on PyMC-Marketing, explained in plain English.
| Channel | Current | Proposed | Δ | Incremental ROI |
|---|---|---|---|---|
| Google Ads - brand | €24,000 | €18,000 | -25% | 6.2x |
| Google Ads - generic | €31,000 | €36,000 | +16% | 3.1x |
| Meta prospecting | €22,000 | €14,000 | -36% | 1.4x |
| Meta retargeting | €12,000 | €15,500 | +29% | 4.8x |
| LinkedIn brand | €8,000 | €13,500 | +69% | 5.4x |
An analytics team's monthly output, ready to forward.
On the first business day of every month, a 14-page PDF lands in your inbox. It tells the story of what happened to your marketing, what we did about it and what should change next. Written by AI on your real numbers, validated by a senior data scientist before it reaches you.
Monthly marketing attribution report
October closed at €328,500 in attributed revenue, up 14.2% versus September and ahead of plan. Brand-driven channels, in particular podcast sponsorship and Dutch billboards, carried more of the growth than last-click reporting suggested. We recommend pulling 18% of Google Ads brand budget back and reallocating it to LinkedIn brand and the Q4 podcast slate, where the calibrated incremental ROI sits above 4.8x.
- 1Reduce Google Ads brand by 18%, reallocate to LinkedIn brand
- 2Extend Q4 podcast slate by two episodes; incremental ROI 5.4x
- 3Run a billboard holdout in Noord-Holland for two weeks
- 01Executive summary in plain languageThree paragraphs your CMO and CFO can read in two minutes. No jargon, no dashboards, just the story.
- 02Per-channel attribution and incrementalityHoldout-validated contribution per channel with confidence intervals. What is real, what is overstated.
- 03Per-country and per-goal deep divesMargin-weighted ROI per market and per business goal, so the recommendations land in euros, not just ROAS.
- 04Concrete recommendationsPrioritised actions with expected impact. Scale this, cut that, run this holdout next.
- 05Model accuracy and audit trailMAPE, RMSE, residual patterns and every assumption we made. Defensible in any board meeting.
An analytics team that works while you sleep.
Everything below is shipping in beta today, except where marked coming soon. Together they replace what an in-house data team would do, at a fraction of the cost.
Independent, transparent, with a human in the loop.
There is no shortage of AI tools that promise marketing miracles. Most are wrappers around the same platform APIs they are supposed to evaluate. We are built differently.
Plug it in. We handle the pipes.
API-first ingestion across 200+ marketing, CRM, e-commerce, finance and analytics sources. No exports, no spreadsheets, no engineering tickets.
One price, and it is right here on the site.
1% of your monthly ad spend, with a floor of €999 and a ceiling of €1,999. No plans to pick between, because everyone who uses this seriously needs the same things, the direct line to a human included.
Spending less than about €50k a month on media? Say so when we talk. We will look at whether there is enough signal in your data before you commit to anything, rather than after your first invoice.
- Full mix model: adstock, saturation, confidence bands
- Every channel in the model, offline included, no cap
- A separate model per country or market
- Geo holdouts we design and analyse for you
- Budget allocation and what-if scenarios
- Runs weekly and on request, unlimited history
- Every connector, from Google Ads and Meta to Klaviyo, Search Console and your CRM
- MCP connector for Claude and ChatGPT, unlimited, including run_sql
- Questions and anomaly alerts in Slack or Teams
- Read-only database access with stable v1 views
- Nightly push to BigQuery, Snowflake or your own Postgres
- Boardroom-ready monthly report, checked by a human first
- Unlimited users, at no extra cost
- A direct line to a senior data scientist, no support layer in between
- Escalate from any answer, picked up within one working day
- Onboarding and model review by that same person
Not happy after the first month? You get it back.
One month is enough to see what you actually bought: your sources connected, the model run on your own history, and the split per channel with its confidence bands in front of you. That is the part most people are unsure about before they start. Decide it does not hold up and we refund that month in full. You do not have to justify it and we will not argue about it.
What it is not: one month is too short for a geo holdout, so you will not have causal proof per channel yet. That is exactly why this covers your first invoice and not your result after six months. Reports and model output you already received are yours to keep, refund or no refund.
Built by someone who’s been in your seat.
Jeroen Oosterveld previously led data and growth work inside Coolblue, one of the largest e-commerce companies in the Benelux. Datacompany exists to make the same caliber of analytics available to businesses that do not have a hundred-person data team. A small group of senior freelancers supports the build.
Connect on LinkedInJoin the beta. We onboard a few companies each month.
Tell us a little about your business. We will reach out personally when your slot is ready. If your media budget is below roughly €50k a month, say so here: we will look at whether a mix model can tell you anything yet before either of us spends time on it.
Prefer to call? +31 50 211 5978