See how all your channels work together, beyond clicks
A cross-channel measurement platform built for marketing managers.
- Understand what actually drives results
- Pay as you go, no sales contracts
- Get answers, not dashboards you never use
- Built for actions, not just insights
20 min
From signup to knowing what works
See which channels drive results across Meta, Google, TikTok, YouTube, AI search and more.
0
Engineers needed
Upload your data and run cross-channel analysis yourself. Nothing to install, nothing to maintain.
8 clicks
From connecting data to cross-channel results
Understand how your channels work together and get budget recommendations you can act on this week.
The problem we solve
Measurement tools were built for a web where people clicked. Five things we do differently.
Affordable
Traditional attribution and MMM tools cost thousands per month, pricing out the businesses that need them most. MarSci is pay as you go.
Easy to implement
Complex setup requiring data scientists, months of configuration, and ongoing technical maintenance. MarSci runs on daily aggregates and gives results in 20 minutes.
Transparent
Black box algorithms. Most platforms do not show how results are calculated. You just have to trust them. MarSci is open source and shows every range.
Built for the new era
Users do not click anymore. Audio ads, AI search results, video first channels: click based tracking misses it all. MarSci measures channels that never get a click.
Privacy first
Cookie dependent tracking is increasingly blocked by regulations and browsers, making data less reliable every day. MarSci uses no cookies and no personal data.
This is what you get back
Your analysis reports Platform CPA reflects reported conversions. Causal CPA is the true cost of the conversions a channel caused. Lowest causal CPA in the account 665 causal conversions on $632.63 of spend. Small budget, punches above its weight 2.39% of paid conversions on 0.24% of spend. Highest causal CPA in the account 89 causal conversions on $40,264.55 of spend. Media mix model on a second account: 6 channels, two years of weekly spend, conversion pay_active_customer. Weekly. The model reproduces the past before we trust what it says about the future. Conversions that would have happened without advertising. A channel whose conversion share beats its spend share is pulling more than its weight. Extra conversions as spend changes. Dot is current spend, dashed is beyond it. Flattening curves are saturating. Bubble size is spend. Higher marginal ROI means the next euro still works. Decisions should be based on causal CPA. Ad platform: each platform’s own conversion count. MarSci estimation: the conversions the model finds each channel caused, over the same days. What you are looking at. Facebook reports a $7.79 cost per purchase. The causal model finds $452.57, because most of those purchases would have happened anyway. Google non-brand search reports $57.97 and actually costs $0.95. That is the difference between a dashboard and an answer. No model settings, no tagging, no tracking code. Connect your platforms or upload a CSV, and the toolkit sets up the models, runs them and emails you when results are ready. Questions about your numbers go straight to the people who built the models, in a chat inside the toolkit. No ticket queue, no account manager in between. Happy to walk you through your Performance Max numbers right now. MarSci has an MCP server, so you can question your results from any MCP-compatible AI assistant, in plain words, without opening a report. You: Where should next month’s extra budget go? Google Search non brand. Causal CPA $0.95 against $57.97 reported, with 28.95% of paid conversions on 0.93% of spend.MarSci
Which paid channels are worth the money?
Google Search non brand
UNDERVALUEDAudience Network
UNDERVALUEDFacebook
OVER-ATTRIBUTEDCost per acquisition by paid channel ?
Move budget out of Facebook and into Google Search non brand.
The platform reports a $57.97 CPA, the causal model finds $0.95. This is where the next budget goes.
Reported CPA is $7.79. Most of what Facebook claims would have happened anyway.
Causal conversions: 146 (6.38% of paid conversions, total 1,299)
Performance Max: 3.09% of conversions on 12.92% of spend. Causal CPA $124.18 against $20.46 reported. Watch it. Pmax reports six times more conversions than it causes.
Audience Network: 2.39% of conversions on 0.24% of spend. Causal CPA $3.01 against $7.86 reported. Undervalued.
Messenger: 1.37% of conversions on 1.68% of spend. Close enough. No action.Expected vs. actual conversions ?
Contribution by baseline and media ?
Spend and conversion share by channel ?
Response curves, top 5 by spend ?
ROI vs. effectiveness ?
ROI vs. marginal ROI ?
Platform-reported against causal CPA
AD PLATFORM MARSCI ESTIMATION CHANNEL REPORTED CPA CAUSAL CPA DELTA CPA INTERPRETATION Google Search brandGoogle Ads $3.07 $5.31 +$2.23 Over-attributed Google Search non brandGoogle Ads $57.97 $0.95 −$57.01 Undervalued Performance MaxGoogle Ads $20.46 $124.18 +$103.72 Over-attributed Audience NetworkMeta $7.86 $3.01 −$4.85 Undervalued FacebookMeta $7.79 $452.57 +$444.78 Over-attributed InstagramMeta $6.95 $112.89 +$105.94 Over-attributed
George, MarSci founderOnline nowYou connect the data. Everything after that is on us.
Connect your data and you are done
Talk to a person, live

George and BoiOnline nowAsk your AI assistant through MCP
Our approach
We believe macroeconomic models should not be only for Fortune 500 companies. Every marketing manager deserves to see how their channels work together.
We believe marketing measurement should be privacy first. That is why we only use aggregated information. Nothing to integrate, no personal data, no cookies.
We believe you deserve to know what to do next. Not another dashboard. A clear recommendation you can act on this week.
# Daily aggregates. That is all we need.
date, impressions_facebook, spend_google, conversions_total
2025-10-01, 128300, 1540.22, 413
2025-10-02, 120911, 1320.10, 399
How it works
Four steps. Only two of them are yours.
Log in
Create an account and log in. No credit card, no sales call, no setup project.
Connect your data
With an API connection to Meta, Google, TikTok, Shopify and more, or without one: upload a CSV of daily aggregates.
We run the models
Attribution and the marketing mix model run on our back end. Nothing to install, nothing to configure.
You get notified
When everything is set up you get an email and your results are waiting. You do not have to do anything in between.
Why MarSci
Multiple models, not just one
We are open-source and we let you run different measurement approaches side by side. No single model that only benefits us. You compare, you decide.
Designed for actions
Most tools give you charts. We give you clear recommendations: which channels to increase, which to cut, and how much to shift. Built to change what you do on Monday morning.
Results in minutes, not months
Upload your data and get answers straight away. No onboarding calls, no setup projects, no waiting for a data team. You are in control from day one.
What our partners say
“We were looking for a media mix analysis for a while. As a travel company and with a complex business model, none of the available solutions could answer our business needs. MarSci managed to not only offer a smooth onboarding process but also to answer which marketing channel is more impactful to our very long booking window.”SStef, Co-founder and Managing Director, travel company
“MarSci has transformed the way we measure and manage our marketing performance. For the first time, we have a clear view of how our online campaigns and offline sales activities work together to drive results. Their platform helped us identify new opportunities, optimise our budget allocation, and make data-driven decisions with confidence. It has been a true step change in how we approach growth.”MGMarketing and Growth, e-commerce and offline sales
Your call is with the people who built the models
There is no sales team at MarSci. When you book a call you speak with the founders, on your channels and your numbers.

George, PhD
Founder. Marketing scientist

Boi, PhD
CTO. AI engineer
Your setup 5 min
Which channels you run, what data you already have, and the decision you are trying to make.
Live walk-through 15 min
We open the toolkit on a real dataset: the journey map, causal attribution, the mix model, and what the recommendations look like.
Your first run 10 min
What to upload, what you get back in 20 minutes, and whether it is worth your time. If it is not, we tell you.
- You get a PhD statistician who has sat on the platform side at Facebook and TikTok, not a sales script.
- You leave with a plan for your first run, not a quote.
- No contract afterwards. Pay as you go if you use it, nothing if you do not.
Prefer email? info@mar-sci.com
Start measuring what matters. No credit card, no sales calls, no code needed.
Log in now or reach out for a demo.
