The short answer
Attribution follows tracked people and assigns credit to what happened. Incrementality compares people who saw your ads with a comparable group who did not, to estimate what the ads caused. Marketing mix modelling uses aggregate spend and sales over time with no per-person tracking. They answer different questions, large teams use all three, and every one of them depends on clean inputs.
DataCops is the tracking solution for ad-funded businesses: it keeps the click, filters junk visits, matches the sale and sends it back to Meta, Google Ads, TikTok and LinkedIn, so your ads learn from real outcomes and not just from what a browser managed to report.
How DataCops covers the attribution chain:
- It keeps the click. One script and one DNS record collect on your own domain. Click IDs (gclid, wbraid, gbraid, fbclid, ttclid, li_fat_id) stay on the server for up to 90 days, and a signed server-set cookie lasts up to 400 days where enabled.
- It checks the visit. Every visit gets a bot verdict against an IP database covering 360+ billion IPs, with a Real people only switch per ad platform (off by default), and form emails are checked.
- It matches the sale. By click ID, or by hashed email and phone, from Shopify orders, HighLevel, any CRM by webhook, or a CSV.
- It sends the sale back. To Meta, Google Ads, TikTok and LinkedIn, counted once against the pixel, with every send in a delivery log.
- It sends the stages that happen later. Booked, showed, won and paid, so the ad platform learns who becomes a customer.
- Ads Warmup. Upload past customers (only email is required, up to 20,000 rows), see a 0 to 10 match score, and send them to Meta, Google Ads and TikTok so a new campaign starts with real buyers.
Best for: stores and lead-gen businesses that want the ad platforms themselves to learn from real sales. It is not a reporting dashboard.
The three methods in plain words
Attribution traces people. A visitor clicks an ad, buys, and the tool credits the ad. It is the most detailed view and the most affected by missing data: a lost click, a blocked pixel or a bot visit all distort it.
Incrementality testing runs an experiment. You hold ads back from a comparable group, a set of regions for example, and compare what each group bought. The difference is what your ads caused. It needs no tracking of individuals.
Marketing mix modelling (MMM) uses statistics. It takes weeks or months of spend by channel and sales, and estimates how much each channel contributed. It needs no tracking of individuals either, and it can include offline media and seasonality.
What question each answers
| Method | Question | Good for | Limit |
|---|---|---|---|
| Attribution | Which tracked touch gets credit for this sale? | Daily and weekly decisions, ranking ads and campaigns | Cannot tell what would have happened anyway |
| Incrementality test | Would these sales have happened without the ads? | Checking whether a channel really works | Takes time and volume; tests one question at once |
| Marketing mix modelling | How much does each channel contribute overall? | Large budgets, offline media, yearly planning | Needs history and an analyst; slow to update |
This is why the three are not rivals. Attribution is fast and detailed but can credit sales that would have happened anyway. Incrementality is slow but answers cause. MMM sees everything at a coarse level. Together they cross-check each other.
How an incrementality test works
- Choose what to test. One channel or one campaign, not the whole mix.
- Split into comparable groups. Regions that behave alike, or a random holdout of an audience.
- Run ads in one group and hold them back in the other for several weeks, long enough to cover your sales cycle.
- Compare sales in each group, from your own store or CRM, not from the ad platform.
- Read the difference as the lift. Compare it with what attribution credited.
Meta and Google both offer conversion lift studies inside their platforms, which handle the split for you. A regional test needs nothing but your own sales data, which is why it is a good first test. It is also useful when the platform cannot see your sales, as in a restricted health account.
Marketing mix modelling
MMM looks at how sales moved when spend on each channel moved, controlling for things like seasonality and promotions. Because it uses aggregate data, privacy limits and blocked pixels do not hurt it. Because it needs a long history and statistical skill, it usually fits companies with large budgets, several channels and some offline media. It tells you where the next budget should go at a coarse level, not which ad to turn off on Tuesday.
Which to use when
| Situation | Start with |
|---|---|
| Small budget, one or two channels | Clean attribution plus your own revenue, and a simple "how did you hear" question |
| Several paid channels, medium budget | Add an attribution dashboard and one regional test a year |
| Large budget, offline media | Add marketing mix modelling and regular lift tests |
| Long sales cycle through a CRM | Stage-based attribution first, then tests |
What they all need: clean inputs
This is the part that gets skipped. Whatever method you choose, it reads your data, and bad data gives confident wrong answers.
- A sale recorded once, not twice because a pixel and a server both reported it.
- A real source, not one lost to a redirect or overwritten by a later touch.
- Bots out, so a fake session is not credited to your best channel.
- Revenue that matches the store or CRM, so a test compares to the truth.
That chain is the subject of the complete attribution guide.
The cheap fourth method: ask people
A single question at checkout, "How did you hear about us?", adds a self-reported source that tracking cannot see: podcasts, word of mouth, a friend's post. It is imperfect because memory is, but it often reveals channels a dashboard shows as direct. Read it alongside tracked data, not instead of it.
Where DataCops fits
DataCops is not an incrementality or modelling tool. It keeps the click on your own domain, checks each visit, matches the sale, and sends it back to Meta, Google Ads, TikTok and LinkedIn with a log. Those clean inputs are what attribution, tests and models all lean on. It also helps the ad platforms themselves learn, which none of those methods do on their own.
Tools you will meet
Each row says what the type does and where it stops. We checked the named tools' own pages in October 2026; vendors change features, so confirm before you choose.
| Type | What it does | Best for | Limit |
|---|---|---|---|
| The ad platform's own reporting | Credits sales by its own rules and window | Ranking ads against each other | Overstates across platforms; sees only what it was sent |
| GA4 or first-party analytics | Sessions and paths on your site | Behaviour and funnels | Browser-only tools miss blocked and non-consented events |
| Attribution dashboards, such as Hyros, Triple Whale, Northbeam and Cometly | Journeys and revenue by channel; several also send conversions back | Cross-channel budget decisions | Check what each sends, to which platforms, and whether bots are filtered |
| Call tracking, such as CallRail and WhatConverts | Ties a call to a source | Phone-led sales | The call is one stage, not the sale |
| DataCops | Keeps the click, filters junk, matches and sends sales and CRM stages, with a delivery log | Making the ad platforms learn from real outcomes | Not a reporting dashboard |
What else the same install does
The attribution chain is one part of a wider setup. Collection runs on your own domain with one script and one DNS record; with your DNS on Cloudflare, the free, optional DataCops Cloudflare Worker reads click IDs and UTMs at the edge before the page runs, and it captures without blocking. On Shopify, the DataCops Shopify app sends every paid order from Shopify's server. A first-party consent manager built on IAB TCF v2.2 and first-party analytics, a GA4 alternative, come from the same script. Each form email is checked for disposable providers and domains with no mail server, and LeadCops can enforce those rules on a form. Every send appears in a delivery log with the reason it was sent, held, skipped or failed.
Sources
- developers.facebook.com/docs/marketing-api/conversions-api
- support.google.com/google-ads/answer/3123169
- ppc.land/meta-rewrites-click-attribution-rules-finally-aligning-with-google-analytics/
FAQ
What is the difference between attribution and incrementality?
Attribution assigns credit to what happened, based on tracked journeys. Incrementality estimates what your ads caused, by comparing people or regions who saw ads with a comparable group who did not. Incrementality answers whether they would have bought anyway.
What is marketing mix modelling?
A statistical method that uses aggregate spend and sales over time to estimate each channel's contribution, without tracking individuals. It suits large budgets and offline media and usually needs an analyst.
Do small businesses need incrementality or MMM?
Rarely at first. Start with clean attribution, your own revenue and a simple regional test or a "how did you hear about us" question. Add formal tests when budgets are big enough to justify them.
How do I run a simple incrementality test?
Pick comparable regions, run ads in some and pause them in others for several weeks, and compare sales. Meta and Google also offer conversion lift tests inside their platforms.
Can these methods disagree?
Yes, and that is useful. Attribution may credit a channel that incrementality shows would have converted anyway. Teams compare and trust the disagreement over any one number.
Is DataCops an incrementality or MMM tool?
No. DataCops keeps the click, filters junk, matches sales and sends them to the ad platforms. Its clean data can feed those methods, which depend on correct inputs.
What does a post-purchase survey add?
A self-reported source, which can reveal podcasts, word of mouth and other touches that tracking cannot see. Memory is imperfect, so read it alongside tracked data.
