Your ad platform claims one number. Your analytics tool reports another. Your CRM, the system where your sales and customer records live, shows a third. Every tool is confident. None of them agree.
The usual response is to go shopping for an attribution tool that promises to settle the argument. Most teams do not need one. The disagreement is not a software problem. It is an operations problem, and you can fix most of it with discipline you already own.
Why the numbers will never match
Attribution is the process of deciding which marketing touch gets credit for a conversion, meaning the sale or lead you count as a win. Every tool answers that question with different rules. Different rules produce different numbers. That part is math, not malfunction.
Windows
An attribution window is the amount of time a platform is allowed to look backward and claim credit. One platform may count a purchase if the buyer clicked an ad within the past week. Another may look back a full month. Another may count a view rather than a click, meaning the buyer only saw the ad and never touched it.
Two tools with different windows will report different totals from the same sales. Neither is lying. They are answering different questions.
Models
An attribution model is the rule for splitting credit when a buyer touched more than one channel. Last click gives everything to the final touch. First click gives everything to the opening touch. Data-driven models split credit using the platform's own math, which you cannot inspect.
Here is the mechanic that burns most teams. A buyer clicks a social ad, comes back through a search ad, then converts from an email. The social platform claims the sale. The search platform claims it too. Email claims it as well. Add up the platform reports and you have three sales that were really one.
Deduplication
That triple count is a deduplication problem. Deduplication means making sure one real sale is counted once. Ad platforms cannot do it for you, because each platform only sees its own slice of the journey. Only a system you control, your analytics or your CRM, sees the whole thing.
Timing adds one more layer. Some tools report a conversion on the day of the click. Others report it on the day of the purchase. Pull a report on Monday and the same sale can land in different weeks depending on which tool you ask.
The real fix is boring: naming and UTM discipline
Before you evaluate any attribution software, look at your links and your naming. In most accounts we audit, the mess starts there.
UTM parameters are the small tags added to the end of a URL that tell your analytics where a visitor came from, tags like utm_source and utm_campaign. When they are missing, your analytics files the visit under direct traffic or some vague bucket, and the channel that earned the click gets no credit.
When they are inconsistent, it is worse. To your reporting, google, Google, and google-ads are three different sources. FB_Spring_Sale and fb-spring-sale are two different campaigns. The tracking did not fail. Five people typed tags five different ways, and the reports are faithfully recording the chaos.
The fix costs nothing:
- Write one naming convention for campaigns, ad sets, and ads. Structure it so the name itself tells you the channel, the audience, and the offer.
- Build a UTM template, a shared sheet or link builder that generates the tags, so nobody types them by hand.
- Use lowercase everywhere. Capitalization differences split your data silently.
- Tag everything that points at your site, including email, social posts, and partner links. Not just paid ads.
- Audit the tags on a regular calendar. Naming rot creeps back the moment a new hire builds a campaign from memory.
This is the operational layer we rebuild in a marketing operations audit, and it is the same layer that makes a paid ads system legible months later. Naming and tagging sound trivial. They are the plumbing every report sits on.
Pick one source of truth
You cannot make the tools agree. You can decide which one settles arguments.
A source of truth is the single system your business decisions come from. For most companies it should be the system closest to revenue: analytics tied to actual orders, or the CRM. Not an ad platform, because ad platforms grade their own homework.
Then give every other number a job:
- Platform numbers are for optimization. Inside Google or Meta, use the platform's own reporting to compare campaigns against each other. The rules are at least consistent inside those walls.
- Source of truth numbers are for decisions. Budgets, channel mix, and what you tell leadership all come from the one system you chose.
Write the decision down. One page. Which system is truth, which model it uses, which window applies, and who owns it. The document matters more than the choice. Most attribution fights inside companies are not about data. They are two people quoting different systems, and neither knows the rules the other is playing by.
This is also where reporting that writes itself pays off. When reporting is automated from the source of truth, the weekly number is the same number every time, produced the same way. Nobody rebuilds it in a spreadsheet with their own assumptions baked in.
Living with imperfect attribution
Here is the part nobody selling software will say out loud. Attribution will never be complete. Privacy rules, blocked cookies, people switching devices, a link shared in a group chat. Part of the journey is invisible and always will be.
That is fine, because you do not need perfect. You need consistent.
A consistent, slightly wrong measurement beats a precise number that changes definitions every quarter. If your source of truth undercounts a channel, it will undercount it the same way next month, and the trend line still tells the truth. Direction and movement over time are decision grade. Absolute precision is not on the menu from any vendor at any price.
And when you truly need to know whether a channel works, run a simple holdout. Turn it off in one region or one window of time and watch what happens to real revenue. A crude test against reality tells you more than any model arguing with another model.
So resist the tool. Fix the layer underneath: clean names, disciplined UTMs, one source of truth, one written set of rules. That is how we approach marketing systems in general. Fix the operational layer and the results compound.
Not sure where your numbers are lying?
Our Performance Audit starts exactly here. We trace how data moves from click to report, find where the story breaks, and hand you a dollar-weighted plan to fix it. The plan is yours to keep whether or not we do the work together. Start with a Diagnostic Call.