Google Ads · Multi-Touch Attribution
Multi-touch attribution promises to credit every step of the journey instead of just the last click. The idea is right. The way it is usually implemented quietly counts the same customer several times, because each platform grades its own homework and none of them can see the others. Here is how the models actually work, why the numbers inflate, and what we do instead of chasing fractional credit.
Single-touch attribution hands all the credit to one interaction, usually the last click. Multi-touch spreads it across the path. The common models are first-touch, which credits the introduction; linear, which splits credit evenly across every touch; time-decay, which weights recent touches more heavily; position-based, which loads credit onto the first and last touches; and data-driven, which uses a model to assign fractional credit based on observed patterns. Each answers a real question about which touches move people toward a sale. The trouble is not the models. It is where the data comes from.
Run Google and Meta at the same time and both will claim the same conversion. Google's attribution sees the Google touches; Meta's sees the Meta touches; neither sees the other. So the customer who clicked a Facebook ad, then searched your brand on Google, then converted, gets counted once by Meta and once by Google. Add the numbers up and combined reported revenue can be nearly double what actually landed in the bank. Multi-touch attribution inside a walled garden is not multi-touch at all. It is single-platform attribution wearing a nicer label, and it inflates exactly where you can least afford it: the budget-allocation decision between channels.
The mechanism behind the double count: enhanced conversions for leads →
Google's data-driven attribution is now the default, and it is genuinely more sophisticated than last-click. It is also modeled, proprietary, and unauditable: you cannot see why a touch received the credit it did, and you cannot reconcile it against your own revenue. As the primary basis for moving budget, that is a problem, because you are trusting a model's story about your funnel instead of the outcomes your CRM can actually confirm. Data-driven attribution is a fine input. It is a dangerous single source of truth.
The way out is not a better fractional-credit formula. It is a deterministic spine underneath the models. We tie every touch we can to an explicit click identifier, reconcile conversions against real revenue in the CRM, and deduplicate a customer to a single record across platforms with a shared order ID. Once a conversion is counted once and valued at real dollars, the argument about which model to use gets much smaller, because the total is finally honest. Then we optimize each platform to one number, profit-based ROAS on collected revenue, instead of letting three tools each bid toward their own inflated version of events.
A shared order ID means the customer who touched both Google and Meta is counted once, not once per platform. The combined revenue number finally matches the bank.
Conversions carry collected revenue from the CRM, not form-fill counts. The credit being split is at least credit for something that actually happened.
With one deduplicated, revenue-valued conversion per customer, Smart Bidding on each platform optimizes toward profit instead of toward its own overcount. Budget follows real return.
Because the spine is deterministic and tied to your own revenue, you can reconcile it. The model becomes a lens on top of numbers you trust, not a story you have to take on faith.
On every account we capture click identifiers where they survive, back them with hashed first-party data where they do not, and reconcile each conversion against collected revenue in the CRM. Every event carries a uniform order ID so a customer is counted once no matter how many platforms touched them, and the whole thing flows through one unified pipeline that owns deduplication end to end. The result is not a prettier attribution report. It is a single honest total that each platform's bidding can optimize against without competing to overcount.
The signals underneath this: cookieless tracking → · The tracking layer we build on: Hyros implementation →
A way of crediting more than one interaction in a customer's path, instead of giving all the credit to a single touch like the last click. Common models include first-touch, linear, time-decay, position-based, and data-driven, each of which distributes credit across the journey differently.
First-touch credits the introduction; last-touch credits the final click; linear splits credit evenly across all touches; time-decay weights recent touches more; position-based loads credit onto the first and last touches; and data-driven uses a model to assign fractional credit from observed patterns.
Because each ad platform only sees its own touches and claims the conversion for itself. A customer who interacts with both Google and Meta gets counted by each, so combined reported revenue can be far higher than the revenue actually collected. Without a shared deduplication key, the totals inflate.
It is more sophisticated than last-click, but it is modeled and proprietary, so you cannot audit how credit is assigned or reconcile it against your own revenue. It works well as an input and poorly as a single source of truth for moving budget.
You need a deterministic way to count each customer once and value the conversion at real revenue before you need a fancier model. Fix the double counting and the revenue values first; the choice of attribution model matters far less once the underlying total is honest.
Google Ads has its own multi-touch model, data-driven attribution, and it is now the default on most conversion actions with every other model except last click retired. It genuinely fixes the ordering problem inside Google's inventory. What it cannot do is see the Meta touchpoint, the phone call, or whether the conversion became revenue, which is the double-counting problem this page is about.
Data-driven attribution: what it measures and what it misses →
We measured 203,043 transactions that settled through client merchant and bank accounts over twelve months and asked how much could be tied back to a first-party ad click. Across all accounts, 79.1%. For lead generation businesses, 95.9%. For retail, 63.9%. The gap explains why platform reports and bank records never reconcile.
Next Step
If you are spending $30,000 or more per month across channels and your platforms disagree about what your advertising produced, we should talk.
Request a Proposal →