Voices · · 7 min read
Why Your Attribution Model Might Be Measuring the Wrong Customer
Every attribution model assumes the person who converts is the person who drove the decision. In businesses where the buyer and the user are different people, that assumption quietly misdirects your budget. Here is how I found it at Tynker, and a three-step audit you can run.
Growth & marketing leader
Every attribution model rests on an assumption almost nobody states out loud: the person who converts is the person who drove the decision.
In a lot of businesses, that assumption is simply wrong. My claim is this: when the person who pays is not the person who uses, your attribution model is probably measuring the wrong customer, and it steadily moves your budget away from the thing that actually creates demand.
Forrester's 2026 Buyer Insights research, released on January 21, 2026, found that a typical B2B purchase decision involves 13 internal stakeholders and nine external influencers, and that more than 60% of business buyers use a trial before they commit. You can read the Business Wire release, the same release on Forrester's investor site, and Digital Commerce 360's coverage from January 22, 2026. That is a crowd, not a single converter. Yet most models still hand the credit to one name on one touchpoint.
The growth world has known a version of this for years. Product-led growth, a term Blake Bartlett of OpenView coined in 2016, is built on the sequence where someone uses the product first and the purchase follows. The person who experiences the product and the person who signs the invoice are often not the same.
Most teams will agree with that in a strategy meeting. Then they build their attribution model as though the buyer and the user were one person anyway, because the buyer is on the invoice.
At Tynker, that gap was not a theoretical nuance. It was the entire business.
“The kid was the user and the champion. The parent was simply the one who paid.”
Who pays, who uses
Our earliest growth strategy targeted parents directly. It felt like the obvious, defensible move. It was also the slower one, and it took us longer than it should have to see why.
| Targeting the Buyer | Targeting the User | |
|---|---|---|
| Who the message is built for | The parent, the one paying | The kid, the one experiencing the product |
| What drives the eventual purchase | A direct pitch about value | The kid's own enthusiasm and advocacy |
| What attribution captures | The parent's last touchpoint | Nothing, by default |
| What actually converts faster | A slower, harder sell | A demand the parent did not have to be sold on |
The better strategy was to reach kids first. We met them through Minecraft and other content they already loved, built real engagement, and let them become champions inside their own households. Once a kid was the one asking a parent for more, the upsell to a paid subscription became dramatically easier to close.
The kid was the user and the champion. The parent was simply the one who paid.
That pattern showed up in how kids found us, too. As I described in an October 26, 2023 interview with the eMarketing Association, kids often arrived through their love of Minecraft, and once they started building their own modifications, that kicked off a flywheel around creativity, helped along by a community where kids help each other.
Why the dashboard could not see it
Here is the part that matters for anyone running a measurement stack.
Last-touch and most conventional multi-touch models are built to assign credit to whatever channel acquired the parent, or whatever touchpoint came right before the purchase. Both of those point at the buyer.
Neither one captured the loop that was actually driving the business: acquire the kid, build engagement, turn that engagement into advocacy, let the advocacy create demand in the household, then convert that demand into a subscription. The purchase was just the last visible step in a chain of causation that started weeks earlier, with a person the attribution model was not even looking at.
This is worse than a blind spot, because of what it does to budget.
When the model credits the parent-facing channel that happened to sit closest to the purchase, that channel looks efficient. It gets more money. Meanwhile the content and community work that created the demand shows up as cost with no return, so it gets trimmed. The model does not just miss the real driver. It steadily moves budget away from it.
If you have ever watched a team cut the thing that was working and scale the thing that was merely nearby, this is often the mechanism behind it.
What fixed it
The fix was not a better attribution tool. It was accepting that we were measuring the wrong person.
The turning point came from pairing the multi-touch data we already had with a simple question asked directly at the moment of conversion: how did you hear about us?
In my experience the answers were consistent, and consistently unhelpful to the existing model. Parents described a path that ran through their child's enthusiasm, not through any channel a dashboard would have credited. Content and community had driven the actual discovery, even when no click, no ad impression and no trackable touchpoint had occurred anywhere in the parent's own journey.
“The model does not just miss the real driver. It steadily moves budget away from it”
So we rebuilt measurement around the behaviors that created a champion: engagement depth, time in product, and advocacy signals. Those behaviors turned out to be the real leading indicators of a parent's eventual purchase.
That flywheel became a major growth engine, once the team stopped optimising for the wrong side of the transaction.
Where else this shows up
This split extends well past consumer products aimed at children. Anywhere the person who benefits is not the person who pays, the same distortion is waiting.
Enterprise software. An individual contributor discovers a tool and champions it inside the company. Procurement signs the contract months later. Forrester's 2026 Buyer Insights research found that more than 60% of business buyers run a trial before committing, which means someone uses the product before anyone signs for it. Most attribution models credit only the signature, so the person who did the persuading never appears in the data.
Healthcare. In many cases a patient researches a treatment and asks for it by name. An insurer's approval process decides whether it happens. Almost no model captures how much the patient's own pull shaped the outcome.
Education technology. In many districts a teacher adopts a classroom tool on their own. A district administrator controls the budget that eventually pays for it district-wide. The pilot that built the case rarely gets credit for the contract.
Freemium and B2B2C products generally. The free user does the work of proving value. The paying account gets the credit.
In each case, the model is measuring the transaction when the thing worth understanding is the demand behind it.
The cost of measuring the right person
This is not free to implement, and it would be dishonest to pretend otherwise.
Measuring champion behavior instead of purchase-adjacent behavior means living without a clean, immediate return on investment number for weeks, sometimes longer, while the real loop plays out. The early indicators do not look like revenue yet.
That is a harder story to defend in a board meeting mid-quarter than a report of what converted this week, even when the harder story is the accurate one. It takes real trust in the team making the case.
Most organizations drift back to measuring the buyer directly. Not because it is more accurate, but because it is the number a board can react to immediately. If you take one thing from this piece, make it that: the pull toward the easy-to-track person is organizational, not analytical, and it will keep winning unless someone decides on purpose to resist it.
The champion signal audit
Here is the version I would run in any business with a split between the person who pays and the person who uses. Call it what I'd describe as a champion signal audit, a short exercise your team can finish in a couple of weeks.
- Name the buyer and the user separately. If a single persona document describes both, your model has already merged two people who may have almost nothing in common. Write one profile for each: what they care about, where they spend time, and what they do before a purchase happens.
- Ask at the point of conversion. A direct question, how did you hear about us, will surface the real path faster than any attribution platform reconfiguration. Read the free-text answers yourself for a few weeks before you code or categorize them. The language people use will tell you which person was doing the persuading.
- Find the leading indicator on the user's side. Engagement, advocacy or usage depth on the user's side, not the buyer's, is usually the earliest reliable signal that a conversion is coming. Pick one or two that you can actually measure, and track them next to your existing numbers for a full cycle before you change any budget.
- Check where the money is going. Line up your channel spend against the buyer-side credit your model gives each channel. Then ask which spend is feeding the user's experience and which is simply sitting closest to the purchase. Any channel that earns credit mostly by proximity deserves a closer look.
- Agree on the slower number in advance. Before you present the new measures, get your CEO or board to agree that the early indicators are a legitimate basis for decisions. Doing this up front is what keeps the first soft week from sending everyone back to the old model.
The question to take back to your team
The model you use decides what your team can see. If it only sees the person who pays, you will keep funding the channels closest to the invoice and starving the ones that created the wish to buy in the first place.
