The Attribution Problem Wasn’t a Dashboard Problem

At Neighbor, a long-standing attribution gap could not be solved by optimizing campaigns harder. Meta, first-touch reporting and warehouse analysis were answering different questions. I led the initiative from Marketing in close partnership with Product Enablement and Data to audit the event, the value and the attribution logic before seasonal scaling.

An experience I am grateful for

I am deeply grateful for my experience at Neighbor and for the opportunity to work with smart people across Marketing, Product Enablement and Data on one of the most consequential measurement questions of my career. Paid host acquisition did not have a campaign problem. It had a truth problem. Platform reporting could show strong performance while internal reporting told a more conservative story. Neither system was necessarily broken. They used different attribution windows, different definitions of conversion and different views of value.

I started with the business question, not the channel question

The first question was not whether Meta was taking too much credit. It was: what outcome is the business actually willing to pay for? An early host activation event was being asked to carry several meanings at once. It functioned as a campaign conversion, a modeled revenue proxy and an implied statement of host quality. Each choice could be defended alone. Together, they created an illusion of precision. More spend does not fix that. More spend amplifies it.

The most important audit was the event itself

Advanced Meta work begins before campaign structure. An algorithm can only optimize toward the signal it receives. If an event is too early, too broad, duplicated, delayed, poorly matched or economically mispriced, the platform can become very efficient at finding the wrong thing. I brought the business question and the initial diagnosis. Product Enablement and Data engaged with the hypothesis, helped trace the signal across systems and worked with me to correct the event, value and reporting logic.

Then I challenged the LTV assumption

The marketplace was not economically uniform. Hosts differed in listing potential, market demand, speed to first reservation and long-term economic value. An average can be mathematically correct and strategically wrong when it flattens the differences that should guide capital. The goal was not to pretend lifetime value could be known on day one. It was to make the proxy more honest, calibrate it against cohort behavior and create a clearer hierarchy between an early signal, qualified supply, a successful reservation and realized value.

Attribution was not one number

I stopped treating attribution as a contest in which one dashboard had to win. Meta was useful for real-time optimization. First-touch reporting helped explain initial discovery. Warehouse analysis served cohorts and long-term economics. Incrementality answered whether the channel created outcomes that would not otherwise have happened. Those are different jobs. The operating answer was to define which source governed which decision and to distinguish modeled value from realized value.

Fix the ruler before increasing spend

The timing mattered because seasonal investment was approaching. Changing event definitions, value logic or attribution interpretation in the middle of a peak would have made it difficult to separate market movement from measurement movement. We fixed the ruler, established the baseline and then scaled. The benefit was not a prettier dashboard. It was better capital allocation, a more credible ROAS conversation and a measurement system the business could trust.

Leadership made it a three-team effort

I led the initiative from Marketing, but the result belonged to a three-team collaboration across Marketing, Product Enablement and Data. My role was to see the pattern, frame the problem, bring the right teams together and keep the work moving until the business had an answer it could trust. The teams engaged, investigated and helped correct the system. That is what cross-functional growth leadership should make possible.

Platform fluency is not the same as growth leadership

Campaign management asks which audience to test, which creative to pause and how to improve delivery. Those skills matter. Growth leadership asks whether the platform is optimizing for a conversion the business truly values, whether the revenue signal represents quality and whether the measurement system is stable enough to support scale. Operating Meta matters. Knowing when the real constraint sits outside Meta matters more. I know Meta deeply enough to know when the problem is not Meta.

The principle I carry forward

Before asking an algorithm to scale, make sure it has been taught what value means. Audit the event. Audit the value. Audit the attribution model. Decide which system governs which decision. Then scale with conviction. My job is not to make Ads Manager look good. It is to help the business put capital in the right place. I am proud of the growth we delivered and grateful for what Neighbor allowed me to learn. Certain proprietary implementation details have been generalized.

FAQ

Why can Meta and internal reporting disagree?

They may use different attribution windows, conversion definitions, identity matching and views of value. A disagreement does not automatically mean either system is broken.

What should a leader audit before scaling paid media?

Audit the conversion event, the value passed with it, matching quality, attribution logic and the relationship between the early signal and downstream economics.

Did attribution alone produce Neighbor’s growth?

No. Strategy, creative, audience learning, funnel work and execution all mattered. Measurement made the growth more trustworthy and improved the decision about where to invest the next dollar.

Who led and implemented the work?

Stephany led the initiative from Marketing in partnership with Product Enablement and Data. The investigation and correction became a cross-functional, three-team effort.

A note when there’s something worth sharing.

New articles, things I’m building and the occasional lesson from the work. I’ll write when I have something worth sending.