The Agency That Deleted Traffic From Its Own Reports

There is a slide in almost every agency’s monthly deck. Sessions, up and to the right, pulled from Ahrefs or Semrush or Analytics, with a percentage change in green. Clients have been trained to expect it. Agencies have been trained to produce it.

Jan Cankusic stopped producing it.

His Tampa firm, Animal House Marketing, reports two numbers to clients and nothing else: manually verified rankings for a named list of commercial keywords, and the count of new customers who told the front desk they found the business online. Traffic does not appear anywhere in the report. Neither does domain rating, impressions, or share of voice.

The firm works exclusively in healthcare, but the reasoning behind the decision is a general argument about attribution that applies to anyone buying organic search, and it is worth examining on its merits whether or not you work in that vertical.

The case against the traffic slide

Cankusic’s objection is not that traffic is unmeasurable. It is that traffic is almost always produced by the wrong pages.

“High traffic numbers come overwhelmingly from informational content,” he said. “Those pages are not location-specific, so they rank nationally. You can win one, get a few thousand visitors a month, and almost none of those people are anywhere near the business.”

That decoupling is the core problem. For a local or regionally bounded business, a national audience is not a smaller version of the right audience. It is a different audience entirely, and a report that counts it is measuring an outcome nobody is paying for.

His second objection is about tooling. The third-party platforms most agencies report from cannot reliably see the specific, localized, high-intent queries that convert. So the agency reports the metric the tool can produce rather than the metric the client cares about, and over time the deck optimizes for the tool instead of the business.

This is a familiar shape to anyone who has run B2B demand generation. MQL counts behaved the same way for a decade: easy to produce, easy to grow, loosely connected to revenue, and politically difficult to abandon precisely because everyone had agreed to look at them.

What replaces it

The substitute is less elegant and considerably harder to fake.

Rankings are checked by hand in Google’s Anonymous Ad Preview, a tool that lets the operator set a geographic location and see the first page of results as a searcher in that city with no account history would see them. No personalization, no logged-in bias, no estimated position averaged across a country.

The step that matters most is sequencing. Before any work begins, the firm builds the list of commercial queries a target customer would actually type when ready to buy, runs the location-set check on every one, and records the starting position. That baseline is the control.

“Because the baseline exists, we cannot take credit for a ranking the client already had,” Cankusic said. “The conversation is: here are the valuable queries, let us check them together, you are not ranking, are these the customers you want, yes, now let us go get them.”

Clients are given the method and the link so they can run the same check themselves, which is an unusual amount of auditability to hand someone who is paying you.

The firm applies the identical procedure to AI assistants. A target query is rewritten as the sentence a person would actually type into ChatGPT, opened in an incognito session logged into no account, and the response is checked for whether the client appears among the three to six providers the model typically names. Baselined before, re-checked after publication.

That last point is where the method earns the most attention right now, because AI answers compress the available real estate dramatically. A conventional results page shows ten listings and a map, so a mid-ranked business is still visible. An AI response names a handful of options in one paragraph and there is no second page to reach. Being adjacent to the answer is worth nothing.

The second reported number is attribution at the point of sale. Every client of the firm is required, as a condition of the engagement, to add a “how did you hear about us” question to its intake process. Only the people who name online search get counted.

“If a client will not track that, I would rather not take the account,” Cankusic said. “Otherwise in six months we are arguing about a graph instead of counting customers.”

The commercial structure that follows from it

A measurement philosophy this specific tends to force changes in how work is sold, and it has here.

Operating as a medical search engine marketing company, the firm sells a $97 entry test rather than the multi-month agreements standard in the sector. The client selects one query cluster it does not currently rank for, receives finished pages within 24 hours, publishes them, and reviews results on day 10. If no keyword in the cluster has reached page one of Google or earned an AI recommendation by then, the fee is refunded with $500 in service credits. Ongoing work starts at $495 per month on a month-to-month basis, with a 60-day refund provision on qualifying packages under which the firm restores the client’s site to its prior state.

Roughly 95 percent of clients operate without a contract. “We have barely written one and have never enforced one,” Cankusic said. “If you need a contract to retain a client, the product is not good enough.”

The firm reports results including a practice that reached page one of Google in five days and the top citywide position for its core terms by day 49, now attributing three to five additional new customers per day to search, and a client with only 15 reviews that reached second position citywide in 35 days and subsequently canceled a $1,500 per month paid search campaign producing fewer than three qualified leads monthly. These are the firm’s own figures rather than independently audited ones.

The transferable part

Strip out the vertical and two things survive that any buyer of organic search can apply this quarter.

Make your agency baseline every target query before work starts, in writing, with the method documented. An agency that will not establish a control before the experiment is reserving the right to claim credit for what already existed.

Then decide what counts as a result before anyone reports one. If the answer is a number your finance team already tracks, the reporting conversation is effectively over. If the answer is sessions, you have agreed to spend a year discussing a metric that does not appear on your income statement.

Guest Author