We Audited Our Own Reporting and Both Halves Were Broken
This is not a list of things other agencies get wrong. In September 2026 we ran the same audit on our own website that we run for clients, and found that the two numbers a marketing report leans on hardest were both lying to us. If it happened here, with the tooling and the incentive to get it right, it is worth thirty minutes of your time to check yours.
Number One: Leads You Never Saw
Our analytics reported 656 form starts and one recorded lead across the period. The obvious reading is a catastrophic conversion problem. The real cause was that 8 of our 12 forms never reported a submission at all. The tracking helper worked perfectly; most forms simply never called it. Our /practice-website/ landing page had taken 235 sessions across its ad variants and recorded one key event.
Everything downstream of that was wrong. Cost per lead, conversion rate by page, which channel was working — all computed from a number that was missing two thirds of its inputs. We were not converting badly; we were counting badly.
How to Check It in Ten Minutes
Submit every form on your own site, one at a time, and confirm three separate things for each: the email arrived, the analytics event fired, and someone followed up. Most practices have never tested all three in one sitting, and the failure is almost always in the middle one — the form works, the notification works, and the event never fires, so the agency's report shows nothing.
If your site has a phone number, check that too. Ours had 76 telephone links and zero tracking on any of them — for a business whose patients mostly call rather than fill in forms, that is the wrong half of the funnel to be blind on.
Number Two: Rankings That Are Not Rankings
Search Console told us we sat at position 1 to 10 for 149 non-branded queries carrying 6,785 impressions. Any reasonable person reads that as page-one visibility. Those queries produced zero clicks.
We searched four of them by hand, signed out, on a phone. We were not in the top 25 for any of them. A reported average position is not a ranking — it counts appearances inside image carousels, "people also ask" panels, local packs and results sitting below a screenful of ads. A page can be "position 4" in a module nobody scrolls to.
The One Question That Separates a Real Report From a Nice One
Ask your agency to search the term live, in front of you, on a phone, signed out. Not a rank-tracker screenshot — the actual result page a patient would see. It takes ten seconds per keyword and it settles the argument permanently.
If the page is not there, the impressions are real and the position is a phantom, and any strategy built on "we just need to nudge it from 6 to 3" is built on nothing. We nearly rewrote every title and description on our site before running that check.
What Good Reporting Looks Like
Three things, in this order. Leads, with a definition. Not clicks, not impressions, not "engagements" — how many people asked you to contact them, and what event fires when they do. Where they came from, honestly, including the ones that came from your existing patients telling someone. What changed since last month, and why, in a sentence a human can check.
A report that leads with impressions and average position is not dishonest by design; it is reporting the numbers that are easiest to produce. But those are the two numbers that broke on our own site, and we build this software for a living.
Fix the Measurement Before You Change the Strategy
The reason to care is not accuracy for its own sake. It is that every decision after this one — spend more here, cut that, the site needs rebuilding — is made from these numbers. We spent time considering a content strategy for queries we were never actually ranking for, and we could have concluded our forms needed redesigning when they converted fine.
Check the forms. Check a live search. Then decide what to change.
The Questions Worth Asking Your Agency
What are the key metrics to focus on? Leads, cost per lead, and the source of each. Everything else is diagnostic. If a report leads with impressions, reach or “engagements”, those are inputs, not outcomes.
How do I know if my agency's numbers are real? Two tests, both free. Submit your own form and confirm the event fires. Then ask them to search a ranking keyword live, on a phone, signed out, in front of you. Both of those failed on our own site.
What is a reasonable reporting cadence? Monthly is enough for a practice. Weekly reporting on a channel that takes a quarter to move produces noise and panic decisions.
Should I pay for a dashboard? Usually not. A dashboard makes bad data prettier; it does not make it true.
What are the key performance indicators in marketing? For a practice, four: new patient inquiries, cost per inquiry, the share that become booked appointments, and the channel each came from. Impressions, reach, followers and average position are not KPIs — they are inputs, and all four of ours can be wrong if the tracking is.