Measurement
What to count after a page goes live
A handful of counts, read per page, will tell you more than a dashboard full of blended traffic.
After a landing page is public, the useful question is not “how is the campaign doing?” in the abstract. The useful question is whether this page, for this offer, is receiving the visits you paid for or earned, and whether those visits become submissions your team can act on. That question can be answered with a few counts. It cannot be answered by a wall of charts that mix every page, every source, and every definition of a conversion you have ever used.
Keep the grain at the page. A subdomain or URL that carries one offer should have its own view of visits and completed forms. When you blend a spring window campaign with a fall roofing campaign, the better page hides the worse one and you “optimize” a ghost. If your tool only shows a site-wide total, export by page or stop pretending you are measuring the thing you shipped. This article is about which counts earn a weekly look. It is not about buying more analytics products.
Visits, starts, and finishes
Count visits to the page, people who began the form, and people who finished it. You will not always have a clean “start,” and a finish is the one that matters most, but the gap between visits and finishes is the health of the page. A page with many visits and almost no finishes has a broken promise, a broken phone layout, a broken load time, or traffic that was never going to qualify. A page with few visits and a high finish rate may be fine and simply underfed. Those are different problems. A single conversion rate, divorced from the volume, pushes teams to “fix” a page that is merely quiet.
Define a finish as a submission you can see in the system your team actually works from, not as a button click in an ad platform. Button clicks include mis-taps, refreshes, and people who failed validation. The submission row is the event the business can staff. If the ad platform says fifty conversions and the submission list shows thirty, believe the list until you find the missing twenty. The gap is a measurement bug or a filter, and either way it is more important than a new headline.
Quality, not just completion
A finished form can still be a bad lead. Pick one quality mark your follow-up team can apply without a research project: in the service area or not, reached a person or not, request matched the offer or not. Review it on a sample every week, not on every row if volume is high, and not never. The page can be “converting” while filling the queue with people who wanted a different product because the headline drifted from the ad. Completion will not show that. The people on the phones will, if you ask them a yes-or-no question and write the answers down.
Do not invent a quality score with twelve weighted factors on week one. You will not maintain it, and you will start negotiating with the formula instead of reading the calls. One or two marks, defined in a sentence, beat a score nobody trusts. If out-of-area submissions dominate, the page’s qualification copy or the traffic targeting is wrong. If people cannot be reached, look at the phone-field label and at whether you are calling inside the window you promised. The count points at a place to look. It does not fix the page by existing.
Speed and failure
Watch for the dull failures that masquerade as low interest. A page that times out on a phone, a form that errors after the button, a thank-you screen that never appears, a certificate warning: each of these looks like a traffic problem if you only stare at visits. Submit a test on a phone once a week while the campaign is spending money. Count failures of that test as incidents, not as anecdotes. If the test fails, pause the clever work and repair the path. Buying more clicks for a form that drops the submission is a straightforward waste.
Note changes on the same timeline as the counts. A consent-text edit, a new required field, a switch of ad destination, a long weekend when nobody returned calls: write them on the same sheet as visits and finishes. Otherwise you will explain a dip with a story you prefer, such as “the audience got worse,” when the dip started the hour you made the phone number optional and the follow-up team stopped dialing. Annotation is the cheapest analysis you can do.
What to ignore at the start
Ignore vanity totals that you cannot act on this week. Total time on page is easy to misread, because a confused person and an interested person can both linger. Ignore leaderboards of cities unless geography is the qualification and you are prepared to change the page or the targeting. Ignore comparisons to an industry benchmark you found in a roundup article. Your offer, your price of the next step, and your traffic source are not the benchmark’s. Compare this page to itself last week, and this source to the other source hitting the same page.
Ignore any metric that requires you to export customer phone numbers into a personal spreadsheet “just to calculate it.” If the count cannot be produced inside the access rules you already set, choose a coarser count. Measurement is not a reason to spread a lead list across laptops. Aggregate numbers and a small reviewed sample are enough for the decisions in this guide.
A weekly page review
- Visits and finished submissions for this URL only.
- The gap between the ad platform’s conversions and the rows you can open.
- One quality mark from the follow-up team, on a sample.
- One real test submission from a phone, noted as pass or fail.
- Any copy, field, or traffic change since the last review, written on the same sheet.
- A single decision: leave it, fix a named defect, or stop the traffic.
End the review with a decision. A meeting that “keeps an eye on it” for a month is how a broken phone layout survives. If the page is doing the job, leave it alone and do not fidget with the headline out of boredom. If it is not, name the defect in a sentence: the first screen does not show the offer, the form errors on mobile, the clicks are from outside the service area, the finishes do not match the ad. Then assign the sentence to a person. Counting is useful only when it causes a specific change or an explicit choice to keep going.