Most teams measure programmatic output by counting pages. That is the wrong unit. A hundred thousand published URLs tell you nothing about whether the work earns trust, traffic, or revenue. Quality is the only number that compounds, and it is harder to read because it hides inside aggregates. This is how we read it on the directory sites we run, and how you can build a scorecard that tells you the truth before a search engine does.
Why page counts lie
A template stamps the same structure across thousands of rows of data. When it works, you get scale that no hand built site can match. When it fails, you get the same failure repeated thousands of times, and the aggregate metrics smooth it into a flat line that looks fine. Sessions hold steady because a few strong pages carry the set. Average position drifts down so slowly that no one panics. Then a core update lands and half the template vanishes in a week. The decline was always there in the per template data. The page count hid it.
The fix is to stop trusting totals and start reading distributions. Group every URL by the template that produced it, then look at how that group behaves over time. A good template has a tight, healthy distribution. A bad one has a long tail of pages that never got indexed, never got a click, and never should have shipped. Our wider view on what separates the two lives in the quality bar for auto built pages, which pairs closely with the scoring work here.
The four signals that matter
1. Indexation that holds
Indexation is the first vote a search engine casts on quality. If you submit fifty thousand pages and the index settles at fifteen thousand, the engine is telling you that thirty five thousand of them are not worth keeping. Watch the curve, not the snapshot. A template that climbs to full indexation and stays there is earning its place. One that gets indexed in a burst and then bleeds out over the following months is being quietly demoted. Read indexation by template and by the date a batch was published, so you can tie a decline back to a specific change in the build.
2. Unique value per page
This is the signal everyone claims to care about and few actually measure. The honest test is simple. Strip the template chrome, the nav, the boilerplate intro, and the shared footer, and ask what is left that this exact page gives the reader that no sibling page gives. If the answer is one swapped city name and a number pulled from the same table, you have a doorway, not a page. We quantify this as the ratio of genuinely page specific words and data points to total words. Templates with a thin ratio are the ones that get hit first.
3. Engagement from real people
Rankings can be gamed for a while. Attention cannot. Once a page gets impressions, the behaviour of the people who land on it is the cleanest read you have. Are they scrolling, clicking through to a listing, using the on page filters, or bouncing straight back to the results. You do not need a perfect attribution model. You need to know which templates produce pages that hold a visitor and which produce pages that visitors abandon. The second kind drags the whole domain down over time.
4. Query coverage
The last signal is breadth. A programmatic set exists to answer a family of related queries at scale. Coverage asks how much of that family you actually win. Pull the queries your template ranks for, compare them against the intent map you built before you wrote a line of code, and find the gaps. Low coverage usually means your data is too shallow to satisfy the longer, more specific searches that are the whole reason to go programmatic in the first place.
Building the scorecard
Put the four signals in one view, one row per template, and review it weekly. The point is not a single composite grade. The point is to see which signal is failing for which template, because each failure has a different fix. Thin indexation is a value problem. Weak engagement is a usefulness or a relevance problem. Poor coverage is a data problem. When you read them together you stop guessing and start treating the template like the product it is.
- Indexed share by template: indexed pages over submitted pages, trended monthly.
- Unique value ratio: page specific words and data over total words, sampled.
- Engaged sessions: the share of landings that scroll, filter, or click a listing.
- Coverage: ranking queries over the target query family you mapped up front.
The weekly hand sample
No dashboard replaces reading the pages. Every week, pull a random handful of live URLs from each template and read them as a visitor would, on a phone, cold, with no knowledge of how they were built. You are looking for the small failures that aggregates never surface: a missing data field that leaves a sentence dangling, a heading that promises something the body never delivers, two pages that are near twins because the source rows were nearly identical. Ten minutes of this catches problems that a month of metrics would only hint at. When the sample turns up a pattern, it is almost always a template fault, and fixing the template fixes it everywhere at once.
Tie quality to money, not vanity
The final discipline is to connect quality to the outcome the site exists for. On a directory, that is listings viewed, enquiries sent, or advertising inventory that gets seen. A template can rank and still be worthless if the pages it builds never move anyone toward the thing you sell. We weight our review toward templates that produce both rankings and downstream action, and we are willing to prune a template that ranks but converts nothing. This is the same operator logic we apply across the portfolio, set out in how we build: scale is only worth having when each unit of it earns its keep.
Quality measured this way becomes an early warning system rather than a post mortem. You see the thin template before the update punishes it, you see the coverage gap before a competitor fills it, and you see the engagement problem before it drags the domain. For the failure modes this scorecard is designed to catch early, read when programmatic SEO backfires, and start the practice now, while your set is small enough to fix.
We score every template against the Kings Hospitality Group Page Value Test: a page earns its place only if it answers the query faster than the result above it, carries data a competitor cannot copy in an afternoon, and would still be worth publishing if no search engine existed. Across our directories, templates that fail two of those three rarely hold rankings past a refresh.
Common questions
What is a good indexation rate for programmatic pages?
There is no single number, but a healthy template usually holds most of its submitted pages in the index after a few months. A steady decline is the signal that value per page is too thin.
Should I measure quality per page or per template?
Per template first. Programmatic pages share a skeleton, so a weak template fails the same way thousands of times. Fix the template, then sample individual pages to catch data gaps.