01
Marketplace
Location × category
Real inventory, real pricing, real availability per combination. The classic case, and the one most often faked.
Needs live data to stay honest.
Programmatic SEO
Scale, but only where scale deserves to exist
Programmatic SEO is the fastest way to build a thousand pages and the fastest way to get a site penalised. The difference is entirely whether each generated page carries information a user could not get from the page above it.
3
conditions a dataset must meet before we build a template at all
1 in 4
programmatic opportunities we assess and then advise against
Per-page
gain check, run before publication and after
The honest version
I
Templates are not the product, data is
A page generated from a thin database row is thin no matter how good the design is. The question is never "can we generate this" but "does each row hold enough distinct, useful information to justify a URL".
II
Index bloat is a real cost
Ten thousand near-duplicate pages consume crawl budget your genuinely valuable pages need. We would rather ship eight hundred pages that all get crawled than eight thousand that get ignored.
III
It only works on real demand
Long-tail volume has to actually exist. We validate demand per template pattern before building anything, and we tell you when the answer is no.
The Three Conditions
Real search demand across the pattern. A dataset rich enough that each page differs meaningfully. And a reason for a human to be glad they landed there.
All three, or we build something else instead. This is the conversation most programmatic pitches skip.
Validated demand
Rich dataset
Genuine user value
How we build it
Data first, template second, scale last.
We model the full URL pattern space, then validate real demand across a representative sample rather than extrapolating from a head term. If only the top ten percent of the pattern has demand, we build ten percent of the pages and put the effort into depth instead.
Pattern modelling
Sampled validation
Cannibalisation check
Most source datasets are too thin on their own. We enrich: joined third-party data, computed comparisons, aggregated user signals, editorial snippets for the highest-value rows. The goal is that every page can answer something the category page cannot.
Data joins
Computed fields
Editorial layer on top rows
Answer-first structure, real internal linking between related rows, schema per page type, and dynamic sections that only render when there is genuine data behind them. Empty states are removed, not filled with filler prose.
Answer-first
Row-to-row linking
Per-type schema
No filler states
We release in cohorts and watch indexation, engagement and query coverage per cohort. Pages that never get crawled or never satisfy a query get consolidated or removed. The library shrinks as often as it grows, and that is the system working.
Cohort release
Indexation watch
Scheduled pruning
Where it genuinely works
Four patterns with enough underlying data to justify the URL count.
02
Saas
Integration and comparison
Product A with Product B, where the setup, limits and gotchas genuinely differ per pair.
Highest commercial intent.
03
Education
Programme and outcome
Course, fee, eligibility, placement data per combination. Buyers compare exactly these fields.
Data freshness is the whole game.
04
Tools
Calculators and lookups
Pages that compute something. The strongest form, because the user gets an answer nowhere else generated it.
Also the most cited by AI engines.
The build
- 01 Fit assessment Three conditions tested against your data and your market. Sometimes the honest answer is no. Week 1
- 02 Data model Source, enrichment plan, refresh cadence and ownership. Agreed before any page is designed. Week 2-3
- 03 Template build One template, tested against twenty real rows including the ugly ones. Week 3-5
- 04 Cohort release Highest-value rows first, in waves, with indexation monitored per cohort. Week 6+
- 05 Prune and extend Cut what fails, deepen what works, refresh the data on schedule. Ongoing
Next
Find out whether scale is even the right move
Show us your dataset. We test it against the three conditions and tell you honestly whether programmatic is worth building, including when it is not.