01
Flagship
Original research
A survey, panel or dataset only you could gather, written up with methodology visible enough to be checked.
The asset that gets cited for years.
Content Systems
An editorial engine, not a content calendar
AI made it possible to publish a hundred articles a month. It also made those hundred articles worth nothing. We build the other kind of content operation: fewer pages, each carrying information the internet did not already have.
< 40
Gain Score below which we refuse to publish a page
1 : 6
ratio of pages we kill to pages we ship, at brief stage
Named
human expert on record behind every substantive claim
Why content programmes fail now
I
The cost of average went to zero
Anything a competitor can generate in an afternoon cannot defend a position. If your content could have been written without your company existing, it does nothing for your company.
II
Refresh beats publish
Most libraries have twenty pages that could rank and two hundred that never will. Fixing the twenty is faster, cheaper and more durable than adding to the two hundred.
III
Editorial systems outlive editors
A calendar depends on whoever is running it. A system, briefs, gain scoring, expert review, refresh triggers, keeps producing when the person changes.
The Gate
Every brief is scored for Information Gain before a word is written. Below the threshold, it does not get commissioned, however good the keyword looks.
Original data, first-hand testing, named expertise, net-new framing, proprietary benchmark. The same rubric we publish openly.
Gain scored at brief
Expert review gate
Refresh triggers
The editorial engine
Four components. Remove any one and the output reverts to a content calendar.
A territory you can credibly own, broken into pillars and clusters with an honest assessment of where you have genuine expertise and where you would be bluffing. We would rather you own three subjects completely than have a shallow opinion about thirty.
Pillar and cluster
Credibility audit
Competitive whitespace
Each brief names the question, the structure, the sources, the specific expert to interview, and the gain target it must hit. Writers are not asked to invent insight from a keyword, they are given the raw material and told what shape to make it.
Question-led
Named expert
Gain target
Source pack
Interviews recorded, data pulled, tests run. The unglamorous collection work is the whole product. Drafting is the fast part, and yes, AI helps with drafting, it just cannot supply the thing that makes the page worth reading.
Practitioner interviews
Data pulls
First-hand testing
Every page has a review trigger: ranking decay, factual expiry, or a competitor overtaking the answer. Pages that cannot be brought back to standard get consolidated or removed. A smaller, sharper library reads better to engines and to people.
Decay alerts
Consolidation
Deliberate pruning
What comes off the line
Four formats we return to, because they hold up under both human and machine reading.
02
High Intent
Comparison and alternatives
Honest, specific, including where you lose. Buyers and models both punish the version where you win everything.
Converts hardest, ages fastest.
03
Authority
Practitioner teardowns
A named operator walking through a real decision, with the numbers. Nothing a model can synthesise.
Interview-led, quote-rich.
04
Coverage
Definitional pages
The clean, liftable answer to a category question. Unglamorous, and the most-cited thing on most sites.
Answer-first by construction.
How the engine gets built
- 01 Library audit Every existing page scored: keep, refresh, consolidate, retire. Usually the biggest quick win. Week 1-2
- 02 Territory map The pillars you can credibly own, and the clusters underneath them. Week 2
- 03 Brief system Templates, gain rubric, expert roster, review gates. Handed over so your team can run it without us. Week 3-4
- 04 Production cadence A sustainable rhythm, usually four to eight substantive pieces a month, not forty. Month 2+
- 05 Refresh loop Decay monitoring and scheduled updates, so the library compounds instead of rotting. Ongoing
Next
Find out what your library is actually worth
Send us ten URLs. We score each against the Gain rubric and hand back what to keep, what to fix, and the three assets we would build first.