A topic queue that can show its working.
Every content tool produces a ranked list. Almost none of them can tell you what the ranking is made of. Docket's score is built from five measurable signals, and it prints how many of them it actually had.
A confident number nobody can check is worse than no number.
Ask a language model to score eighteen topics and it will hand back a cluster in the seventies, because it has no way of knowing which of them your business can win. That number then becomes the sort order of the screen, and the week's work is decided by it.
The failure is quiet. Nothing looks broken. The queue is full, the scores look considered, and the team writes what is at the top. Six months later the pieces that worked and the pieces that did not are indistinguishable in the plan that produced them, because the plan was never anchored to anything measurable.
The question is not whether a tool can rank topics. It is whether the ranking is made of anything.
Five measurements, one capped opinion, and a penalty that can sink a topic.
- 1
It reads your site
Industry, services, audience, tone and a named competitor set, inferred from the site itself and editable everywhere. This is what business fit is measured against.
- 2
It gathers evidence
Your Search Console demand and positions, and the answers assistants gave to your tracked questions. Three queries for a whole batch, so grading eighteen topics costs what grading one does.
- 3
It validates before it proposes
Each candidate is checked for real demand, for overlap with a page you already rank with, and for whether assistants already recommend you. Duplicates are rejected by measurement rather than by asking the model nicely.
- 4
It scores, and says what on
Weighted terms, renormalised over what was actually measured, capped by how much that was. Every card carries its basis line and the full breakdown is one click away.
Specific enough to argue with.
Every line here describes something the software actually does. If one of them turns out not to be true of your setup, that is a bug and we want to hear about it.
The planner's opinion can never exceed 15 of 100.
It is held out of the renormalisation deliberately. Renormalising it would make a model's rating half the score in a workspace with nothing connected, which is the exact failure the score exists to end.
A topic competing with a page you already rank for is caught.
It becomes a recommendation to refresh that page, with the URL, its position and its impressions attached. Publishing a second page onto ground you already hold usually loses to the page that was already there.
Ranking weighs the score against what the piece will cost you.
A 70 that Docket can finish today ranks above an 80 that needs an interview nobody has booked. Effort discounts rather than disqualifies, so a heavy piece with real demand behind it still surfaces.
The queue is balanced by search intent, not by whatever was easiest.
Three topics per intent per week across informational, commercial, transactional, navigational, local and seasonal. A queue that is all explainers is a blog nobody buys from, and that is what an unconstrained model produces.
Seasonal topics are dated from an almanac, never by a model.
Over a hundred occasions with computed dates, filtered by your market and your industry, each carrying the lead time the piece needs to be live in time. A model asked for a date states a wrong one with complete confidence.
What it does not do
- It does not buy third-party search volume. The demand it scores is the demand your own site can already see, which is free and more truthful about your situation than a vendor's estimate.
- It cannot score demand at all until Search Console is connected, and it says so on every card rather than filling the gap with a guess.
- It does not write the piece for you at this stage. It decides what is worth writing; the drafting is a separate step you trigger.
- It has no model of your commercial priorities beyond what your profile says. A topic that matters because of a deal you are chasing is one you should add by hand, and you can.
Answers, Not Hedging.
The parts either side of this one.
Twenty minutes, your website on the screen.
Not a slide deck. We run the checks against your actual domain before the call and spend the time on what came back. If it turns out you do not need this, we will say so.