GEO-040·Issue 04·Strategy & Market·40 / 132
Planning Content Capacity: What Comes After 132 Articles?
Stop counting articles. Plan GEO content in answer units, run three supply lines with different decay rates, and use four pools instead of a keyword sheet.
Issue focusAnswer Units: Writing Category and Product Pages for GEO
Late March, a quarterly review at a brand selling portable power stations into the US. The content lead is asked what to publish next quarter. Her keyword sheet has hundreds of rows; after deduplication, barely a third point at different questions. The paragraph an assistant cited most last quarter came from an engineer’s ticket reply, typed in four minutes, never on a plan.
GEO content capacity is not publishing speed. It is three supply lines running at once: templates set the floor, experts earn the citation, and first-party data cannot be copied.
Capacity is answer units, not articles
A 3,000-word category guide can hold zero extractable answer units; a 600-word compatibility note can answer three questions. Plan in answer units per week and the number drops.
A keyword proves demand, not where the answer comes from. The repeatable material is the condition only you know: whether this model starts an air-conditioner compressor, how charging falls below freezing, what differs between an EU and a UK plug. Ask how many first-hand judgments the team can produce this week. The honest answer is smaller than the plan.
The answer unit framework fixes the shape. Capacity planning asks who supplies the sentence inside.
Three supply lines, three decay rates
| Line | Raw material | Cost, copying, decay | Owner |
|---|---|---|---|
| Template | Answer-unit skeleton, comparison frame, troubleshooting template | Low and reusable; a competitor can write the same page. By the fifth sibling comparison, new ones stop being cited | Content editor |
| Expert | Engineering, support, buying, and repair experience | High; interview, edit, byline, review. Slow decay; cited for years | Content lead plus a business expert |
| Data | Tickets, compatibility lists, measured dimensions, regional voltage and compliance | Medium; access, cleaning, quarterly refresh. Yours alone | Operations or data |
A template-heavy start is fine: templates arrange what you know and make the fourth page cheap. The test is month three. If expert and data still produce nothing, doubling output doubles page count while the cited set stays the same.
The expert line costs most because it needs a named person, a date, and test conditions. That is why it survives. An assistant can quote a conditional, verifiable claim; it has no reason to quote a generic one.
Four pools instead of a keyword sheet
- Customer questions. Verbatim pre-sale and post-sale tickets, exported weekly, deduplicated, tagged. Closest to how buyers ask.
- Site-search dead ends. Queries with no result or near-zero clicks. Closest to traffic you already have.
- Expert agenda. Monthly: what engineering and merchandising are asked most, and what gets misunderstood.
- Off-site corpus. Objections and comparison requests recurring in communities, review sites, and marketplace Q&A.
Four checks before production: one specific question rather than a noun; a named source, whether expert, data, or public test; no existing page answering the same question, or merge and re-link; a clear decision stage. Fail two checks and drop it.
The counterexample: split “best portable power station” into twenty long-tail phrases and publish twenty guides on one outline. The site looks complete. To an assistant it is one answer repeated twenty times.
Set the ceiling before the quota
Start by admitting a limit. Plan at 80 percent of the last four weeks’ actual output, not the number someone wants to be true.
| Step | Action | Owner | Passes when |
|---|---|---|---|
| 1 Ceiling | Answer units per week, at 80 percent of the last four weeks | Content lead | Delivery within 20 percent of plan |
| 2 Pools | Assign an owner and cadence; when a pool is empty, stop publishing | Content, support, engineering | Every topic traces to a pool and source |
| 3 Quotas | Quarterly quota per line; expert plus data at least 30 percent | Content lead | Cited pages attributed by line |
| 4 Refresh | Citation pages quarterly, conversion pages monthly, long-tail pages every six months or merged | Editor and engineer | A record of merged and retired URLs |
| 5 Review | New answer units, interviews, data refreshes, not article count | Growth lead | Two of the three rising |
Step 2 is the one teams skip. An empty pool means booking an engineer, requesting data access, waiting for an export: harder than writing. That discomfort is the signal. The line is broken; more scheduling multiplies near-duplicates.
Step 4 needs judgment. Prices, stock, and policies change with facts; an opinion essay changes only when its conclusion is overturned. The content refresh policy sets the triggers; the page priority matrix orders the work.
Example: one quarter at a three-person desk
Example: a hypothetical outdoor energy-storage brand, three content people, about twenty pages a quarter. No real brand, no performance figures.
- Templates, about half. Power-tier selection pages (“how many watts do I need”), one answer-unit structure, two weeks. The job is coverage, not differentiation.
- Experts, about a quarter. Two 30-minute interviews a month with the power engineer, questions from the highest-friction tickets: parallel expansion, cold-weather charging, air-conditioner start. Bylined, with test conditions; who is accountable makes the claim checkable.
- Data, about a quarter. Return reasons exported into a compatibility and troubleshooting table, dated, refreshed with each new model.
Review by line, not traffic. Citations almost all on templates means differentiation is thin. An expert line at zero usually means the byline, date, or conditions are missing; fix those before raising output.
Where this breaks
A small team in one market with few SKUs does not need three lines. One interviewable expert and one template will carry the year.
Slow-moving industrial parts have a different bottleneck: re-verification, not supply. Schedule how often a page is checked, not how much is published.
Arbitrage and dropship sellers have nobody to interview; a manufactured expert persona gets tested by the model and the buyer alike. Run the data line first and let templates lead.
A model filling the quota is the last trap. It speeds drafts and localization, not judgment; near-duplicates do not widen the cited set, only maintenance. In medical, children’s, or battery-shipping categories, the expert line is a precondition to publishing, not an option.
This month, list every page an AI assistant cited in the past twelve months, tag its source as template, expert, or data, and find the line at zero. Give it a fixed quota next quarter and put the interviews on the calendar. The budget for that argument belongs to the strategy and market lane.