GEO-125·Issue 11·Content & Semantics·125 / 132
2027 Content Planning: From Products to Category Knowledge
Stop spreading the 2027 content plan across SKUs. A four-node category knowledge map, a five-step build, and a clean ownership split with product pages.
Issue focusScale It: SOPs, Organization, and the 2027 Roadmap
At an October planning meeting, a DTC pet-supply team reviews the year: 180 SKUs, each with a page, forty rewritten around answer units. Someone asks an assistant: “I have three cats and travel for work — do I need two fountains?” The answer cites a US review site’s guide and a forum thread. None of the 180 pages appears.
The 2027 content moat is not on product pages but in category knowledge: a product page answers “is this one worth it,” category knowledge answers “how do I choose among these,” and the second is what AI reaches for before the shortlist forms.
Why a product page cannot answer a category question
A product page proves itself: these specs, this improvement over the last generation, this price. One position to defend: buy mine. A category question wants trade-offs: what to give up on a tight budget, whether a small apartment needs the large unit, when the category is wrong for you. A brand selling only its own product cannot easily write “don’t buy this.”
The usual counter-example is a product grid plus a brand blurb, or forty product pages relabeled with the category name. Page count doubles; AI still gets forty self-justifications, no comparison frame.
A category answer needs three inputs: decision dimensions (what these products are compared on), exclusion conditions (when not to choose one), and checkable measures (a number and the test behind it). None is a product-page KPI, so more product pages do not close the gap.
The category knowledge map: four kinds of node
Category knowledge is not one long guide but a set of nodes, reusable across pages and languages. Draw the map before choosing topics.
- Buying dimensions. The five to seven fields that separate products. For each, write when a high value helps and when a low one does — not “bigger is better.”
- Use scenarios. Who uses it, under what conditions, in the buyer’s words — not marketing segments.
- Budget tiers. What each price band buys and what it forces you to give up. The cell planners skip and models quote most.
- Failure modes. Return reasons, installation traps, compatibility conflicts. “When not to buy” enters an answer faster than “why buy me.”
Ownership follows the data: content owns dimensions and scenarios, merchandising owns budget tiers, support owns failure modes.
Five steps from map to reusable asset
| Step | Action | Owner | Passes when |
|---|---|---|---|
| 1 Draw the boundary | SKUs buyers compare together, plus one decision question | Content lead | One sentence names who compares what |
| 2 Fill the map | Mark every cell: have, missing, contradictory | Content, support, merchandising | Each cell points to a page or says “missing” |
| 3 Rank topics | Frequency × answer gap × unique evidence, not search volume | Growth, content | Each top-ten topic points to “no page answers this” |
| 4 Make it an asset | One module = one table + one answer unit + one checkable proof | Content | Hide the logo; it is still the clearest answer in the category |
| 5 Split ownership | Knowledge to content; specs, price, stock to merchandising; one master data set | Content, merchandising | Any two pages show one value per field |
Step 3 is where planning drifts. High-volume questions already have answers; the model does not need yours. Medium-frequency questions every competitor answers vaguely are the citation opening.
Step 4 separates a campaign from an asset. An article is rewritten in six months; a table plus a definition feeds a product page, a category page, a comparison page, and a second language — the job of the modular content system.
Step 5 needs enforcement, not a meeting. If the category page says twelve hours of battery backup and the product page says ten, the build should fail before a customer notices. That is a CI check, not an editorial guideline.
Example: one category, three assets
Example — brand and figures are invented; the structure is the point.
A fountain brand collapses forty SKUs into one decision question: how many pets, and how many days is the house empty? It builds three assets.
First, a decision table: capacity in ounces, noise at one meter, filter cost per year, parts to disassemble weekly, battery backup in hours. Each cell says “choose high when…, choose low when…”, never “bigger is better.”
Second, three answer units: a three-cat household, a single cat with a nine-to-five owner, a frequent traveler — each 80 to 120 words with a recommendation, an exclusion, and one number a stranger could check.
Third, forty product pages keep specs, price, and checkout, with one line above the fold — “this model suits X, not Y” — and a link to the category page. They finish the comparison, not the category education.
Acceptance is not a ranking. Re-run the same twenty questions: does the brand appear in the comparison paragraph, and do those numbers match the page? Treat it as a format check, on your quarterly review cadence.
Where this breaks
First, the category has no stable dimensions. In a new form factor, buyers do not yet know how to compare; a forced table is wrong by spring. Start with scenarios and product content.
Second, sales happen mostly inside a marketplace. Platform AI reads listings, reviews, and fulfillment metrics; on-site category knowledge drops down the queue.
Third, master data is not unified. One citation of a contradictory number costs more trust than ten pages earn.
Fourth, category knowledge is treated as third-party evidence. Your own buying guide is still self-justification; reviews, community threads, and certifications are not optional.
Fifth, five categories launch at once. Every map needs support and returns data. Take the one or two carrying the most revenue, finish a review cycle, then copy the pattern.
This month: one category, twenty questions
Pick the category with the largest revenue share. Pull twenty real questions from the last 90 days of support chats and site search, and mark each against existing pages: answered, half-answered, or nobody. The “nobody” questions become the first 2027 topics, each paired with a table or an answer unit. That is the map’s first version.
Set the year’s cadence against the 90-day roadmap, and keep templates on the content and semantics hub. Planning by SKU produced 180 pages answering nobody’s question; planning by category produces fewer pages that answer the question buyers ask before they know your name.