The Right Unit of Content Strategy Is a Customer Decision
Keyword research reveals language and demand, but it does not define site architecture. Customer decisions are a better unit for choosing pages, sections, tools, or nothing.
Marketing
8 min
The short version: Plan content around a distinct customer decision or task, then choose the smallest maintainable asset that completes it. A keyword variation is evidence of language. It is not automatic permission for another URL.
Keyword research still matters. It shows how people describe a problem, how intent changes, and where demand may exist. The mistake is turning every phrase into site architecture. That leap produces multiple pages that repeat the same advice, compete for the same internal links, and age independently.
The safer planning question is: what is the reader trying to choose, diagnose, calculate, implement, or verify right now?
The current inventory shows why the unit matters
On August 10, 2026, I reviewed 181 records in this site's Thoughts CMS and the public corpus around them. Several families contain closely related URLs: three treatments of the agency exception queue, two LLMs.txt-as-moat arguments, and three founder-outbound trigger-library pieces. Some may serve different depths or stages. The inventory alone does not prove cannibalization. It does prove that another keyword-shaped variation needs a stronger page-versus-section case.
The five newest Thoughts exposed the opposite failure. Their titles described distinct frameworks, but the pages had almost no body. Architecture and substance failed in different ways. A decision map helps with both: it asks whether the URL is distinct and what complete job the asset must perform.
Write the decision brief before the content brief
A useful decision brief has five lines:
Actor. Name the person close enough to the work: a demand-generation lead, founder, product marketer, deliverability operator, or engineering owner.
Moment. State what triggered the need: a placement drop, a stalled deal, a tooling renewal, a failed migration, or a quarterly planning decision.
Options. List the real choices, including wait, repair, consolidate, build, buy, and stop.
Evidence needed. Define the facts, first-hand input, product behavior, or primary-source synthesis that would change the choice.
Qualified next action. Specify what a helped reader can sensibly do next.
“Email deliverability” is a topic. “Should I pause sending after placement drops while authentication still passes?” is a decision. The latter tells you which diagnosis, evidence, caveats, and next step are required.
Run the page-versus-section test
Before approving a URL, answer four questions:
Does the task have a distinct reader, trigger, set of options, or success condition?
Can the asset contribute unique evidence, a tool, a workflow, or accountable expert judgment?
Would the answer be useful as an independent search or navigation destination?
Can someone maintain the material claims and route when reality changes?
If the case is weak, update an existing page or add a section. If two proposed pages support the same decision with the same evidence, consolidate them before drafting. If the current page has authority but the format does not solve the task, reformat it rather than minting a competing URL.
This is where the evidence contract and publication ceiling work together: one defines what the page may claim; the other gives the system permission not to create it.
Choose the asset that removes uncertainty fastest
Product or venture page: the reader must decide whether an offer fits and can take a product action.
Benchmark: a trustworthy dataset changes the decision and the method can be inspected.
Comparison: the options and tradeoffs are stable enough to evaluate explicitly.
Calculator or diagnostic: the answer depends on the reader's own inputs.
Template or documentation: the reader needs to complete or hand off a workflow.
Case study: a permitted example has inputs, intervention, outcomes, limits, and accountable contributors.
Thought: first-hand judgment or primary-source synthesis changes an operator's choice.
Feed item: a timely signal needs a concise operating implication, not an evergreen URL.
Section or no asset: the existing journey already owns the task or the evidence is not ready.
Content teams default to articles because articles fit production. Readers care about resolving uncertainty. The format should follow the decision; length should follow the information required.
Build reader journeys, not mechanical wheels
Connect each asset to the reader's next likely decision. A strategic article may lead to a template; a diagnostic result may lead to the right repair and then a verification step. The bottom-of-funnel content map is useful when it reflects a real buying task, not when every page links to every other page.
Use descriptive anchors where they answer a real next question. Do not force the same anchor into the opening paragraph or repeat it mechanically. A small number of contextual links can create a better journey than a large synthetic cluster.
Measure progression by decision stage
Define the useful action for each task: diagnostic completed, template copied, comparison viewed, product workflow started, qualified consultation requested, or later-stage page reached. Preserve the baseline and annotation, then report the cohort median and per-URL outcomes at 28, 60, and 90 days. Keep branded and non-branded discovery separate when Search Console data becomes available.
A low-traffic page can be valuable if it consistently helps the right reader progress. A high-volume definition page may still be useful without direct conversion, but that role must be explicit. At day 90, decide whether to expand, improve, reformat, merge, maintain, or stop.
The next planning action
Take one proposed editorial cluster into the next roadmap meeting and replace its keyword list with decision briefs. Distribute the briefs to sales, support, product, and the accountable expert for correction before assigning an asset. Publish only the gaps with a distinct task and qualified next action. For search-facing decisions, connect approved assets to the AI Search, GEO, and AEO hub. Measure the resulting URLs as a decision cohort—not as another contribution to whole-property traffic.
Sources and limits
The proprietary input is the August 10 site and CMS inventory. It identifies overlap risk but does not prove query cannibalization, customer demand, or ranking impact. No genuine Search Console export or fresh permission-cleared sales, support, or customer-language corpus for this domain was available in this run, so I am not presenting the framework as validated customer research. Google's crawlable-link guidance and people-first content guidance support usable journeys and reader value; the customer-decision model is an operator recommendation.
