A two-hundred-row topic sheet, running from "what is an invoice" all the way to "how to connect an invoicing API". The team can publish two posts a week, which puts the last row in week one hundred.
The sheet is not sloppy. Every row has a primary term, a search volume, and an intent label. What it assumes is that capacity is unlimited and that no post relates to any other.
A topic list answers what to write. A topic map also has to answer which page goes first, and why the next one depends on it.
A two-hundred-row sheet is not a map yet
Two hundred rows divided by two posts a week is one hundred weeks, close to two years. Lift output to five a week and the same sheet becomes forty weeks, nine months. The library did not change. The project changed shape twice.
The sheet fails as a map because it has no basis for ordering. Sort by volume and you start with the biggest term, which is also the hardest, and three months pass with nothing moving. Sort by intent and the site opens with decision pages, asking a first-time reader to buy. Sort by competitors and you get a comparison list where every post answers a question someone else already answered.
Our own blog went through a re-cut. From December last year to July, we published two or three posts a month and accumulated twenty. In October we moved to one a day, thirty in the month. The same sheet was an eighteen-month project at three a month and a ten-week project at one a day. Output changed, so the slicing had to change. Without that, the back half of the sheet never comes up while the front half gets rewritten again and again.
Capacity is the boundary condition of a map.
A sheet that takes more than two quarters to walk is what we would call a wish list. A map should not stretch past two quarters, because product, price point, and channels all move, and the reasoning written on the sheet expires.
Start the map from pages you already have
At the end of October, a site selling link building and content services sorted its own 353 posts into seven groups and joined each group to three months of Search Console data. The grouping was crude, based on words like link or backlink in the URL, and the author noted that some posts could belong to two groups.
The result looked like this. City and town listicles, 177 posts, half the archive, took 96,536 impressions, eighty percent of the site's 121,374, and turned them into 27 clicks at an average position of 40.9. Link building and backlinks, the thing they actually sell, was 45 posts, 8,431 impressions, 2 clicks, average position 71.7. Technical SEO had 17 posts and AI search 16. The site carried 10,704 internal links.
Plenty of links, pointing the wrong way. A reader landing on the homepage would conclude this was a directory of local agencies, not a link-building company.
For an established site, the first version of a map is this inventory picture. Three counts per group: how many pages, how many clicks in the last three months, and whether the group has an entrance page that covers the whole subject. That third count is the one people skip. Supporting pages with no entrance page are posts hanging nowhere, with no line between them.
Impressions mislead as well. 177 posts took eighty percent of impressions and produced 27 clicks at an average position near forty. Past position forty, extra impressions are drive-by traffic. Rank the map by impressions and capacity keeps flowing to the group that earns least, and every year someone cites the biggest traffic number as proof it should continue.
Eight columns, one question each
The template only helps after the inventory. Ours has eight columns, one question per column. A cell you cannot fill is a topic you have not finished thinking through.
| Column | What goes in it | What an empty cell tells you |
|---|---|---|
| Page | The URL or path this page will finally live at | If the URL is undecided, the page format is undecided too: standalone page or a section of another |
| Intent | Awareness, comparison, decision, or use. Pick one | If you cannot pick, the page is still answering two questions |
| Primary term | One term, carried over from the keyword sheet, unique inside the group | Two terms competing for one page usually means split it or merge it |
| Supporting evidence | What this page can actually show: a test you ran, a screenshot, an official passage, real numbers | Empty means it does not belong in this batch |
| Upstream and downstream | One page this one receives readers from, one it sends them to | If neither exists, it has no relationship with the group |
| Link direction | Which pages link to it, which pages it links to, by page name | If it only links out and nothing links in, it is an orphan |
| Status | Draft, awaiting link-up, live, needs revision | Stuck in draft for more than a quarter, delete the row |
| Priority | P1 this quarter, P2 next, P3 watch | Everything at P1 means nothing is prioritized |
Supporting evidence is the column people skip. It holds what the page can put on the table: a record of something we measured, a screenshot from a console, a line from official documentation, real numbers from a project. Google's content guidance opens its self-assessment with a question about whether the page offers original information, reporting, research, or analysis, and follows immediately with whether it gives a substantial, complete, or comprehensive treatment of the topic. Translated into spreadsheet language, those two questions are this column.
A row with no evidence writes the same article under a different headline. The more of those a site accumulates, the harder it is to explain why anyone should treat it as a source on the subject.
Upstream and downstream is where the user journey enters the sheet. Someone who finishes a page of invoice templates asks next whether to buy software. Someone who reads a free invoicing tool comparison wants to know where the free tier starts charging. Those two sentences are the upstream and downstream, written as page names rather than a string of related posts. If they cannot be written, the page has no relationship with anything in the group, and it belongs elsewhere or nowhere.
Primary terms come from the keyword sheet produced in Keyword Research and Search Intent: From Demand to a Working Keyword Set. That article covers how to mine terms and sort intent. This one only cares that each cell holds one term and no term repeats inside a group.
Status records draft, awaiting link-up, live, or needs revision. When to revise an existing page and when to publish a new one has its own criteria in Content Strategy in the AI Era.
Priority is not sorted by volume in our version. P1 goes to two kinds of page: the group's entrance page, and any page that already earns impressions but almost no clicks. The entrance page decides whether the group is visible at all. For the page with impressions and no clicks, the work is a title and description change, which is far cheaper than writing something new. A group's P1 pages finish before the next group starts.
The journey sets the order, capacity sets the page count
The user journey usually gets drawn as four stages, awareness, comparison, decision, and use, followed by two posts per stage for a total of eight. It looks tidy. What it produces is four posts that are each too thin to be useful.
What the journey actually sets on a map is sequence and dependency. It does not set page counts. In a long B2B decision chain, awareness and comparison need more pages, while the decision stage needs very few: a comparison page and a path to a quote will do, because the last step usually happens with a salesperson. Fast consumer decisions invert this. Price, specifications, and availability pages carry the weight, and awareness pages can be lighter.
Whether two terms deserve two pages comes down to the shape of the answer. If one passage can answer both, merge them. If one term needs a comparison table and the other needs a procedure, the shapes differ, so split. The gain from splitting has to cover the cost of maintaining another page, and if it does not, merge. This runs against the popular "one keyword, one page" rule. The boundary is competitive intensity: when the top ten results are all deep topical treatments, the merged page will not hold.
Do not lay the whole thing out on the first pass. Three to five pages that can reach each other is a group you can start on. Four tests decide whether a group can start: any page reaches any other in two clicks or fewer, one entrance page covers the whole subject, every page has at least two outbound links and one inbound link inside the group, and there are no orphans.
Fewer than three pages and there is nothing to compare against, so the supporting relationship collapses. More than five and the first pass will not finish inside a quarter, which leaves a set of pages that do not connect, worse than not having started.
Expand after the group ships. Publish the entrance page plus two supporting pages, watch whether clicks inside the group move, add the fourth and fifth if they do, and fix the entrance page if they do not. Adding a page is the most expensive optimization available, because it means writing one. Fixing the entrance page is cheap. How to place links and write anchor text inside a page is covered in On-Page SEO: Titles, Page Structure, and Internal Links.
Conversion pages first, or authority first
This is the argument that breaks out in every planning meeting. Sales wants landing pages. Content wants long-form depth. Both have a case.
We check three things.
Whether any page already sits between positions eleven and thirty for the target term. If so, patch that group's conversion pages first. That range means search engines already associate the site with part of the subject, and what is missing is the page that catches the intent. One page is the cheapest possible fix.
Whether the top ten results are all deep treatments from large sites. If so, build the awareness side first. A few hundred words of product copy will not break in, so a set of pages that covers the subject has to exist before a conversion page has somewhere to hang. There is a second variable on that road: how to earn links and brand signals belongs to Backlinks, Digital PR, and Brand Authority.
Cash flow and the review cycle. Quarterly reviews and tight cash push toward conversion pages. Annual budgets and sales cycles beyond six months push toward awareness, because conversion pages only produce numbers after procurement finishes, which is not inside the quarter.
Our own order is to patch holes before building towers. A hole is a page with impressions and no clicks, traffic and no conversion, or a group missing its entrance page. A tower is a new subject started from zero. Holes first, because towers need the ground.
The boundary matters too. If one person writes everything, do not do both. Finish one three-page group before opening the next. A half-built group stays on the site consuming links and crawl attention, and when something goes wrong you cannot tell which part caused it.
Slice the map to weekly output
The slicing method is plain. Line up the P1 rows by group, treat each group as one delivery package of an entrance page plus supporting pages, and cut the plan at whatever the team can ship in a week. Every slice should contain at least one complete group.
An illustration, purely arithmetic. Three posts a week for six weeks is eighteen. At four pages per group, eighteen covers four and a half groups. A quarter of capacity is four complete groups, not eighteen loose pages. Plenty of maps shrink by half at this point, which is a good outcome: the map has started to match capacity.
One more rule for slicing: do not take one post from each of three groups every week. Three groups open at once means three half-finished groups three months later, each missing an entrance page. Finish one group, then open the next, and the site accumulates whole subjects instead of parts.
Set the horizon at two quarters. We redraw at the end of each one, deleting or downgrading rows that will not fit. What a map produces is a written statement of what this quarter will not do.
Deleting rows follows three categories: a primary term that duplicates another page in the group, a supporting-evidence cell that stays empty, and a status stuck in draft for over a quarter. The first two go without discussion. The third deserves one question: was there no time, or nothing to say.
Measure by group, not by page
Once the map is set, someone has to read the data through the same grouping. Page by page, a term slipping from eighth to twelfth may only reflect a different device or location.
Search Console's Performance report can filter on the Pages dimension with a regular expression, pulling a whole group of URLs at once. Google's help page notes the syntax is RE2, that matching is partial by default, and that anchoring the start takes a caret. Pull clicks, impressions, and average position for each group, compare against the previous quarter, and keep one row per group beside the template.
Read click share before position. When a group's impressions rise and clicks do not, the problem sits in titles and descriptions, covered in Title, Meta Description, and H1. When impressions never moved at all, the group is missing pages, or its entrance page does not cover enough of the subject.
For the first two months after a group ships, we look at two things only: whether pages were crawled, and whether they entered the group. Position can wait. A group typically takes a few weeks to two months before its numbers settle, and watching weekly will only make you anxious.
Definitions for impressions, clicks, and CTR live in Rankings Rose but Revenue Didn't, which builds the ROI case on GSC and GA4. Here they only serve comparisons between groups.
A map earns its keep by what it cuts
Judge a map by how many rows it removed this year, not by how many it lists. Listing takes an afternoon. Cutting well requires knowing what each group absorbed and what came back.
What a map can guarantee is limited. It keeps the site from repeating itself or leaving gaps, gives every page a way in and a way out, and gives scheduling something to stand on. Ranking depends on other factors, and we will not claim that writing to a map produces one.
The problem that will stop you next is probably not a shortage of ideas. It is three half-finished groups, capacity for one, and no record on the map of which group deserves it. At that point the missing piece is not more topics. It is a record of what each group cost and what it returned.
Related reading
- Content Strategy in the AI Era: Topical Authority, Quality, and Updates, the wider frame for topical authority, quality judgments, and content lifecycle.
- Keyword Research and Search Intent: From Demand to a Working Keyword Set, where the primary-term column comes from.
- On-Page SEO: Titles, Page Structure, and Internal Links, structure and linking inside a single page.
- Rankings Rose but Revenue Didn't: Building an SEO ROI Case with GSC and GA4, definitions for impressions, clicks, and CTR.