A tiny site should build a content inventory before it obsesses over analytics.
Analytics can tell you what happened.
An inventory tells you what exists.
That distinction matters when the site is still young, the corpus is changing every day, and the publishing system is moving faster than public memory can keep up.
If the public site has thirty posts, four topic lanes, author archives, schema feeds, RSS, image credits, and a search index, then the first operational question is not “which chart went up?”
The first question is simpler:
Can we prove what is live?
Why inventory comes first
Small sites often jump straight to dashboards.
Pageviews. Search clicks. Referrers. Average position. Bounce rate. Time on page.
Those numbers eventually matter, but early analytics can be noisy and strangely seductive. A single shared link can distort the week. A crawler delay can make a strong page look invisible. A quiet but useful article can look like a failure before it has had time to earn search trust.
A content inventory is more boring.
Good. Boring is easier to trust.
The inventory answers:
- Which posts are published?
- Which posts are drafts?
- Which topic lane does each post support?
- Which author owns the piece?
- Which public image and license are attached?
- Which related posts should the reader see next?
- Which schema node represents the page?
- Which search-index entry points to it?
- Which live URL returns 200?
That is not vanity measurement. That is site truth.
What current AI news keeps teaching
The current AI news cycle keeps rewarding the same pattern: the harness matters as much as the model.
AINews is tracking coding-agent progress through harness engineering, verification loops, memory, citation grounding, and agent infrastructure. Future Tools keeps surfacing deployment stories where the interesting question is not the announcement, but whether the tool survives real use.
That lesson applies to a small website too.
The article is the model output. The site system is the harness.
If the harness cannot prove what it published, what it linked, what it credited, and what it exposed to crawlers, the site is operating on vibes.
Vibes are not a publishing strategy.
The inventory a tiny site actually needs
Start with the corpus.
Every published post should have a title, date, excerpt, author, tags, image source, image credit, image alt text, and draft state. If any of those fields are missing, the site is not ready to trust analytics because the basic object being measured is not clean.
Then inventory lanes.
For this site, the working lanes are Practical AI, BIM and Revit, Creator Tools, and Editorial Ops. A good inventory shows whether those lanes are balanced, which lane is stale, and where the next post should land.
Then inventory authors.
Bylines should not be decoration. Each writer needs a public archive, a profile, a beat, and a body of work. When readers click Anton, Cara, Zack, or Ahmed, they should land on a coherent trail instead of a name badge.
Then inventory discovery surfaces.
The sitemap, RSS feed, llms.txt, search index, topic hub, author schema, post schema, and Start Here page all need to agree with the corpus. If one surface drifts, readers and crawlers receive a weaker map.
Then inventory proof.
For each new article, the publishing run should verify:
- the built HTML exists
- the live URL returns 200
- the page shows the right title, author, image, and credit
- the search index includes the title
- the schema endpoint includes the article
- related links render
- smoke tests and SEO growth checks pass
That is the checklist before analytics gets a vote.
Where analytics can mislead
Analytics can make a small team reactive.
If one post gets a spike, the temptation is to copy the surface topic. But maybe the spike came from a one-time share, not search intent. Maybe the post worked because it had a clear checklist, not because of the keyword. Maybe the better lesson was structure, not subject.
Analytics can also hide maintenance debt.
A dashboard might show traffic climbing while image credits are missing, author pages are stale, schema points to the wrong person, or the topic hub has not caught up with the newest posts.
That is why inventory comes before interpretation.
Do not ask analytics what worked until the site can prove what shipped.
What Anton should watch each day
Anton should treat the inventory like a morning instrument panel.
Check the published count. Check draft count. Check the backlog. Check which lane is thin. Check whether the latest post appears in the search index and schema. Check whether author pages reflect the current body of work. Check whether public claims still match repo reality.
Then pick work.
If the Practical AI lane is strong but Editorial Ops is thin, publish an operations piece. If Creator Tools has fresh posts but no first-read route, improve the topic hub. If the site claims daily publishing but the board is stale, update the board. If images are bland, source public images with credits.
The inventory should decide the next move.
A small-site rule
Do not let analytics become a mood ring.
Let analytics become evidence after the inventory is clean.
For a tiny technical site, ranking grows from a set of repeatable habits: publish useful pages, cluster them, link them, credit them, expose them through schema, verify the live result, and keep public surfaces honest.
That is not glamorous.
It is how small sites stop looking accidental.
Verdict
A tiny site needs a content inventory before analytics because analytics cannot fix confusion in the corpus.
First prove what exists.
Then prove it is discoverable.
Then prove readers and crawlers can move through it.
Only then should the dashboard start making decisions.
— Anton