Measure and Iterate: Reading Your First Month of Data
By Timmy · AI-first SEO practitioner · 11 years in SEO ·
In the first month, your data is too small to prove anything about conversion. It is big enough to answer four questions: which pages are indexed, which queries are appearing, which pages get shown but not clicked, and where checkout stalls. Answer those, change one thing at a time, and wait a full window before judging.
Take DressFisher four weeks in. Say 12 of its pages are indexed, 38 queries have impressions, two collections get clicks, and the size guide is the most-viewed page. That is enough to decide the next four weeks. It is not enough to call the store a failure.
The short version
- Use a 28-day window. Daily numbers on a new site are noise.
- Read direction and counts, not percentages.
- One change at a time, then one full window before the next.
- Indexed pages and queries with impressions come before sessions.
- Fix, merge, or delete before you add another page.
- Do not change URLs or redesign because week two dipped.
1. Four sources, four different truths
| Source | What it is good for | What it gets wrong |
|---|---|---|
| Search Console | Queries, impressions, clicks, index status | A 2-3 day reporting lag; no data for very low volume |
| GA4 | Sessions, source, on-site events | Breaks easily with tracking issues; useless at tiny samples for rates |
| Store backend | Add-to-cart, checkout, returns, revenue | Only sees what happened after arrival, not why |
| Messages and returns | The words customers use | Not quantified; easy to dismiss as anecdote |
When they disagree, trust the store backend for money and Search Console for search. GA4 is for direction, not for judging a page on 100 sessions.
2. Four questions, every week
| Question | Where to look | If the answer is yes |
|---|---|---|
| Are new pages indexed? | Search Console → Pages | Fix crawl or internal links on the ones stuck |
| Are new queries appearing? | Performance → Queries | Add the query to an existing page, not a new one |
| Do pages get impressions but no clicks? | Performance → Pages | Rewrite the title and description to match the query |
| Does checkout stall? | Store backend → abandoned checkouts | Check shipping cost, delivery time, and return policy |
Twenty minutes a week is enough. The monthly decision is what to do with what you saw.
3. What the pattern tells you
| Pattern | Likely cause | First fix |
|---|---|---|
| Impressions up, clicks flat | Title or description does not match intent | Rewrite the snippet |
| Clicks up, add-to-cart flat | Page does not answer fit, price, or delivery | Add measurements, photos, shipping and returns detail |
| Add-to-cart up, checkout flat | Cost or friction appears late | Show shipping and returns before checkout |
| Everything flat | Not enough pages indexed, or not enough distribution | Publish or promote; do not redesign |
| One page falling, others stable | That page lost a query | Check the query and the competing page; do not touch the rest |
4. What I would not do in month one
- Change a live URL. Redirects, lost signals, and no upside at this volume.
- Run an A/B test. You do not have the traffic to reach significance. Pick, ship, and move on.
- Redesign the homepage. A new site’s problem is rarely the hero section.
- Add ten more pages. Fix the pages that already have impressions first.
- Check rankings daily. Check impressions weekly. Rankings bounce for reasons that have nothing to do with your work.
The AI workflow I use
The useful AI job here is joining three exports that do not talk to each other: Search Console, GA4 landing pages, and return reasons. AI can align queries to pages and propose what to merge or rewrite. It cannot see live data, so export first.
I run a small Shopify store. Here are three exports from the last 28 days.
1. Search Console, Queries: query, clicks, impressions, position
2. Search Console, Pages: page, clicks, impressions, position
3. Returns and support notes: page or product, reason text
Constraints:
- 12 indexed pages, about 100 sessions in the period.
- I can change two things this month.
- I will not add more than one new page.
Do this:
1. Map each query to the page that should own it. Flag queries that
belong to a page already, and queries with no page.
2. List pages with impressions but a low click share, and draft a new
title and description for the top two.
3. Group the return reasons. Say which page or product detail would
prevent each group.
4. Recommend two changes, ranked by expected effect, and say what to
check after 28 days.
Use only the numbers in the exports. Do not estimate search volume.
Do not recommend a redesign.
Check three things before acting: whether the “no page” queries are just variants of a page you have, whether the return reason repeats or appears once, and whether the change is something you can actually ship this week. AI will also happily suggest a new page for every query, which is how sites end up with fifty thin pages.
What most guides will not tell you
- Impressions move before rankings do. A page can gain impressions for weeks while its average position looks unchanged. Watch impressions to see if a page is being discovered at all.
- On a new site, the most important numbers are not sessions. They are how many pages are indexed and how many distinct queries produce impressions. Those two predict next month’s traffic better than today’s session count.
- Iteration is mostly subtraction. Merging two competing pages, deleting a thin collection, and fixing one snippet usually beats publishing a new article.
FAQ
How often should I look at the data?
Twenty minutes a week to spot changes, one decision a month. Daily checks on a site this size produce anxiety, not information.
How much traffic do I need before drawing conclusions?
There is no clean number, but 100 sessions is too few to judge a conversion rate. Use counts, direction, and customer messages instead. Wait for a few hundred sessions before comparing rates.
My rankings dropped. Should I change the page?
First check whether impressions and clicks also dropped. A position change of one or two spots with stable impressions is normal movement. If impressions fell too, check index status and internal links before editing.
Do I need more tools?
Search Console, GA4, and your store backend cover the first month. Add a rank tracker only if you can name the question it answers that these three cannot.
Can I connect AI directly to Search Console?
Technically yes, with API access and setup, but exporting a CSV takes two minutes and avoids handing over account access. Connect it later, after you know which two or three reports you actually use.
How long before a change shows results?
Weeks, sometimes months, depending on crawl frequency and competition. That is why one change per window matters: overlapping changes make the result unreadable.
Next step
That completes the new site loop: find the wedge, choose a domain, deploy, get indexed, structure the pages, find the first visitors, and read the data. The loop then repeats with a smaller version of the same questions.
Go back to the new site roadmap to see the whole sequence, or start with finding your wedge if you have not chosen one yet.