11 weeks of Search Console data from a new site, including the part that failed

Most write-ups about ranking a new site come from people who already won. They tell you what worked, in order, as if it were a plan rather than a reconstruction. I want to publish the other version, with the actual numbers, while the outcome is still genuinely uncertain.

I launched a browser-based file tools site on 8 May 2026. Everything runs client-side — no uploads, no server, WebAssembly and canvas in the tab. Eleven weeks later I have 42 pages, 1,445 impressions in the last 28 days, and 2 clicks.

That number isn't the interesting part. The interesting part is what the data says about why, and how badly wrong I was for the first two months.

Part 1: two months of building the wrong thing

The launch plan was thirteen tools: image compress, resize, crop, format convert, PDF merge, split, compress, PDF to JPG, and so on. The reasoning was that these are high-volume evergreen queries, the tech is straightforward, and TinyPNG and Smallpdf have proven the model.

Here's what Search Console said after two months, 28 days to 5 July:

Sitewide: 5 clicks, ~550 impressions.

Position 80 to 95 is not "ranking slowly." Nobody has ever seen page nine of a search result. Those are the positions Google assigns when it has indexed your page, understood it, and concluded there are eighty better answers.

The pages weren't bad. They're fast, the schema is right, the tools actually work and don't upload your files. That's precisely what made the data useful: quality wasn't the variable. I had picked a fight with TinyPNG, iLoveIMG, Canva, Adobe and Google's own Squoosh, from a domain with no authority and no links. A better-executed version of the same page would have ranked 84 instead of 87.

But two rows in that same export looked completely different:

Same domain. Same age. Same zero backlinks. Position 9.

The only difference was the topic. HEIC is the format Apple made default on every iPhone in 2017 that Windows still can't open properly. It's a real recurring problem, and the big converter sites treat it as one row in a format list rather than something to own.

Part 2: the pivot

On 10 July I stopped building generic converters and repositioned the entire site around what I started calling orphan formats: files one ecosystem creates that nothing else can open. HEIC from iPhones. AVIF from web pages. JFIF, which is Windows' own bizarre habit of saving JPEGs under a different extension. Apple's .numbers spreadsheets, which Excel refuses outright.

Three days of work, shipped 13 July: three new converter tools, four explainer articles, and a homepage that stopped saying "free image and PDF tools" and started saying "open the files your device can't."

The thirteen generic tools are still live and still indexed. I just stopped linking to them from anywhere on the site. Deleting them would throw away whatever long-tail Google already gives them, but a site that visibly is one thing reads as more authoritative than one padded with eleven me-too converters.

Part 3: what the data says now

28 days to 18 July:

• 1,445 impressions, up from ~550
• Impressions per day went from 11–19 to 60–260 after the 13 July push
• 21 queries ranking positions 1–10

Every one of those 21 is an orphan-format or informational long-tail:

Pages published on 13 July were on page one within days. No backlinks pointed at them.

Meanwhile the head terms remain exactly as locked as they were:

That contrast is the whole lesson, and it's a sharper version of what I got wrong in May. It isn't that a new site can't rank. It's that a new site can rank immediately for questions nobody has bothered to answer well, and never for the terms everyone is fighting over. Same site, same week, positions 3 and 90.

Part 4: so why only 2 clicks?

Because the queries I can win are tiny.

Each of those page-one queries carries 1 to 6 impressions over 28 days. At position 8 with 3 impressions, expected clicks is about 0.3. Two clicks across the whole site is statistically unremarkable, not a failure signal.

I want to be careful here, because this is exactly the kind of correct math that lets you feel fine for six more months while nothing happens. So let me state the uncomfortable version too.

Run the arithmetic forward. If each new guide wins 20 to 30 of these long-tails, and each carries a handful of impressions, a guide is worth maybe 5 to 10 clicks a month at maturity. Publish two a week for six months and you have roughly 50 guides and maybe 400 sessions a month. At display-ad rates that's two to five dollars a month.

The strategy is working and the strategy's ceiling is also very low. Both are true. Coverage compounds, but it compounds from a small base, and I don't yet know whether the curve bends before I run out of patience.

Part 5: four things in the data I hadn't expected

Blogs massively outrank tool pages. My articles sit at positions 15–21. The tool pages they link to sit at 53–76. Same site, same internal linking, same technical setup. Google appears far more willing to rank an explanation than a utility. I now treat the articles as the front door and the tools as what happens after.

Search Console hides most of your queries. My HEIC article shows 358 impressions at the page level, but the named queries for it sum to about 18. The rest is anonymised low-volume long tail that GSC won't show you. If you optimise only against the queries you can see, you're optimising against a few percent of reality.

One referring domain is a hard ceiling. I have five dofollow backlinks and all five are from dev.to. Google counts referring domains, not links, so functionally I have one. Everything sitting at positions 11 to 21 is stuck there because of that, not because of the content. Directories are all nofollow or paid. AlternativeTo rejected the site for lack of notability. This is the actual constraint and I don't have a solution yet.

Google Trends corrected a bet I'd already made. Keyword Planner showed AVIF conversion volume down 90%, which I assumed was a measurement artifact after a 2024 spike. I checked properly. Against heic to jpg on a five-year normalised series, avif to jpg peaked at 8 in December 2024 and now sits at 4: down 50% from peak, down 33% year on year, still sliding, corroborated by avif file to jpg at -80% in related queries. So I was half right and it didn't matter. AVIF is now a maintenance asset, not a growth area. I'd already shipped the pages. Cheap lesson, but it should have come before the build.

What I'd tell myself in May

Competition selection dominates execution quality for a new domain. I don't mean that as a slogan. I mean that the same team, the same stack and the same content quality produced position 87 in one topic and position 9 in another, in the same month, and the only variable was who else was in the room.

If you're building anything that depends on search, look at your Search Console positions before you look at anything else. Not clicks, not impressions. Positions. If your pages sit in the eighties, no amount of writing better will fix it, because writing better was never the problem. Change the room.

I'll post the next set of numbers in a few months, including if it goes badly. I set a kill trigger before starting — under 500 sessions a month by November means I reassess, and I've written it down publicly now specifically so it's harder to quietly move.

The site is fileoholic.com if you want to look at what these numbers describe. Happy to answer questions about any of it, including the parts that aren't working.