← Ethical Analytics

Analytics for a site with (almost) no traffic yet

Every analytics guide is written for sites with thousands of visitors, then read by people with fifteen. This one is for the fifteen. We run a portfolio of small new sites ourselves, so this is the advice we actually follow, including the uncomfortable parts.

First, subtract yourself

On a new site, the most dedicated visitor is you. Your own checks, your phone, the friend you sent the link to — at low volume these can be half the chart. Before reading any number, work out what your own footprint looks like (the days you know you were testing) and mentally subtract it. A launch-day spike of eight followed by three quiet days usually means five people saw your post once, not that you "lost momentum".

Numbers that mean something at low volume

• Referrers, from the first visitor. One visit from a search engine means indexing has started and a query you can rank for exists. One visit from a forum thread tells you where your audience actually talks. At low volume the referrer table is the most information-dense thing you have — read it before any total.
• Which pages get entered. If people arrive on an article rather than your homepage, search engines are doing the introducing, and that article's topic is your signal to write more like it. Entry pages are a ranked list of what strangers wanted.
• Direction over weeks, not days. 3 → 7 → 12 → 19 weekly visitors is a real trend. 12 → 9 is nothing. A useful rule of thumb: with counts this small, treat anything within ±5 as "the same number", and only re-read the trend every Monday rather than every morning.
• Any conversion at all. One signup, one email, one purchase teaches you more than a month of pageviews: someone crossed the gap from curious to acting. Look at what page they entered on and where they came from — that path is your first working funnel, however small.

Numbers that are noise at low volume

• Bounce rate. At seven sessions, one person changes it by 14 points. It starts meaning something in the hundreds, not before. (Also: on a single-page tool, a "bounce" is often someone who got exactly what they came for.)
• Device and browser splits. "60% mobile" from ten visitors is six phones. Check your site works on a phone — you should do that anyway — and ignore the ratio until the sample grows.
• Daily anything. Days are for spotting that a post went up somewhere; weeks are for decisions.
• Averages. Average session length across five sessions is one long visit in a trench coat. Medians need volume too.

What "normal" looks like (the honest part)

A brand-new site that nobody is promoting gets close to zero organic visitors for weeks — search engines take time to index and longer to rank, and until then traffic only arrives when you put the link somewhere. Single-digit weeks in month one aren't failure; they're what starting from nothing looks like. The question that matters isn't "is fifteen good?" but "is anything repeatable in it?" — one referrer that sent two people beats ten untraceable visits, because you can go earn more of the first kind.

A weekly routine that takes five minutes

Once a week, same day: note total visits and the trend arrow; read the referrer list and ask "can I get more of whichever of these was a stranger?"; read the entry pages and ask "what does this say people wanted?"; then decide one thing for next week — write the next article on the topic that got entered, or show up properly in the place that referred someone. Then close the tab for six days. At low volume, checking daily doesn't just waste time; it teaches you to see patterns in noise, and you'll act on them.

By CAAC, August 2026 — we build small web tools in public, and every number-reading habit above came from staring at our own single-digit weeks. Related: do you need a cookie banner just for analytics? what "cookieless" actually means, and Google Analytics alternatives compared.

Privacy · Terms · Docs · CAAC