AI Analytics

An analyst that reads your traffic every morning

Lomen AI does not generate more charts. It reads the pageviews you already collect, finds the two or three things that changed materially, and tells you what to do about them in language you can forward to your co-founder.

How it works

Four steps, no black box

Every insight shows its inputs. If Lomen claims your changelog is your best entry page, you can click through to the exact pageview and pages-per-visit figures behind the claim.

1 · Detect

Statistical baselines watch pageviews, unique visitors, entry pages, referrer mix and device split, flagging deviations that fall outside your site's normal weekly variance.

2 · Connect

Each deviation is read alongside the pages and referrers around it, so an insight points at a probable cause rather than restating that a line went up.

3 · Explain

A language model writes the finding in plain English using only the figures Lomen computed, with every number traceable back to your own collected pageviews.

4 · Qualify

Lomen attaches a confidence level, ranks the findings so the top item is genuinely the top item, and says plainly when there is not enough data to claim anything.

Sample output

Insights from the demo website

These are verbatim examples of the kind of summary Lomen produces for an illustrative site's traffic.

PagesHigh confidence

Your changelog is quietly your best entry page

Sessions that start on /changelog view 2.3× more pages than sessions that start on the home page, and they are growing 18.4% month over month. Linking it from the top navigation is the cheapest experiment available to you.

ReferrersMedium confidence

One referrer is driving last week's spike

62% of last week's extra visits arrived from news.ycombinator.com, and those visitors read a single page then leave. A tailored call to action on /blog/how-we-ship is where that attention is being lost.

DevicesHigh confidence

Mobile visitors drop off before /pricing

Mobile is 38.0% of visitors but only 21.4% of pricing pageviews. The gap has widened for three weeks, which usually points at layout or load time on the path into pricing rather than intent.

Ask anything

Conversational analysis, with citations

Ask “why did traffic jump last Tuesday?” and Lomen answers with the referrer that sent the visits, the page they landed on and how many read a second one — each figure drawn from your own data.

Why did traffic jump last Tuesday?
Pageviews rose 24.1% (1,392 extra views) on Tuesday. 71% of the increase arrived from news.ycombinator.com and landed on /blog/how-we-ship. Those visitors averaged 1.2 pages per visit, against 3.76 for the site overall.Sources: Pageviews · Top pages · Referrers

Your data is never used for training

Prompts and results are processed for your account only. We do not train models on customer data and we do not share it with anyone else.

Numbers come from computation, not the model

Every figure in an insight is calculated from your collected pageviews. The language model writes the sentence around numbers it is given; it never invents them.

Insights say when data is thin

If there is not enough traffic to support a claim, Lomen says so instead of manufacturing a pattern out of a handful of visits.