Time-to-signal, and Rituel

Two questions, one answer · 17 August 2026 · Russian market · method and caveats at the foot

The honest answer to "how long until we know" is: far sooner than the site takes to mature — but only if you stop asking about the site. A site's traffic is a lagging, blended number that can hide a dead investment inside a growing total. We have that exact case in our own portfolio, measured. Judge each bet separately and the verdict arrives in weeks instead of years.

For Rituel specifically, the timing question is close to moot — and the naming question is not. It is a social-first product, where a verdict costs weeks and no domain at all. But «ритуал» in Russian is the funeral trade, Rituals Cosmetics runs rituals.ru in this very market, and that company has already won a court case over the mere word "ritual" in cosmetics.

Part one

1. How long does a site take to reach its potential?

The published figures are discouraging, and they are all about Google:

Russian-market advice puts the Yandex "sandbox" at 2–12 months, commonly 6–12 for a young domain. But the mechanism matters more than the number: Yandex is described as deliberately lifting a page toward the top to watch how users behave on it. That behavioural test is not an obstacle — it is the early signal, arriving long before stable rankings.

The unit of judgement is what's wrong

The most useful thing in our own data is not the growth. It is what the growth concealed.

Over the same window, anekdotnik's single largest body of SEO work — a 60-topic taxonomy, an LLM tagging run across 30,000 jokes, re-tagging passes, canonical fixes, 58 tag URLs in the sitemap — produced 16 visits, 0.64% of traffic, at a 62.5% bounce rate, the worst section on the site. Not one of the 60 topic queries appeared anywhere in the search report. Meanwhile three curated sort-order pages carried 28% of all search traffic.

Both facts are from the same twelve days. Ask "is the site working?" and the answer is a triumphant yes. Ask "is each bet working?" and one bet was already dead, visibly, in under two weeks — and it was the expensive one. Site-level traffic is the slowest possible way to learn something you can know far sooner.

Can the timeframe be shortened? Yes — a ladder of leading indicators

Each rung arrives before the one below it and answers a different question. Instrument them and you buy months.

1. Before you buildSearch demand. Wordstat costs an afternoon and kills more bad ideas than any other check. On anekdotnik the demand research predicted the taxonomy failure before the traffic did — bare modifiers dwarfed topic queries, and live data later reproduced that prediction exactly. This is the cheapest kill available and it is available on day zero.
2. Week 1–3Index coverage. The one datum that separates "not indexed yet" from "indexed and not ranking" — which is the difference between too early and wrong. Without it you cannot tell a patience problem from a product problem, and you will wait out a bet that was never going to work.
3. Week 2–6Impressions on non-branded queries. Visibility precedes clicks; impressions are the standard leading indicator that the engine is willing to show you. Rising impressions with flat clicks is a title/intent problem, not a ranking one — and it is fixable in a day.
4. Week 4–8Behaviour on arrival — bounce, depth, time. In Yandex this is both a leading indicator and the causal lever, since behaviour is what the test traffic is measuring. anekdotnik's 62.5% tag-page bounce said the pages underdelivered regardless of ranking.
5. Month 3–6Traffic. The lagging number. By the time it speaks, every rung above it has already told you the answer.

What cannot be shortened, and one caution

Domain age and accumulated authority are real and are not compressible — that is what the 72.9% figure is measuring. What is compressible is the time to a decision. And one caution against over-reacting: our own taxonomy verdict rests on twelve days at ~57 organic visits/day. Directionally strong, statistically thin. A ladder rung is a reason to investigate, not to kill on a week of data.

Part two

2. Rituel

A visual makeup-recipe platform: looks from makeup artists and influencers rendered as step-by-step storyboards — what to apply, where, with which tool, in what quantity, to what effect, with illustrations, Reel timestamps, technique notes and checkpoints.

The demand, and its intent

Measured today via the Yandex Direct v4 Wordstat endpoint, Russia, monthly, broad match. The head term макияж is 1,306,840/mo — but the head is not the market, because most of it is not looking for a tutorial:

The single largest sub-term, перманентный макияж, is a salon service — someone booking a cosmetic tattoo, not learning a technique. The genuinely instructional band — как сделать макияж, макияж пошагово, уроки макияжа, the occasion and technique terms — runs in the low tens of thousands each.

For scale: a single term in the games category we are already in, дурак онлайн, is 474,404/mo. Rituel's entire addressable search band is roughly an order of magnitude smaller than one keyword in the other portfolio. That is not disqualifying. It does mean search should not be the plan.

The competitive field

The finding that matters most: this is not a search product

The risk is the mirror image of that gift: a product whose distribution is someone else's platform, over source material — other people's looks and Reels — that it does not own. That is a rights question and a dependency question, and both are worth answering before the first storyboard is drawn, not after.

The name has three separate problems

How to know in six weeks, not six months

1. Weeks 1–2 — publish ten recipes where the audience already is. No site, no domain, no build. The product's whole claim is that a storyboard makes a look easier to recreate; that claim is testable as ten posts.
2. The metric is not reach, it is saves and returns. A recipe is used later, at the mirror. Save rate and repeat visits are the behaviours that distinguish this from ordinary beauty content; likes do not.
3. Weeks 3–4 — charge someone. The category proves willingness to pay through courses. One paid pack of recipes tests the business, not the format.
4. Only then build the site, and build it for the terms measured above rather than for the head term — entering at макияж пошагово and the occasion terms, where intent matches what the product actually is.
5. Settle the name before any of this reaches print. Points 1 and 2 above are not softened by traction; they get worse with it.

Method, and what I did not check

Demand figures come from the Yandex Direct v4 Wordstat endpoint, Russia (GeoID 225), monthly, broad match, pulled today; volumes are nested inside their head term and must not be summed. Portfolio figures are quoted from anekdotnik's own SEO performance review against live Яндекс.Метрика (counter 108508005) — and that review's own caveat stands: twelve days at ~57 organic visits/day is directionally strong and statistically thin. Domain status came from live WHOIS against whois.tcinet.ru and whois.verisign-grs.com; occupancy was determined by fetching each site and reading what came back. Page weights and TTFB are single unauthenticated fetches from a European datacentre — ratios, not absolutes. The ranking-timeline figures are third-party studies of Google, not Yandex, and the Yandex sandbox figures come from RU agency blogs rather than from Yandex; treat both as indicative. Web search ran in a US region and under-represents the Russian SERP, so the competitor list is a sample, not a census. I did not search the Rospatent trademark register, did not obtain a price for any parked domain, and have no traffic data for Rituel because it does not exist yet.