51,000 hh.ru applications: what drives an invitation
51 тысяча откликов на hh.ru: что на самом деле влияет на приглашение
B · Useful, limited The founder of XAGO, an auto-apply service for hh.ru, publishes a clear method and outcome tables on 51,583 applications, but the sample is his own customers and the analysis is descriptive.
Summary based on the full text·Original: about 6 minutes to read
In short
A Habr post (19 October 2026) by the maker of XAGO, a service that sends hh.ru applications automatically from the job seeker's account. Between 8 August and 30 September 2026 it sent 242,000 applications for 284 people; hh.ru returned an outcome for 51,583 of them from 191 people. The author uses these outcomes to test common job-search advice. The largest effect is the difference between individual candidates, not timing or cover-letter length.
Key numbers
- 5.2%applications that received an hh.ru invitation status, about one in 19
- 38.9%applications the employer never opened, or the vacancy closed first
- 10.9% vs 3.8%invitation rate for sales vacancies versus software development
- 1.7% vs 10.2%invitation rate of the bottom and top 10% of job seekers (174 people with 50+ outcomes)
- 2.2% vs 5.7%invitation rate for vacancies with screening questions or a test versus ordinary vacancies
As reported by the source.
Russian·Web page·opens habr.com
What it reports
- Funnel per 100 applications: 38.9% never opened or vacancy closed, 17.7% opened with no reply, 38.2% rejected, 5.2% invited.
- Employer response time: median 30 hours; 45% within a day, 75% within 5 days, 90% within 11 days.
- Hour of sending: 6:00-7:00 Moscow time gives 6.0-6.4% invitations, noon 3.7%, 19:00 7.6% (best hour). Friday 6.8%, Tuesday 4.2%.
- Cover letters under 400 characters: 5.4% invitations; 400-800: 4.0%; 800-1,200: 5.2%. Longer letters add nothing.
- Volume per person does not lower the rate: under 300 applications 5.9%, over 2,500 applications 5.8%.
- Vacancies with questions or a test: 2.2% versus 5.7% for ordinary vacancies, a 2.6x gap.
- By field: sales 10.9%, design 8.8%, marketing 8.1%, software development 3.8% (only 53.5% opened), product and project management 3.6%.
- Per-candidate spread: bottom decile 1.7%, median 4.0%, top decile 10.2%. The author cannot separate CV, profession, city or vacancy choice as the cause.
How they got it
Counts come from the service's own database. The bot sends from the candidate's hh.ru account and then reads the status in the account's applications page. Data were pulled on 13 October, leaving at least two weeks for employer reactions. Invitations sent only as chat text without an hh.ru status are not counted. The author checks plausibility against hh.ru's limit of 200 applications per profile per day, which was never exceeded on a single profile.
Limits
- The sample is the author's customers: active, high-volume job seekers who chose their own filters. The author says it is not a labour-market study.
- Outcomes exist for 191 of 284 people; hh.ru did not return statuses for the rest.
- Single-variable tables with no controls: profession, city and filters may drive the timing and volume patterns.
- The author sells the auto-apply product whose behaviour the results bear on.
Why it matters for companies
From the employer side this is a rare outside view of a recruiting funnel under bot-driven application volume: four in ten applications are never opened, and development roles are opened least. It also shows that a mass-apply tool does not dilute invitation rates per candidate, which means screening load rises without an obvious quality signal in volume. Recruiting teams can use the response-time distribution and the effect of screening tests as a benchmark for their own funnel data.
Key passages in translation
Russian original
Почти 40% откликов работодатель даже не открывает.
Almost 40% of applications are never even opened by the employer.
Russian original
Разница между людьми — в шесть раз. Час отправки даёт два раза, длина письма — ещё меньше.
The difference between people is sixfold. The hour of sending gives twofold, letter length even less.
Russian original
Сразу о том, чего здесь нет: это не исследование рынка труда.
First, what this is not: it is not a labour-market study.
Summary based on the full text. Read the original for the full argument and data.Read the original on habr.com
- Tags
- Labour market