Transparency · rules version 1
How company statistics are calculated
Company pages show how often applicants hear back, reach an interview or get ghosted. The numbers come from JobStacky members tracking their own applications. Raw shares can mislead: if many people apply to a job they were never a fit for, the company looks harsh when it is not. So JobStacky weights each application by how well the applicant's profile fits the job. Everything below is the complete rule set. It uses fixed rules only, no AI, and it is the same for free and paid members.
1. Which applications count
Only real applications count. Jobs a member only viewed, marked as not interested, or where the posting expired before they applied are left out.
2. Fit level of each application
Each tracked job gets a fit level from up to four checks. Each check that has enough data scores good (2), partial (1) or poor (0). The fit level is the average of the checks that ran: 1.5 or more is good, 0.75 or more is partial, lower is poor. If no check could run, the level is unknown.
| Check | How it works | Result |
|---|---|---|
| Kind of role | The job title and the member's three most recent job titles are each sorted into a role family (engineering, data & AI, product & design, sales & growth, marketing, customer success, finance & operations, people & HR, legal & compliance) by keyword rules. | Same family: good. Different family: poor. Skipped when either side has no recognisable family. |
| Seniority | The job's level is read from its title (intern, entry, mid, senior, lead/manager, director, executive). The member's level comes from the years covered by their work history: under 1 year intern, 1–3 entry, 3–6 mid, 6–10 senior, 10+ lead. | Job 2+ levels above the member: poor. Job 3+ levels below: partial. Otherwise good. Skipped when the title states no level or the profile has no dates. |
| Skills | The skills listed on the member's profile are looked up as whole words in the job title and description. | 3 or more found (or at least 30% of their skills): good. 1–2: partial. None: poor. Skipped without listed skills or job text. |
| Language | When the job states a language other than English, it is compared with the languages on the member's profile. | Listed: good. Not listed: poor, and the whole fit becomes poor. Skipped when either side is missing. |
The fit level is recalculated when a job is added or edited, and every night for members who changed their profile. Each member can see the fit level of their own tracked jobs and the reasons behind it in their job tracker. Nobody else can: company pages only show totals, and the split by fit level only appears once at least 3 different people have applied.
3. How much each application counts
| Fit level | Weight |
|---|---|
| Good fit | 1.0 |
| Partial fit | 0.6 |
| Unknown (no profile or not enough detail) | 0.5 |
| Poor fit | 0.2 |
Two more rules keep a few people from dominating a company's numbers:
- At most 2 per person. Each person counts for at most 2 applications per company. Someone with 10 applications at one company counts 0.2 for each.
- Spray applying. A person with 10 or more applications in total, of which at least 70% are a poor fit, counts half.
4. Which numbers are weighted by fit
- Weighted by fit: reached interview (including offers and rejections after an interview), got an offer, rejected without an interview. These mostly reflect whether the applicant fit the job.
- Not weighted by fit: heard back (any assessment, interview, offer or rejection) and ghosted. Whether a company answers is its own behaviour, whoever applied. The per-person and spray rules still apply.
5. Small samples
Every rate comes only from that company's own applications. Nothing from other companies is mixed in:
rate = Σ (weight × outcome) / Σ weight
Outcome is 1 when the application reached that result and 0 otherwise. A company with one or two reports could look perfect or terrible by chance, so the rates and the verdict only appear once there are at least 5 applications from at least 3 different people. Until then the page shows “Not enough reports yet” with the raw counts.
6. What is always shown alongside
Every company page also shows the unweighted shares, the number of applications and people behind them, and the site-wide interview rate as a comparison only (it never changes a company's own numbers). Statuses are reported by members themselves and can be wrong; treat the numbers as signals, not facts.
Related
See companies hiring now on Companies hiring now, or browse all company pages. Questions about these rules? Contact us.