METODOLOGIA // DOKUMENTACJA ALGORYTMU TOPLISTA SCORE
How we calculate Toplista Score. Formula, sources, audit.
The algorithm is public and versioned. The same formula scores every company — no exceptions, no paid placements.
Recomputed: 4 Nov 2026 • Average score: — • Formula version: v1
Toplista Score formula
DETERMINISTIC_SCORING_v2.4
score = 40 × reputacja + 25 × stabilność + 20 × doświadczenie + 15 × oferta
reputacja = 0,60 × bayes(rating, m=30, C=4,5)
+ 0,25 × log10(opinie + 1) / log10(1000)
+ 0,15 × wskaźnik_odpowiedzi
stabilność = 0,35 × min(1, lata/15) + 0,25 × przychód(log)
+ 0,20 × rejestr+VAT + 0,20 × brak_wpisu_KRZ
doświadczenie = 0,60 × min(1, log10(realizacje+1)/log10(2000))
+ 0,40 × min(1, lata/15)
oferta = 0,40 × min(1, gwarancja_lat/25)
+ 0,30 × udział_marek_Tier1 + 0,30 × własna_ekipa
total = NULL gdy coverage < 60% (brak oceny przy niewystarczających danych — nigdy fałszywe 0)
Data sources
- Google Places — rating and review count (aggregate only)
- KRS, GUS, CEIDG — revenue, headcount, company age
- Own reviews — from 5 published they feed reputation
- Debtor registers — no KRZ entry
Rules we enforce
No paid placements in the ranking
Declared data counts at weight 0.5
Below 60% coverage means “No score”, not zero
Formula version and recompute date published next to the score