How Nichify scores a Vinted niche

The 4 signals

1. Coherence (cluster purity)

Coherence measures intra-niche dispersion: how closely the listings that make up a niche actually resemble one another. A coherent "Y2K" niche groups pieces that share the same visual cues (typography, loud colours, baggy fit). A phantom niche contains a hodgepodge of listings the algorithm bundled together by accident.

Scale: 0 (incoherent) to 100 (perfectly coherent). Confidence threshold: 60.

Why it matters: a low-coherence niche means the algorithm is still trying to isolate the signal, so the median price recommendations are noisy.

2. Monthly volume

Estimate of the number of transactions visible in the niche over the past 30 days (confirmed sales only: status Sold or Reserved at the time of the scrape).

Note: Vinted does not publish total sales, so Nichify reconstructs this estimate from the state transitions observed across the sample.

3. Median sold price

Median of confirmed sale prices over the window. The median is preferred to the mean to neutralise outliers (luxury pieces that slip into an entry-level niche).

The price shown in the UI is always the sold price, never the asking price, which can stay open for months without market validation.

4. Trend

Comparison of volume over the past 30 days against the previous 30 days. A trend >+15% flags a niche that is taking off; a trend <-15% flags a niche that is saturating.

Important: this signal arrives with a ~7-day delay because Sold to Available transitions (buyer cancellation) are confirmed late.

Data window

• Starter plan: 30 days of history.
• Pro plan: 90 days.
• Expert plan: 365 days, plus access to seasonal signals (a niche's anniversary, Q4 peak, post-Christmas).

The longer window matters because it lets the algorithm tell a one-off spike (a 48-hour TikTok effect) from a structural trend (Y2K since 2024).

Sample per niche

A niche is only kept in the scoring if it has at least 40 active listings AND 15 confirmed transactions over the window. Below this threshold, statistical noise outweighs the signal and Nichify flags it as "emerging niche, insufficient data" rather than fabricating a score.

The 40/15 threshold comes from an internal study on median price stability: below it, the median moves by more than 20% from one scrape to the next.

Refresh cadence

• Active niches: rescanned every 24h.
• User monitors: rescanned every 6h (all plans), 1h (Expert plan).
• Public niche pages (/niches/*): refreshed once a week, on Sunday evening.

The frequency is not higher because Vinted does not move that fast: between two 24h scrapes, fewer than 5% of niches gain or lose their "actionable" status.

What Nichify does not measure (yet)

• Net margin per reseller: we do not have access to the purchase price. A high Nichify score signals a good resale price, not a guaranteed margin.
• Time to sell: we do not surface the "average time to sale". High volume is a fair but imperfect proxy.
• Long seasonality (12+ months): only the Expert plan reaches 365 days. Multi-year seasonality is on the 2026 roadmap.

Sources and accountability

The algorithm and its code are closed (commercial reasons), but the signals shown in the UI are fully traceable: every niche has a "view source listings" panel that lists the Vinted URLs used in the calculation. You can always audit the sample.

Question, disagreement or methodological flaw spotted? [email protected] · reply within 5 working days.