How to interpret a property score
Every score in PropertyInvestiQ is a number from 0 to 100, with a colour-coded band and a confidence percentage. This article explains what those mean and how to act on them.
The five bands
| Band | Score range | Colour | What it means |
|---|---|---|---|
| Excellent | 80–100 | Green | A strong opportunity by every measure we check. Worth shortlisting immediately. |
| Good | 65–79 | Light green | Above average. Likely worth a closer look. |
| Moderate | 50–64 | Amber | Mixed signals. Read the breakdown before deciding. |
| Below average | 35–49 | Orange | Several flags. Investigate carefully or pass. |
| Avoid | 0–34 | Red | Multiple negatives. Usually not worth your time. |
What the score actually measures
The current scoring formula is tuned for the flipper persona — someone buying to refurbish and resell. Every lot starts at 50, and three components move it up or down:
- Price (±35 points) — the guide price per square metre compared with the median for other auction lots in the same postcode district (and property type where there are enough of them). We need at least three other lots to set that benchmark, and a lot never counts towards its own. Floor area comes from the EPC register, the listing’s own size, or an estimate from the bedroom count. If the price per square metre is implausible (below £500 or above £25,000), we treat the size or price figure as wrong and give no price signal.
- Success indicators (up to +30 points) — EPC upgrade headroom, a listing that has been on the market for 60+ days, and refurbishment wording in the listing description.
- Risk indicators (down to −35 points) — a guide price under 60% of the area benchmark (often a hidden defect), an EPC rating of G, and cladding or a lease under 80 years mentioned in the listing description.
The components sum to the final 0–100 score.
For lots listed through property portals we don’t have the listing’s own description, so the refurbishment, cladding and lease checks can’t run on them — check the listing yourself for tenure and cladding. Area trends (growth, sales volume, crime, buyer demand, valuations) and subsidence are part of the formula, but the nightly auction scoring doesn’t fetch them yet.
V2 coming: BTL and HMO persona variants will weight these factors differently (yield matters more than refurb margin, for example). Until they ship, the score reflects flipper economics.
Confidence percentage
Every score also carries a confidence percentage. It reflects how complete the data was for that particular lot — not how good the lot is. We run the checks above, and confidence is the share of the checks we could run for that lot that had real data. A check we can’t run at all (for example the description checks on a portal listing) doesn’t count either way; a check we ran but found no data for (for example no matching EPC certificate) lowers confidence.
Auction listings vary a lot in how much underlying data exists, so confidence varies too:
- Higher confidence — solid comparables and property data; the score rests on more signals.
- Lower confidence — thin comparables or a sparse listing. Treat the score as directional: read the rationale and run a deep analysis before relying on it.
- “Unrated” — no check moved the score off its starting point, usually because there were too few other lots nearby to benchmark the price. Don’t read anything into the absence of a score; check the listing details yourself.
Confidence is a data-completeness gauge. A high score with low confidence means “looks promising, but we’re working from limited data” — always read the rationale.
How to use scores in practice
- Filter the auctions or map view by band first — start with Excellent + Good only.
- Read the rationale on any shortlisted lot — it lists the top 3 factors driving the score in plain English.
- Run a deep analysis on the 1–2 lots you’re seriously considering. That’s where the AI does the heavy lifting — comparable selection, refurb estimate, exit-strategy commentary.
When a score feels wrong
Scores are deterministic — same inputs produce the same number. If a score looks off:
- Check whether the breakdown has a price factor. If it doesn’t, there weren’t enough other auction lots nearby to benchmark the price, so the score rests on the other checks alone.
- Check the floor area. An estimated size (from the bedroom count) can be well off for large or unusual properties.
- Check whether the property has unusual features (Grade II listed, ex-council, non-standard construction, flood risk) that the formula doesn’t yet flag.
- Get in touch if a specific score seems materially wrong — outliers help us improve the model.