Scoring
How a deal is scored
Every deal gets two things: a verdict for each strategy (strong, ok, weak or fail), which says whether its numbers meet your targets, and a match score from 0 to 100, which ranks deals against each other. Both are estimates to help you decide what to look at, not advice on what to buy.
In short
- The verdict says whether a strategy’s numbers meet your targets; the 0–100 score only ranks deals against each other.
- A strong verdict must also pass extra checks on the valuation and the evidence behind it, so good-looking numbers on thin data aren’t called strong.
- The score blends the best verdict, the price against our value, the AI’s reading of the advert and our confidence; it’s held at 40 if nothing your search wants works.
- Targets and weights are defaults you can change on your Assumptions page; every rule change is in the dated changelog (now version 23).
These are the rules in version 23. Figures marked “by default”, and every target in the table below, are our defaults: you can change them, and the score's weights, on your Assumptions page, and every figure is then worked out with yours. The other rules are fixed. What changed in each version.
1. A verdict for each strategy
Each strategy is worked out with stamp duty, fees, finance and running costs, then compared with these targets (our defaults; each band starts at its figure, so “8%+” means 8% or more):
| Strategy | Strong | Ok | Weak | Fails |
|---|---|---|---|---|
| Flip | Profit on GDV of 20%+ and at least £20,000 profit (£15,000 for accounts opened from 12 November 2026) | 15%+ profit on GDV, or 20%+ with less than the minimum profit for strong | 10% to under 15% | A loss or no profit, or under 10% |
| BRRR | At most £10,000 left in after the refinance and £250+ a month cashflow | 15%+ return on capital left in, or all money out | 5% to under 15% return on capital left in | Fails the lender stress test, negative cashflow, or under 5% return on capital left in |
| Buy-to-let | 8%+ return on cash and £200+ a month cashflow | 5%+ return on cash | 3% to under 5% return on cash | Negative cashflow, or under 3% return on cash |
| HMO | 12%+ return on cash and £600+ a month cashflow | 8%+ return on cash | Under 8% return on cash, with cashflow of £0 a month or more | Negative cashflow |
- HMO is worked out only for houses with 3 or more bedrooms where local room rents are available.
- Below £60,000 by default, most lenders won't lend: buy-to-let and HMO figures assume a cash purchase, buy-to-let is then at most weak (see below), and a BRRR refinance fails.
- The value the figures use: ours, or an independent automated valuation of the home when it's lower. Where one from the last 60 days is below ours, every figure, the below-value wording and the score's below-value signal use it, our value after works is scaled down by the same ratio, and the deal shows both. It's refreshed before it's 60 days old while the listing is shortlisted or matches a search; if it isn't, the figures go back to ours and the deal says it expired. Whether the two agree (section 2) always compares ours with theirs.
- The works estimate: by default £200, £450 or £900 per m² for light, medium or heavy works (the AI's reading of the advert), plus items that scope doesn't include (such as a roof or windows) and 10% contingency. By default it's ×1.20 in London, ×1.08 in the South East and ×1.03 in the East of England, where building costs more; other regions use the UK rate until their build costs are verified. The floor area is the energy certificate's, else the advert's, else a typical one for similar homes on the same street (good matches only), else a typical size for the bedrooms (the works then say “estimated from a typical size”). A floor area under 15 m² a bedroom is ignored, and so is a house's advert figure under 70% of the national minimum space standard for its bedrooms.
- Auction lots are analysed at the higher of the guide plus 20% and 80% of our own estimate, by default: lots usually sell above guide. That floor always uses our value, even when the figures use a lower independent one, so a lower value never lowers the price we assume you'd pay.
- Buy-to-let mortgage: sized on the lower of the price and our own estimate, when our estimate is Fair or Good confidence and more than 10% below the price, as lenders lend on the lower of the price and their valuation. This check always uses our value, not a lower independent one.
- BRRR refinance: by default 75% of the value after works, but no more than the rent covers at the lender's stressed rate (by default the rent must be 125% of the interest), and no more than keeps your cashflow at your minimum when a smaller loan would. The deal card says when either limit applies.
- The stress test depends on the mortgage product in your assumptions: a 5-year fix is tested at the pay rate, or 5.5% if that's higher; a 2-year fix at the pay rate plus 2 points, or 5.5% if that's higher. By default, accounts opened from 12 Nov 2026 start on a 5-year fix and older accounts on a 2-year fix.
- Offers: when a strategy doesn't work at the asking price, the deal says the price below which it would, but only up to 50% below asking. Beyond that it says it doesn't work at any realistic price, and why. The 50% is fixed.
- “Figures look wrong: check the inputs”: an amber flag on a flip loss of more than 60% of the value after works (unless it fails even at the top of our range, when the verdict stands whatever the value). It doesn't change the verdict or the score, but such deals are never used as homepage examples, free-preview picks or the feed's headline figures.
- Flip and BRRR figures are worked out on the low end of our value after works, which is never below the home's value as it is (the value the figures use: ours, or the lower independent one). They're withheld, and the deal says it's too uncertain to value reliably, only when that range is more than 1.6 times as wide at the top as at the bottom, or our valuation is Low confidence and the low end had to be lifted by more than 15% to reach the value as it is. Buy-to-let and HMO figures still show.
- Other Low-confidence values, and ranges more than 1.3 times as wide, show their figures with a “Wide range: check carefully” caution.
- These withholding limits (1.6 times, 15% and 1.3 times) are fixed, not assumptions you can change.
- A flip that loses money even at the top of our value range says it fails even there, with that best case, rather than the bigger loss at the low end.
When a verdict is lowered
Some checks lower a verdict below what the numbers above give. The deal card then says so beside the verdict, e.g. “BTL: weak — would be ok, but the rent is an area-wide estimate”, and every fail gets a one-line reason from the card's own figures.
- Any strategy, strong shown as ok: when the value behind it fails one of the checks in section 2 (confidence, a second opinion, a price far below our value), or a listing flag rules strong out (an auction guide far below our estimate).
- Buy-to-let and HMO, strong shown as ok: a yield above 12% gross (buy-to-let) or 15% after bills (HMO) by default, unless both the valuation is Good confidence and the rent estimate is high confidence. Yields that high are often a wrong rent or a problem with the property.
- Buy-to-let, at most weak: when no lender would lend on it, at purchase or after works (cash buyers only, a short lease, non-standard construction, priced below the usual lender minimum, or a flat under 30 m² that needs works, unless you've ticked on the deal page that a lender will lend on it), and the figures assume cash; or when the rent is an area-wide estimate rather than comparable lets. An auction lot isn't counted as one no lender would lend on: it can't be mortgaged in time to complete, but it can be refinanced, so its buy-to-let figures assume cash without the cap (or, if it needs works, a purchase then a refinance).
- A flat under 30 m² that doesn't need works is worked out as a cash purchase, as most lenders won't lend on one, but isn't capped at weak, as it could be refinanced. Tick on the deal page that you've checked a lender will lend on it to see the figures with a mortgage.
- Buy-to-let on a home that needs works before letting (the AI's reading of the advert: heavy works, or a renovation need of 7 out of 10 or more) isn't lowered: the works estimate and the rent lost until it's let (by default 1, 3 or 6 months for light, medium or heavy works) are in the cash put in. An auction lot that needs works is worked out as bought with cash or bridging and refinanced on its value after works.
- Buy-to-let and HMO, at most weak: when the asking price is above the top of our value range for the home as it is: you'd be overpaying, and a lender may value it lower.
2. The checks a strong verdict must pass
Every strategy's figures rest on our valuation, so a strong verdict is shown as ok instead when:
- the valuation isn't Good confidence and the asking price isn't at least 15.9% below it (the Fair band's typical error of 10.9% plus 5 points);
- the valuation isn't Good confidence and there's no independent valuation of the home yet, whatever the gap;
- an independent automated valuation differs from ours by more than 10%, either way (our valuation's confidence then also drops one level: Good to Fair, Fair to Low; Low stays Low);
- our value is more than 30% above the asking price and the independent valuation hasn't confirmed it;
- the asking price is a third or more below our estimate, which usually means something the data can't see (tenure, condition, or it isn't a standard home);
- it's an auction lot whose guide is under half our estimate.
These checks are fixed, not assumptions you can change. Deals that don't pass them aren't featured as strong or sent as same-day alerts. How accurate our valuations are.
3. The match score, 0 to 100
The score blends six signals, weighted as below by default. A signal without data (say, a listing the AI hasn't read yet) is left out and the rest are re-weighted, so a deal isn't marked down for what we don't know.
| Signal | Weight | What it measures |
|---|---|---|
| Best strategy’s verdict | 40% | The best verdict among the strategies your search wants: strong counts in full, ok 70%, weak 35%, fail nothing. |
| Below our estimate | 15% | How far the asking price is below our estimate less its typical error (7.0% at Good confidence, 10.9% at Fair, 18.3% at Low: half of homes sold further from the estimate than that), with full marks at 20% or more by default. The same test as the “X% below our estimate” label on a deal. The value is ours, or the lower independent one (section 1). A price at or above that figure scores nothing here: it isn’t clearly below value. Scores nothing when the valuation is Low confidence (as it is for every studio, for now), or our value is unusually far above the price and an independent valuation hasn’t confirmed it, or the gap is explained by a short lease, a repossession or the home’s condition, so a suspect “discount” never lifts a deal. |
| Kitchen-diner potential | 10% | The AI’s 0–10 reading of the advert’s photos and floor plan. |
| Extra-bedroom potential | 10% | The AI’s 0–10 reading, for example a loft or a large room that could be split. |
| Renovation need | 10% | The AI’s 0–10 reading. Counted only when your search wants a flip or BRRR: if you’re letting as it is, works aren’t a plus. |
| Confidence | 15% | How much evidence the valuation and the AI had. |
- If none of the strategies your search wants has numbers that work, the score is held at 40, however good the other signals.
- For flips and BRRRs, a valuation below Fair confidence (by default) scales the verdict down. Buy-to-let and HMO verdicts rest on the rent and the price, so they aren't cut, but say a lender's valuation could come in below the price.
- A home on the market over a year loses 5 points, so it ranks below an otherwise similar fresh listing; over two years it's left out of the feed by default and out of alerts.
- Scores of 75 or more are our top matches, and the default threshold for same-day alerts on Priority.
- The weights, the 20% for full marks, the Fair-confidence threshold and the alert score are defaults you can change. The hold at 40, the 5-point penalty and the one- and two-year limits are fixed.
Every deal page shows its own score breakdown, signal by signal.
4. Three separate signals on every deal
A deal with attractive numbers on thin evidence shouldn't look the same as one we can back up, so every deal page shows three things side by side rather than one combined figure:
- Attractiveness: the match score above, and the best strategy's verdict.
- Data confidence: Good needs a Good-confidence valuation, no independent valuation more than 10% away, and at most one score signal missing. Low is a Low-confidence or missing valuation, or an independent valuation more than 10% away (which also lowers our valuation's own confidence one level). Anything else is Fair.
- Risk: High with any serious (red) warning, Check first with amber ones only. With none, None found only when the planning and flood checks have run (England and Wales); otherwise Not fully checked, as for a deal checked only for flood zones while the government's planning data service is down. Green belt and Article 4 areas aren't checked in Wales (there's no national map of them), and a Welsh deal's None found says so.
Changelog
Every change to the rules that rate deals, newest first. Dates are when each change was made to our code, in UK time. Deals on our track record keep the verdict and score they had when we featured them, and say which version they were rated under.
Version 23: “Below our estimate”: one test for the label and the match score; one verdict per strategy everywhere
CurrentWhat changed
- The match score’s below-value signal is now “Below our estimate”: how far the asking price is under our estimate less its typical error (7.0% at Good confidence, 10.9% at Fair, 18.3% at Low). A deal that clears it says “X% below our estimate”, even when the price is inside our likely range. The likely range is still the model’s 80% range, and “above our range” is unchanged. The signal’s weight is unchanged.
- A verdict lowered by a check (an unconfirmed value far above the price, among others) says why first, and never says it meets a higher threshold. The deal page’s figures use exactly what the feed, the score and the PDF use.
Why
The label measured “below our estimate” against the estimate less its typical error while the score measured it against the low end of the 80% range, so a deal could say it was 14% below our estimate and score nothing for it.
Version 22: Buy-to-let on a home that needs works: the works and the rent lost until it’s let are in the figures
What changed
- Buy-to-let on a home that needs works before letting is no longer capped at weak: the works estimate and the rent lost until it’s let (by default 1, 3 or 6 months for light, medium or heavy works) are in the cash put in, and it’s judged on those figures.
- An auction lot that needs works is worked out as bought with cash or bridging, the works done, then refinanced on its value after works. A flat under 30 m² that needs works isn’t mortgageable for buy-to-let (at most weak), unless you’ve checked a lender will lend on it.
Why
The cap left the works out of the figures: some homes needing £100,000 of works showed a weak buy-to-let that the numbers couldn’t support, and some good ones were held at weak whatever their figures.
Version 21: Refurb costs: lower default rates for everyone, higher in London and the South East, better floor areas
What changed
- Every account’s default refurb costs are £200, £450 and £900 per m² for light, medium and heavy works (new accounts have had these since 12 Nov 2026; the old default was £250, £650 and £1,100). Rates you set yourself are kept.
- Works cost more where building does: ×1.20 in London, ×1.08 in the South East and ×1.03 in the East of England, on the area works and the items outside the scope. Other regions use UK rates until their build costs are verified.
- A home with no floor area of its own has its works sized from a typical floor area for similar homes on the same street (from energy certificates), before falling back to a typical size for the bedrooms; works sized that way say “estimated from a typical size”.
- A house whose advert gives a floor area under 70% of the national minimum space standard for its bedrooms (1 bed 25 m², 2 beds 42, 3 beds 51, 4 beds 63) has that figure left out of the works estimate. Flats keep the 15 m² a bedroom check, as small conversions are real.
Why
Medium works on a 90 m² house came to £64,350 on the old default (90 m² × £650 = £58,500, plus 10% contingency; a kitchen and bathroom are part of medium works), against typical quotes of £30,000–45,000, so flips and BRRRs failed on the works alone. At £450 a m² they come to £44,550. A UK rate is too low for London and the South East. A third of works estimates rested on a typical size for the bedrooms, or on an advert figure too small to be right.
Version 20: The lower of our value and the independent one
What changed
- Where an independent automated valuation (last 60 days) is below ours, the flip and BRRR figures, the below-value wording and the match score’s below-value part use it, and the value after works is scaled down by the same ratio. Deal pages, PDFs and cards show both values.
Why
On the listings we feature the independent valuation was a median 9.3% below ours: our value is probably about 10% high on deals that look cheapest, so the figures take the more cautious of the two.
Version 19: Buy-to-let: no mortgage, or an area-wide rent, means at most weak
What changed
- Buy-to-let on a home no lender would lend on (cash buyers only, a short lease, non-standard construction, or priced below the usual lender minimum) is at most weak: its figures assume cash, with no mortgage payment, so they aren’t comparable. An auction lot isn’t included: it can be refinanced.
- Buy-to-let ok or strong needs a rent from comparable lets: on a low-confidence rent (an area-wide average, or a regional default for the bedrooms) it is at most weak. A rent you enter yourself isn’t capped.
Why
Flats sold to cash buyers only showed £1,147 a month of “cashflow” on an area-average rent, with no mortgage taken off; buy-to-let verdicts rested on rents that weren’t from comparable lets.
Version 18: Discounts explained by a short lease, a repossession or condition; studio and flat values corrected
What changed
- A listing on the short lease, repossessed or unmodernised lists (or whose advert shows a short lease, a repossession or heavy works) gets nothing for being below value in the match score, and says “Priced below our value: likely reflects a short lease. Check before relying on it.” instead of how far below.
- Studio values made without the address are corrected for the model’s measured overvaluation (×0.771 from the street alone, ×0.900 with a floor area), with wider ranges; other flats’ ranges are widened slightly (×1.03 and ×1.05) so 80% of sale prices fall inside them.
Why
Flats on those lists are listed a median 24% to 28% below our value, against 3.6% for flats that are simply slow to sell: the gap reflects what the model can’t see, not a bargain. In testing, studios valued from the street alone came out a median 28% above their sale price.
Version 17: No below-value boost for a suspect value; long-listed homes and studios
What changed
- A deal flagged “Unusually far below our estimate” (our value more than 30% above the price, not confirmed by an independent valuation), or valued at Low confidence, gets nothing for being below value in the match score. The signal counts as zero rather than being left out, so such a deal ranks below an otherwise similar one without the flag.
- Studios: our valuation’s confidence is at most Low until the model learns them (it doesn’t know a studio from a one-bedroom flat), so they get nothing for being below value, and the deal says only whether the price is below, within or above our value range, not by how much.
- A home advertised over 6 months shows an “On the market N months” flag. Over a year, 5 points come off its match score. Over two years it’s left out of the deal feed by default (a filter shows it) and out of alerts.
- The same home advertised more than once (same postcode, bedrooms and type, prices within 5%) shows once, as its newest advert.
Why
A Leeds LS6 studio asking £60,500 against our £156,000 value ranked among the top matches on its “discount”. Studios and similar flats are often priced far below what sales nearby suggest for reasons the data can’t see, and a gap that size is more often a problem with the home than a bargain.
Version 16: Floor areas too small for the bedrooms aren’t used
What changed
- A floor area under 15 m² per bedroom (from the advert or the energy certificate) is treated as wrong: the value comes from the street and its sales instead of the floor area, the refurb cost from a typical size for the bedrooms, and the deal shows a “Floor area uncertain” flag saying so.
- Such a flat isn’t marked “Small flat” (which makes buy-to-let cash only) on that figure.
Why
A 4-bedroom terrace in Bexhill was listed at 44.9 m², about half its real size. Valued on that area it came out at £153,000 against £210,000 to £249,000 sales on its own street, so its £235,000 asking price looked 17% above our range.
Version 15: Buy-to-let and HMO priced above our likely range: at most weak
What changed
- A buy-to-let or HMO whose asking price is above the top of our likely range for its value as it is (“Above our range”) is rated at most weak, never ok or strong, with the reason first: “Priced 17% above our likely range: you’d be overpaying, and a lender may value it lower.” Deal cards, deal pages, deal packs (PDF) and CSV exports all show it.
- Their maximum offer stops at the top of that range, and “works at offers below …” gives the price at which they would work with that cap.
- Flip and BRRR are unchanged: their figures already set the price against the value after works, so overpaying shows in the profit and the money left in.
- Free plan previews never pick a deal priced above our likely range.
Why
A Bexhill terrace asking £235,000, 17% above our likely range, rated ok as a buy-to-let because the rent covers the mortgage at that price, and was a free preview’s top pick. Buy-to-let and HMO figures take the price as given, so a decent yield on an inflated price read as a good buy.
Version 14: “Below our estimate” only below our likely range
What changed
- A deal says it’s below our estimate only when the asking price is below the low end of our likely range for the value, and says how far below the range it is: a £160,000 price on our £221,000 estimate, whose likely range starts at £185,000, says “28% below our estimate (14% below our likely range)”. A price inside the range says “Within our likely range”, with no below-value claim.
- The score’s below-value signal counts only how far the price is below the low end of that range. It used to count any price below our estimate less its typical error, which included some prices inside the range.
Why
A Brighton flat was shown as 14% below our estimate (our estimate was £221,000) though its £190,000 asking price was inside our own likely range of £185,000 to £230,000. By our model’s standards that isn’t clearly below value.
Version 13: Flips that fail even at the top of our value range
What changed
- A flip that loses money even at the top of our range for the value after works says it fails even there, with that best case, instead of leading with the bigger loss at the low end of the range.
- Those flips are no longer flagged “Figures look wrong: check the inputs”: the verdict holds whatever the value turns out to be.
- Deal packs (PDF) and exports show the same figures and wording as the deal cards.
Why
A £1.15m house showed an £821,200 flip loss worked out on the low end of a Low-confidence range. That loss rests on the most uncertain figure we have; that the flip fails even at the top of the range does not.
Version 12: “Too uncertain to value” for extreme cases only
What changed
- Flip and BRRR figures are withheld only when the range of our value after works is more than 1.6 times as wide at the top as at the bottom, or the valuation is Low confidence and the as-is floor had to lift the low end by more than 15% of the as-is value. This replaces the withholding rules of versions 8, 10 and 11, and is the rule today.
- Other Low-confidence values and ranges over 1.3 times show their figures, still worked out on the low end, with a “Wide range: check carefully” caution.
Why
Version 11 withheld figures at any Low-confidence valuation, over half of listings including ordinary ranges, and nearly every working flip disappeared. Touching the as-is floor is normal (the typical uplift from works is smaller than the valuation’s error band), so only a large lift counts.
Version 11: Values after works never below the as-is value; too uncertain to value
What changed
- The low end of the value after works is never below the home’s value as it is: works don’t make a home worth less.
- When our value after works is Low confidence, or its range is more than 1.3 times as wide at the top as at the bottom, flip and BRRR figures are withheld and the deal says it’s too uncertain to value reliably. Buy-to-let and HMO, which rest on the rent and the price, still show. (Replaced 12 minutes later by version 12, which withholds figures only in extreme cases.)
- On a 5-year fix, a deal card says when the loan was kept below what a lender would lend to keep cashflow at your minimum, which is why its interest cover is above your required ICR.
Why
The range is our estimate plus or minus its typical error, which is wider than the typical uplift from works, so on most listings its low end sat below the as-is value: a £1.15m house showed an £852,700 flip loss. At Low confidence that range is too wide to support a profit or loss figure.
Version 10: 5-year fixes, a loan the rent can carry, and realistic offers
What changed
- A mortgage product assumption: a 5-year fix is stress-tested at the pay rate (never below the floor), a 2-year fix at the pay rate plus the add-on. New accounts start on a 5-year fix; existing accounts keep the 2-year test until they switch.
- A BRRR refinance borrows no more than keeps the rent covering the mortgage and running costs, even when a lender would lend more.
- BRRR figures are withheld, like flip figures, when our value is Low confidence and more than 30% below the price. (Replaced in version 12.)
- An offer more than 50% below the asking price is not shown as one that works: the deal says it doesn’t work at any realistic price, and why.
Why
Most BRRR refinances are onto 5-year fixes, which lenders test at the pay rate, so the 2-year test was capping loans that lenders would make. At the pay rate the largest loan can leave cashflow negative, which no investor would draw. A value far below the price at low confidence made BRRR figures on expensive homes misleading.
Version 9: Value after works as a range, and offers that would work
What changed
- The value after works (GDV) is shown as a likely range: our estimate plus or minus its typical error at its confidence level (7.0% at Good, 10.9% at Fair, 18.3% at Low for our valuation model).
- Flip and BRRR profit, verdicts and maximum offers are worked out on the low end of that range; the figures at our central estimate are shown alongside.
- A strategy that doesn’t work at the asking price says what offer would make it work, checked against the full verdict, not just the target.
- New accounts start from updated default assumptions: a 15% flip profit target, a strong flip needing at least £15,000 profit (not £20,000), refurb costs of £200, £450 and £900 per m², and a BRRR stress test at the pay rate (a 5-year fix). Existing accounts keep theirs. (The refurb costs became the default for every account in version 21.)
Why
Half of homes sell more than about 9% away from our estimate, and on a flip or BRRR an error that size can wipe out the profit, so a single figure gave false confidence. The old new-account defaults were stricter than typical UK practice.
Version 8: Buy-to-let floor and flip sanity checks
What changed
- Buy-to-let fails below a 3% return on the cash put in (it was weak); 3% to 5% is weak.
- A flip loss of more than 60% of the value after works is flagged “Figures look wrong: check the inputs”.
- When our value is Low confidence and more than 30% below the asking price, flip figures are withheld, and a withheld flip is never counted as strong. (This withholding rule was replaced in version 12: see the current rules above.)
Why
A buy-to-let returning under 3% on the cash put in now fails instead of showing as weak. A bungalow valued well below its asking price at very low confidence showed a six-figure flip loss; the maths was right but the inputs weren’t, so figures that depend on a very uncertain value are now withheld.
Version 7: Score caps and buy-to-let lending
What changed
- The score is held at 40 when none of the strategies you want has numbers that work.
- Renovation need counts towards the score only when a search wants a flip or BRRR.
- A Low-confidence valuation no longer cuts buy-to-let or HMO verdicts, which rest on the rent and the price. Flip and BRRR keep the cut, and buy-to-let and HMO deals say a lender’s valuation could come in below the price.
- Flats and maisonettes under 30 m², and studios, carry a lending caution; under 30 m², buy-to-let runs as a cash purchase unless you’ve checked a lender will lend.
- Buy-to-let uses a mortgage on the lower of the price and our estimate when ours is more than 10% below the price. A home needing heavy works is capped at weak to let as it is.
- Auction lots with a guide under half our estimate are never rated strong.
- Your deal-page choices (include the refurb in buy-to-let, a lender will lend) apply to your deal list.
Why
A deal whose numbers don’t work shouldn’t outrank working deals on its other signals, and renovation need is only a plus if you’re doing the works. The buy-to-let changes bring the figures closer to what a lender would actually offer.
Version 6: A plausibility check
What changed
- If our value is more than 30% above the asking price, the deal can’t be strong, featured or sent as a same-day alert unless the independent valuation agrees within 10%. It is flagged “Unusually far below our estimate”.
- Land, plots, garages, farm buildings, commercial and mixed-use property are detected more widely, and get no automated value or below-value figure.
Why
Second opinions put the independent valuation a median 9.8% below ours at average condition, and further below on the listings furthest under our value, so the biggest gaps needed confirming.
Version 5: An independent second opinion
What changed
- Shortlisted deals get an independent automated valuation. If it differs from ours by more than 10%, the deal gets an amber flag, our confidence drops a level and it is shown as ok at most.
- Same-day alerts apply the same valuation rule as strong verdicts.
Why
A second, independent valuation catches cases where ours is wrong before a deal is called strong or sent as an alert.
Version 4: Our valuation model, and what “strong” needs
What changed
- Values come from our valuation model, trained on HM Land Registry sales and energy certificates, with confidence set by the width of its 80% range.
- A strong verdict needs a Good-confidence valuation, or else an asking price below our estimate by at least the Fair band’s typical error plus 5 percentage points. Otherwise it is shown as ok.
- An asking price a third or more below our estimate is never rated strong or counted as below value.
Why
Flip and BRRR verdicts rest on our valuation, so “strong” should only appear when the valuation is reliable enough to support it. A gap of a third or more is more often something the data can’t see (tenure, condition, a plot) than a bargain.
Version 3: Buy-to-let, HMO and lending rules
What changed
- BRRR: the refinance loan is capped at what the rent covers at the lender’s stressed rate, and a BRRR fails below a 5% return on the capital left in.
- Buy-to-let and HMO verdicts, which count towards the score when a saved search wants them (HMO only where local room rents are available).
- A minimum property value for a mortgage (£60,000 by default): below it, buy-to-let and HMO run as cash purchases and a BRRR refinance fails.
- Problem flags from the listing (shared ownership, retirement, cash buyers only, short lease, non-standard construction and others), with red flags kept out of alerts unless a search includes them.
- Commercial and mixed-use property is flagged and not valued automatically: you can still run figures on your own numbers.
Why
A review found the BRRR figures assumed a full loan the rent might not support. Homes below lenders’ minimum value can’t usually be mortgaged, so their figures had assumed finance that isn’t available.
Version 2: Valuations measured against real sales
What changed
- Values are shown as a likely range, and “below our estimate” is measured against the low end of that range.
- Valuations blend in the home’s own last sale and earlier sales in the same postcode, moved by the local price trend.
- Fixes for implausible valuations: certificates matched to the wrong home on street-only addresses, and small flats valued like the large flats in their postcode.
Why
Tested against 1,500 HM Land Registry sales, valuations from nearby sales alone were typically about 15% out and made cheap value-add homes look like bargains. Measuring against the low end of the range cut normally priced homes shown as 15% or more below value from 18% to 8%. Before the fixes, 88 of 453 listings were valued at more than 1.5 times their asking price.
Version 1: The first deal score
What changed
- A 0–100 score blending the best wanted strategy’s verdict (weight 40), how far the asking price is below our estimate (15, full marks at 20% below), the AI’s kitchen-diner, extra-bedroom and renovation-need scores (10 each) and our confidence in the figures (15). Signals without data are left out and the rest re-weighted.
- Flip and BRRR verdicts against your own targets. Below a valuation confidence of 0.4, the verdict counts for less in the score.
- Auction lots are analysed at the guide price plus 20%, and at no less than 80% of our current value.
- Retirement, shared-ownership and park-home listings are shown but not analysed or scored.
Why
The first version. The auction rules came the same afternoon, after the first real listings ranked an auction lot’s low guide price as the top deal.
All figures are estimates, not valuation, tax or financial advice. Do your own due diligence and get a survey before buying.