Value Score: How We Tell Dublin Renters Whether a Listing Is a Rip-Off
In Dublin, a 2-bed apartment in Ranelagh costs 3,200 EUR per month. Is that a deal, market rate, or a rip-off? The honest answer is: it depends on five other things you can't see from the listing photo. Distance to a DART station. Number of identical units that rented for less last quarter. Whether the BER rating is C or G. Whether the "second bedroom" is actually a converted closet.
Daft.ie shows you the price. It does not tell you whether the price is fair. So we built Value Score, a single 0-100 number on every HomeScout listing that answers "is this overpriced?" so renters don't have to do the spreadsheet themselves.
This is how we calculate it, what's in the model, and the things we deliberately left out.
The Renter's Real Question
When you walk into a Dublin viewing, you have about three minutes to decide whether to apply. The decision is not "do I like this place", you've already decided that from the photos. The decision is "is the price reasonable for what I'm getting."
The information you'd need to answer that, off the top of your head:
- What did similar units in this area rent for, in the last 90 days?
- Is this priced above or below the median for its size + location?
- How does the per-bed price compare to the city median?
- Are there hidden cost differentials (no parking, BER G, no heating)?
- Has this exact unit been listed before at a different price?
Nobody has time to look this up at a viewing. So our job was to compress all of it into one glanceable number.
What Goes Into the Score
Value Score is a weighted blend of seven inputs. Each input is normalised to a 0-100 sub-score. The weights below are the current production values; they shift slightly as we tune.
| Input | Weight | What it measures |
|---|---|---|
| Price vs. comparable median | 30% | Listing price compared to the median of the last 90 days of similar units (same area, ±1 bed, ±15 sqm) |
| Price vs. area median | 15% | Same idea, but coarser, by Dublin postcode area |
| Price per bedroom | 10% | Sanity check against the city-wide median bedroom price |
| BER rating | 15% | A-rated units score full; D and below get penalised because heating costs are real money |
| Transport access | 10% | Walking distance to nearest DART, Luas, or 24-hour bus stop |
| Listing freshness | 10% | A listing that has been re-listed three times at increasing prices is a yellow flag |
| Amenity completeness | 10% | Parking, washing machine, dishwasher, balcony, each present amenity nudges the score up |
A listing in Ranelagh at 3,200 EUR with a B BER, 4-minute walk to the Luas, full amenities, priced at the area median scores around 78. The same flat at 3,800 EUR with a D BER, 12-minute walk to anything, no parking, that's a 41.
We deliberately keep the weights public-facing-ish (the table above is on our docs page) because renters who understand why a number is what it is trust it more than a black box.
Where the Comparable Data Comes From
The hard part of Value Score is not the math. It's the data.
We ingest every listing that appears on Daft.ie's Dublin pages, daily, into a Postgres time-series. Each listing keeps a history: first seen, price changes, delisted-and-relisted events, and final disappearance from the market (which we treat as a proxy for "rented"). We don't see actual signed-rent prices, Ireland doesn't publish those, but the asking price at delisting is the closest available signal.
For each new listing, we query the comparable cohort: same Dublin postcode area, ±1 bedroom, ±15 sqm, last 90 days, delisted (not still active). We compute the median, the 25th, and 75th percentile. The listing's price-vs-median score is just where it falls on that distribution, normalised.
Some areas don't have enough comparables, niche size brackets, fringe postcodes. We backfill with a Bayesian prior that pulls from the next-broader cohort and flag the score with a "low confidence" indicator. Better to admit uncertainty than to fake precision.
What We Left Out (And Why)
Three things you'd expect in a rental score that we deliberately omitted:
Reviews. No "users rated this neighbourhood 4.2 stars." Those scores are dominated by a small number of vocal users and they teach renters nothing. The choropleth map and the price data already convey what's good about a neighbourhood; we don't need to add a 5-star average.
Landlord reputation. Ireland doesn't have public landlord ratings, and the half-built reputation systems we looked at had a strong "biased toward complainers" effect. We may revisit if we ever build a verified-tenant feedback loop, but the data quality bar is high.
Predicted rent growth. "This area is hot, score adds 5 points" is a trap. It pushes renters toward overheated markets and rewards landlords who price aggressively. The score is about value now, not speculation about value tomorrow.
How It Surfaces in the Product
The score appears in three places:
- Search result cards. Every listing card shows the score as a coloured pill, green (75+), amber (50-75), red (<50). Renters scrolling 30 listings can spot the deals without reading every description.
- Property detail page. The full breakdown, each input, its sub-score, the comparable cohort size, the confidence indicator. For renters who want to understand why a place got a 47.
- Saved properties dashboard. When the score changes (price drop, new comparable data), the saved entry updates and you get an alert. A unit that drops from 52 to 71 because the landlord reduced the asking price is a useful signal.
We do not let the score ever read above 100 or below 0. A "deal" cannot be infinitely good, and a "rip-off" cannot be infinitely bad. We've found these edge cases produce skepticism, not engagement.
What We Got Wrong On The First Try
Our first version of Value Score weighted price-vs-median at 60%. That sounds right, "the renter just wants to know if it's overpriced", but it produced a brutal failure mode: every place in Dublin 1 got a 90 because Dublin 1 is cheap relative to the city, even when individual flats were dumps. The score collapsed neighbourhood quality into the price comparison.
The fix was the BER + transport + amenity inputs. They're effectively a "is this place worth living in at all" sanity layer that the price comparison doesn't capture. Cheap and miserable scores around 50, not 90. Expensive and excellent scores around 80, not 60.
We re-tuned weights three more times after that based on user click-through. The current values land where renters click on green-pill listings 4x more than red-pill, and where saved-listing rates correlate with score in a way that suggests renters trust it. Empirically, that's our success metric.
Try It
Value Score is on every listing on HomeScout, free. No signup required to see it. Run a search at homescout.io/search, type something natural like "2 bed near Trinity under 2500", and the score will be there, on every card, before you click anything.
If you're a renter, the practical advice is: filter by green pills first, save 5-10, then read the detail page on each. You'll find better places faster than scrolling Daft.
If you're an engineer thinking about building a similar score for a different market, I'm happy to talk through the data pipeline, email's in the site footer.
, Caspar
Top comments (0)