Three Comparables or Thirty: How Much Weight Should You Give a Valuation?

Follow the sample through the filters

CarValue.ie records the scanned pool, the group after vehicle filters and the final pricing cohort. The last of these is the set behind the price bands. A headline saying “100 listings searched” does not mean all 100 are close matches for your car.

The mileage step may reduce a large search to a handful of cars. Engine variants, fuel subtypes and premium trims can reduce it further. That is often appropriate: more data is not better if it describes the wrong vehicle.

A narrow range does not guarantee strong evidence

Imagine three adverts priced at €19,950, €20,000 and €20,100. Their spread is small. They might still all be dealer stock of one specification, or all share an attribute that your car does not. Price consistency is one piece of evidence, not independent confirmation.

With a small group, removing a single unusual car can move the quartiles significantly. Look at the individual listings and ask whether the advertised condition, gearbox, trim and mileage justify including each one.

Read the confidence label as a heuristic

The current engine derives its confidence label from sample size, price dispersion and search expansion. It is not a probability that your car will sell for the middle number. In particular, the existing heuristic can label a group of five to nine tightly priced cars “high”.

For our September editorial research, we flag pricing cohorts below ten as small samples even if the product’s heuristic label is higher. This makes the limitation explicit while keeping the underlying production price calculation unchanged.

What stronger evidence would look like

A useful evaluation freezes the input data and compares old and proposed engine behavior on the same cases. It should record errors by model, mileage, fuel type and sample size rather than averaging away the difficult cases.

Comparing predictions with held-out asking prices tests agreement with adverts. It does not prove accuracy against final sales. To make that stronger claim, we would need reliable completed-sale evidence and a way to match each transaction to its condition and specification.

How to act on a thin result

Use a wider decision range, inspect the matching adverts and seek an actual purchase offer if you need a firm number. Recheck when more relevant stock appears. Do not turn a small dataset into apparent certainty by reporting an extra decimal place.

• Check the final pricing-cohort count, not just the search total.
• Look for mismatched trims, seller types and mileage.
• Treat an unusual top or bottom advert as a question to investigate.
• Keep inspection findings separate from statistical pricing.