Methodology
How we compile the Yellowhite private plate dataset: where the numbers come from, how they are calculated, and where the limits are.
What the data is
The dataset draws on two DVLA sources. The first is DVLA's fixed-price stock, monitored through its published listings; when a registration is withdrawn from sale we treat that as a sale. The second is DVLA's timed online auction results, which DVLA publishes after each auction closes. Fixed-price monitoring runs from January 2024 and auction results from January 2025, both to the present.
In total that is 1,246,290 sales, including 27,656 auction lots across 14 auctions. The figures on this page and across the statistics pages were last rebuilt on 27 July 2026.
The two markets, and why they are never blended
There are two very different markets, and we never combine them into a single average. The fixed-price list is administered by DVLA in set tiers. It reflects DVLA's pricing policy, not what a plate would fetch in open bidding, so it is only ever described here as what DVLA charges.
The auction price is set by bidders competing for a lot, so it reflects demand. Every claim on these pages about what a plate is worth comes from auction data. Any figure taken from the fixed-price list is described only as DVLA's price, never as a market value. Keeping the two apart is the single most important rule in how we report these numbers.
How likelihood is calculated
The likelihood figures, the ones that say a given car is several times more likely than average to wear a private plate, divide a make or model's share of plate purchases by its share of the UK vehicle parc. The vehicle parc comes from the Department for Transport's vehicle licensing statistics.
The division is done on raw vehicle counts, never on percentages that have already been rounded. This matters: an earlier version of this analysis, published in 2025, calculated the multipliers from rounded parc shares, which understated several smaller manufacturers and dropped some entirely. Working from the raw counts is the correction, and it is what makes the current figures trustworthy. To keep small samples from producing silly numbers, any make or model with fewer than 50 plates or fewer than 5,000 licensed vehicles is excluded as too small to measure.
Vehicle matching, and the honest limit on the never-assigned figure
Every plate in the dataset is checked against official vehicle enquiry and MOT records to find the car it was assigned to, if any. Across all plates that should be identifiable, that matching succeeds about 72.7% of the time. Of the 1,246,290 plates we track, 857,066 are matched to a vehicle and 389,224, or 31.2%, are not.
This is the figure most open to misreading, so we are explicit about it. That 31.2% never-assigned share combines two different things: plates genuinely held as assets, on retention certificates, that are deliberately never put on a car; and plates we simply could not match through the vehicle enquiry and MOT lookups. Because identification is not perfect, part of the 31.2% is match failure rather than a decision never to fit the plate. The true share held purely as assets is therefore somewhat below the headline number, which is why we describe it as "up to a third" and never as a flat statement that a third of plates are held as assets.
Matching is redone on a rolling basis, so a plate recorded as not assigned can move onto a vehicle in a later run, and the figures shift a little each night as that happens.
How auction dates are handled
A timed online auction runs over several days as its lots close, so a single sale date is not really "the auction". We group auction lots by the calendar month in which they closed, taken from the sale date of each lot. We do this rather than trust the file names or sheet names, because DVLA has at times republished older result files with adjusted dates. Grouping by month from the sale date keeps every monthly total and ranking correct regardless of how a file was labelled.
What is excluded
A few things are deliberately left out. A plate that is relisted, for example withdrawn and later offered again, is counted once, not twice. Makes and models below the minimum thresholds above are excluded from the likelihood figures. Plates we could not match to a vehicle are excluded from the likelihood calculation, but they are still counted in the sales and value totals; they are only ever dropped from a figure where an unmatched plate would distort it, never from the headline counts.
How often it updates
The dataset is rebuilt in full every night, and again after each DVLA auction closes and its results are loaded. The last rebuild was 27 July 2026. Each of the statistics pages carries its own last-updated stamp so you can see how current the figures on it are.
Using and citing the data
The figures across these pages are free to cite in reporting or research, with a link back to the page the figure came from. If you need a breakdown we have not published, a longer time series or higher-resolution figures, get in touch or email [email protected], and we will usually be able to help. A short description you are welcome to use: Yellowhite tracks every private registration DVLA sells, matched to DVLA vehicle records, and updates the data nightly.