How prices are compared

Normalisation

Each listing’s price is recorded in the seller’s currency and converted to USD minor units at the day’s rate. Only the USD figure is compared across sellers. Prices published on this site are what a customer pays through the API, which includes our margin; wholesale figures are not published because the two side by side would disclose the margin by subtraction.

Price gap

A domain is contested when more than one marketplace sells the same placement type on it. The gap is the most expensive ask over the cheapest, as a percentage of the cheapest; the figure quoted is the median across contested domains, because the mean is dragged past 100% by a tail that reaches ten times. Today 86,282 domains are contested and the median gap is 98%. Comparing a homepage link with a niche edit on the same site would produce a far larger number that means nothing, so it is excluded.

Price bands and plausibility

Listings are grouped by placement type, inventory mode and the seller’s claimed-DR decade. For each band with at least 30 domains we hold the full distribution of asks at five-percent steps. An offer’s plausibility is where its price sits in its band: under the 10th percentile is flagged low: real authority is not sold at that price, so the claimed number is more likely bought than the link is a bargain. Bands under 30 domains are not published; an offer in one gets no verdict rather than a weak one.

Market reference price

The median ask of the offer’s band, with the offer’s own price against it and the spread of asks for that exact domain across sellers. It is called a reference, not a fair value: we hold asking prices, not clearing prices. Within ±15% of the reference is “at”; a band is a distribution, not a price tag. A subdomain is compared with the unrated band because its listed DR is its parent’s. Where plausibility reads low, the reference is not spoken: “95% under the market” and “priced like bought authority” are the same fact, and the second framing is the true one.

The Backlink Price Index

A monthly series, published on the 1st and never revised. The basket is every placement type × claimed-DR decade cell with at least 100 non-suspect root domains in the base month (September 2026), weighted by that month’s domain counts. Each month the basket is re-priced with the new cell medians (cheapest ask per domain, buyer price) and divided by the base month’s value; base = 100. Fixed weights mean the index moves when prices move, not when the mix of listed inventory does. Beside the composite each snapshot carries medians and quartiles by type, DR band, measured-traffic band and subject, the seller-gap dispersion and the lowest-decile-to-mean ratio, and a CSV. Cells that fall under the floor in a later month are re-priced at their last published median and flagged; a cell absent from the base is never added mid-series (a new series starts instead). Index at /research/backlink-price-index.

Price history

Every observed ask is appended to a history that is never overwritten (from 28 August 2026). A listing’s change since first seen is reported once it has at least 14 days of observations; under that, a change describes our crawl, not the market.

What these numbers are not

Not a valuation

Nothing here says what a link is worth to a buyer. It says what sellers ask and how those asks compare.

Not summable

The gap and the reference are medians of distributions. Adding them across a plan as “savings” overstates; the API reports savings separately as the difference against the median seller of each slot actually bought.

Not a supplier comparison

Suppliers are never named. Where their asks differ for one domain, the buyer is shown the spread and the count, and buys the cheapest.