How much is fresh data worth?

In the B2B data market, a contact sells for the same price whether it was verified last month or eighteen months ago. This is an anomaly. The value of data is not a property of the data: it is a function of its age. This article proposes a simple model to calculate it — and draws a direct consequence from it on the price that it is rational to accept.

The expected value, not the nominal value

Data only has value if it is still true. The correct unit of account is therefore the expected value: value if correct × probability that it is still true, minus cost if false × inverse probability. This second half is almost always forgotten, even though it changes everything: false data is not worth zero, it is worth less than zero. An invalid email degrades the deliverability of your domain; an obsolete post costs a salesperson twenty minutes and a little credibility for your brand.

The depreciation curve

The probability of validity decreases approximately exponentially. Let's take a professional email, the validity of which is linked to the position: with a rotation of around 20% per year on qualified functions – an order of magnitude commonly observed, higher in tech – the degradation approaches 1.5 to 2% per month. Concretely:

• Verified 1 month ago: around 97-98% chance of still being valid.
• 6 months ago: around 89%.
• 12 months ago: around 79%.
• 18 months ago: around 70%, and significantly less in mobile populations.

Each field has its own curve: a SIREN number depreciates very slowly, an active job offer expires in a few weeks, a job title falls between the two. Reasoning “by basis” makes no sense; we reason by field.

The calculation that changes the acceptable price

Let's set some illustrative values: a correct contact brings you €5 on average (probability of response × value of the pipeline generated), an incorrect contact costs you €2 (lost time, deliverability, cleaning). The expected value becomes:

• At 1 month: 0.98 × 5 − 0.02 × 2 ≈ 4,86 €
• At 18 months: 0.70 × 5 − 0.30 × 2 ≈ 2,90 €

The same record has lost 40% of its value — and the calculation becomes negative as soon as the probability of validity falls below about 30%: beyond a certain age, paying for anything destroys value. Operational conclusion: accepting the same price for a form of unknown age and for a form verified last month is to accept a blind discount. The age should appear on the label, field by field.

A stock ages one month per month

This model disqualifies a common practice: basic “one-shot” purchasing. A base purchased today with 95% validity automatically drops below 80% in one year without anyone touching anything — its average age increases by one month each month. The only asset that does not depreciate is a flow: a base of which each record is re-verified at an interval less than its half-life.

What we do with it

This is the raison d'être of sourcing.sh's architecture: an index — around 200,000 companies, 123,000 profiles, 1.4 million offers — continuously searched by agents whose job is precisely to keep each field above its value threshold, with a freshness date exposed rather than hidden. You do not pay for retail files: you pay, as a fixed price, to maintain the curve at the right level — delivered to where you work, CRM, ATS, API or Claude via MCP.