How impact figures are calculated

Backed shows employers and funders what work has been created and what it is indicatively worth. This page says exactly how — and where the numbers stop.

What is counted, and where it comes from

Sources of each figure
FigureSourceNotes
People hiredPlacements recorded by the support worker or employer on the networkOne per person per role. A person hired twice counts twice.
Still in workRetention check-ins at 30, 60, 90 and 180 days“Still in work” means no check-in has been answered “not retained”. Unanswered check-ins are not counted as retained.
Hours of paid workRecorded hours per week × weeks in workWeeks in work run from the start date to today, or to the first check-in marked “not retained”, or to the recorded end date.
Wages earnedHours of paid work × recorded hourly wageAn estimate. Where no wage was recorded the London Living Wage default of £13.85 an hour is used.
Training and qualifications recordedExperience lines of kind training or certification on people supportedRecorded by the support worker or issued as a verified credential by a training programme on the network.
People whose profile has since been deletedAn anonymous outcome snapshot written at deletionDates, hours, wage and retention only — no name and no link to the person. Their outcomes still count; they cannot be identified.

Indicative social value

The social-value figure is indicative. It is calculated by multiplying full-time-equivalent years in work by two per-person-year proxies:

Proxies used
ProxyValue per full-time person-yearIn the spirit of
Wellbeing uplift of moving from unemployment into steady work£14,000HACT Social Value Bank wellbeing valuations
Savings to public services (out-of-work benefits, service use)£10,700Unit-cost databases for exchequer savings (for example the Greater Manchester Combined Authority Unit Cost Database)

These proxies are placeholders and unlicensed. They are in the range published by recognised UK frameworks but are not taken from a licensed dataset. Before any figure from Backed is used in external social-value reporting — a bid, an annual report, a Social Value Act submission — the organisation reporting must adopt licensed framework values and record which ones. The shape of the calculation stays the same; only the constants change.

Part-time hours count pro-rata against a 37.5-hour week. Wage arithmetic uses 4.33 weeks per month.

What is deliberately not applied

  • Deadweight — the share of outcomes that would have happened anyway — is not deducted. The figures are gross, not additional.
  • Attribution — how much of an outcome is down to Backed rather than the partner organisation, the employer or the person themselves — is not applied. The figures describe work created through the network, not work caused by it.
  • Drop-off and displacement are not modelled.
  • Discounting of future value is not applied, because no future value is counted: only time already worked.

A funder or employer applying a recognised framework (for example Social Value UK’s principles) should add their own deadweight, attribution and drop-off assumptions on top of the gross figures shown here.

How time in work is capped

Value accrues only for time a person has actually been in work. Time in work runs from the placement start date until the earliest of: today; the recorded end date of the placement; or the day of the first retention check-in answered “not retained”. Time is never projected forward. Prospective estimates (“what would offering this role be worth?”) are shown separately, labelled as estimates, and never added to the retrospective totals.

The same value for every person

The per-person-year proxies are identical for everyone. Backed never prices a person differently by their background, age, disability or any other characteristic — that would be undignified and contrary to the network’s positive-action rules. What varies is time in work, hours and wage.

Where AI is and is not involved

Every number is calculated deterministically in code and can be reproduced. Where an organisation has switched it on, an AI model drafts the prose around the numbers from the calculated report text; it is instructed to use only the figures given and never to invent, extrapolate or round. The built-in wording is used when no AI is configured, and the figures are identical either way.

Related: terms of use · privacy notice.