Why almost all income data is modelled, not declared
Ask most people their annual income in a survey and they either skip the question, round to a convenient number, or give a figure that excludes bonuses, rental income, or a partner's earnings. Direct income questions have high abandonment rates. Even when answered, the figure rarely matches HMRC records. This is not a new problem: academic survey researchers have documented the "income non-response" issue for decades.
Commercial data compilers have responded by building statistical models that infer probable income from variables the individual is far more willing to share or that can be derived from public sources. The result is an income band, not a salary figure, and it is accurate to a band about 65-75% of the time across a whole file. That is good enough for segmentation purposes but should never be treated as a precise financial fact about a named individual.
The distinction between modelled and self-declared income matters commercially, too. A record flagged as "self-declared income over £75,000" carries far higher confidence for a wealth-management campaign than one where the £75k-plus band was inferred from postcode house prices. Good suppliers are transparent about which is which, and that transparency should appear in the data specification sheet before you commit to a count.
How the modelling works: the four main data sources
Income modelling on UK consumer files draws on four primary inputs, combined at either postcode or unit-postcode level and then overlaid on individual records.
1. Census and ONS earnings data
The Office for National Statistics publishes the Annual Survey of Hours and Earnings (ASHE), which provides median gross annual earnings by geography, industry, and occupation. The Census (2021 in England and Wales) adds household income quintiles at Lower Super Output Area (LSOA) level. Both datasets are public, granular, and updated regularly. The postcode average derived from ASHE and Census data forms the baseline for any income model: a record in KT2 (Kingston upon Thames) starts with a higher prior probability of being in the £40,000-plus band than one in DL3 (Darlington), before any individual-level variables are applied.
2. Property value and tenure
HM Land Registry publishes every residential sale in England and Wales. The sold price, linked back to a full postcode, is one of the strongest individual-level proxies for household income. A property worth £850,000 in Winchester points firmly towards a higher income band; a rented flat in a low-value block points the other way. Tenure itself (owner-occupier versus renter, and type of rental) adds a further dimension. The 2021 Census tenure data, crossed with council tax band data from local authorities, refines the estimate further.
3. Occupational coding
The Standard Occupational Classification 2020 (SOC 2020) groups jobs into major groups with well-established median salary ranges. When a consumer file includes a self-declared occupation (gathered at the point of survey or questionnaire completion), that SOC code adds significant precision to the income estimate. "Senior manager in financial services" in SOC major group 1 is a reliable signal of above-median income. "Elementary process operative" in major group 9 is a reliable signal of the lower bands. The occupation variable is not always present, but when it is, it typically raises model accuracy by 8-12 percentage points in the bands where property value is ambiguous.
4. Self-declared lifestyle and financial indicators
On fully opted-in consumer files, individuals often answer profiling questions beyond income itself: questions about savings products held, holiday spend, planned home improvements, or vehicle ownership. Each answer feeds into the model as an additional proxy. Someone who states they invested in an ISA last year, owns two cars, and took a long-haul holiday in the past 12 months is almost certainly not in the sub-£20,000 band. These indicators do not replace the geographic and property signals but they sharpen the band assignment at the individual level.
Standard income banding on UK consumer files
There is no universal standard, but most UK consumer data suppliers use a 5-to-7-band structure. The table below shows a typical scheme and the approximate distribution across the UK adult population based on ONS earnings data and HMRC income statistics (figures are approximations and vary by source and year).
| Income band | Approximate share of UK adults | Model confidence | Primary proxy variables |
|---|---|---|---|
| Under £15,000 | ~25% | High | Benefits claimant indicators, council tax band A/B, rented social housing tenure |
| £15,000-£24,999 | ~20% | Moderate | SOC groups 4-7, postcode median earnings, private-rental tenure |
| £25,000-£39,999 | ~25% | Moderate | Mid-range postcode earnings, SOC groups 2-4, first-time buyer property values |
| £40,000-£59,999 | ~16% | Moderate-high | SOC groups 1-2, property values £250k-£500k, lifestyle indicators (ISA, annual leave) |
| £60,000-£99,999 | ~9% | High | SOC group 1 (senior), property values £500k+, premium vehicle ownership, private school indicators |
| £100,000 and over | ~5% | Very high | Property value £750k+, multiple properties, directorship data (Companies House cross-reference), high-value lifestyle indicators |
The two middle bands (£25,000-£39,999 and £40,000-£59,999) are the hardest to distinguish because many proxy variables converge at broadly similar levels for both groups. A household in a £280,000 semi-detached in Leicester with a mid-ranking public-sector occupation sits in genuine statistical ambiguity between those two bands. Accept band-level accuracy of around 60-65% for the middle range and plan your creative and offer accordingly.
Affluence indicators: a more stable alternative
Rather than forcing every record into an income band, many data suppliers derive an affluence score, typically a 1-to-10 scale where 10 represents the highest estimated wealth. An affluence score aggregates property value, tenure, occupational grade, postcode earnings, and lifestyle proxies into a single ordinal figure. It makes no claim about the specific salary.
For campaign planning, this often works better. A financial-services client targeting individuals with significant investable assets will find an affluence score of 8-plus a more reliable filter than a modelled income band of £60,000-plus, because the score incorporates accumulated wealth signals (property equity, multiple assets) that income alone does not capture. A retired consultant with a modest pension income but a mortgage-free £600,000 property has a low income estimate and a very high affluence score. Income modelling alone would screen him out of a premium investment-products mailing. Affluence scoring would keep him in.
The two variables complement rather than replace each other. For current-earnings-sensitive offers (salary-linked credit products, for example) income band is the right filter. For wealth-sensitive offers (investment platforms, premium travel, estate planning), affluence score is more informative. Our experience is that campaigns combining both filters consistently outperform those using either in isolation.
Use cases: who buys income-banded consumer data?
Three broad use cases account for most demand.
Financial services targeting
Banks, insurance providers, and wealth managers target income bands to match product to affordability. A premium credit card with a £5,000 minimum limit is wasted on sub-£20,000 incomes; a basic current-account offer is wasted on £100k-plus households who already have multiple banking relationships. Income filtering reduces wastage in both directions. The Financial Conduct Authority (FCA) also expects firms to target products appropriately, so income segmentation has a compliance rationale as well as a commercial one.
Premium product and luxury brand campaigns
A Cheshire-based luxury kitchen manufacturer running a postal campaign does not want to pay for 500,000 records and hope affluent homeowners are in there somewhere. Filtering to households with estimated income over £60,000, property value over £400,000, and an affluence score of 7-plus is standard practice. The count drops to perhaps 60,000-80,000 records in the relevant region, but response rates and average order values both improve significantly.
Charity major-donor prospecting
Major-donor fundraising in the UK charity sector relies heavily on wealth screening: identifying individuals with the capacity to give five-figure or six-figure gifts. Consumer files filtered on income band £100k-plus, combined with high property values, directorship indicators (cross-referenced with Companies House), and philanthropic lifestyle declarations, form the starting point for prospect research. Charities are acutely aware of the data-ethics dimension here, and the Information Commissioner's Office (ICO) has issued specific guidance on the use of third-party data for fundraising targeting. Any charity using external data for wealth screening should have a clear privacy notice statement covering this processing and should complete a DPIA.
For a broader view of the full range of consumer data variables available in the UK, including geography, demographics, household composition, and lifestyle profiling, see our guide to UK consumer data: what is available and how it is sourced.
GDPR considerations when using income data
Income data is not listed as a special category under Article 9 of the UK General Data Protection Regulation (UK GDPR). You can process estimated income under a standard lawful basis. For a consumer marketing file, the correct basis for the original data collection is consent under Article 6(1)(a), combined with Privacy and Electronic Communications Regulations (PECR) consent for electronic marketing channels. SortedIQ's consumer file is fully opt-in under UK GDPR and PECR consent: records are sourced from individuals who have actively agreed to receive marketing from third-party organisations.
GDPR risk: income combined with other variables
Income data on its own sits outside Article 9. The risk area is combination. If a campaign targets individuals with high estimated income and a diagnosed health condition (for example, critical-illness insurance prospects), or high income and a political party membership indicator, the combined dataset edges towards processing that engages special-category protections indirectly. The ICO expects a DPIA under Article 35 wherever profiling is systematic and involves sensitive inferences, even if no single variable is formally special-category. Buyers doing this kind of multi-variable targeting should seek legal advice and document their assessment before purchase.
A further consideration is Article 22 of UK GDPR, which restricts fully automated decision-making that produces significant legal or similarly significant effects on individuals. Income modelling used purely for marketing segmentation (deciding who receives a leaflet) does not typically meet this threshold. Income modelling used to determine credit eligibility or pricing does, and must be disclosed and challengeable. Know which side of that line your campaign sits on before you start.
What to ask your data supplier before buying
Before purchasing a consumer file with income data appended, you should confirm the following with the supplier in writing:
- Is the income figure modelled or self-declared? If modelled, what are the primary input variables?
- What is the supplier's claimed band-level accuracy across the file, and on what validation methodology is that claim based?
- What is the date of the most recent model refresh? An income model calibrated against 2019 earnings distributions is materially out of date given subsequent wage growth.
- What lawful basis supports the original data collection, and can they provide the consent record or legitimate interests assessment documentation?
- Does the file include suppression against the Mail Preference Service (MPS) for postal use, or the Telephone Preference Service (TPS) for telephone use?
A supplier unwilling to answer these questions in writing is a supplier you should walk away from. Compliance documentation is not optional under UK GDPR; it is the buyer's first line of defence if the Information Commissioner's Office (ICO) ever investigates a campaign.
If you are also considering targeting by homeownership status or property type, note that our forthcoming guide on UK homeowner data covers how property tenure and value are compiled and how they complement income profiling in direct marketing.