Segments describe an area as it was. Property and household data at the address describes what is true now.
Most marketers have used geodemographic segments at some point. You profile your customers, find the segment labels that over-index, and target more postcodes that carry the same labels. It is a tidy process, and for a long time it was the best option available.
The trouble is what sits underneath. Segments are built largely from census data and area averages, and they are reclassified only every few years. Your campaign, on the other hand, lands on a specific doormat this month. The gap between those two things is where response quietly leaks away.
How segments are built, and why they made sense
Geodemographic systems group small areas, usually postcodes or clusters of them, by shared characteristics: age profile, household type, housing, income indicators and so on. Each area gets a label, and every address in it inherits that label.
That made sense when address-level data was hard to come by. A national census that runs once a decade gives a consistent, nationwide picture, and grouping areas by it is a reasonable shortcut for planning media or choosing where to post leaflets.
But it is still a shortcut. Two things are baked in from the start: the label describes an area rather than a property, and it reflects the moment the underlying data was collected.
Where the drift comes from
Segments drift in two directions at once.
Over time. Neighbourhoods change faster than reclassification cycles. New-build estates appear. Family homes become rentals. Large houses are converted into flats. Owners retrofit insulation, fit solar panels or install a heat pump. Households move in and out. None of that shows up until the next refresh, and by then the area may have moved on again.
Within the area. Even on the day a segment is built, it is an average. A single postcode can hold an owner-occupied detached house, a privately rented terrace and a flat above a shop. The segment gives all three the same label, because it has no way to tell them apart.
So when you target "people like your best customers" by segment, you are really targeting places that looked like your best customers' places, some years ago, on average.
What it costs you
The cost rarely shows up as a single line in a report. It shows up as a campaign that underperforms for reasons nobody can quite pin down.
- Wasted spend. Print, postage and media go to addresses that were never a fit, such as renters for a home improvement offer or flats for a product that needs a driveway.
- Missed prospects. The good addresses sitting in "wrong" segments never get mailed at all, because the label rules them out.
- Weaker models. Propensity scores trained on area labels inherit the same blur, so they hold up less well when you act on them.
- Muddy learning. When a test underperforms, it is hard to tell whether the offer failed or the audience was never right in the first place.
None of this means segments are useless. They remain a reasonable way to describe a market in broad strokes. The problem comes when a broad-strokes tool is asked to make address-by-address decisions.
What address-level data does instead
The alternative is to describe the property and the household directly, rather than the area around them. That is what PropertyDB is built to do.
PropertyDB covers 32.7 million UK properties, with over 100 variables per address. Instead of a label, you see the attributes that actually decide whether someone is likely to buy:
- Property type, age and build features
- Tenure: owner-occupier, private rent or social rent
- Estimated value and value band
- EPC rating and energy efficiency
- Bedrooms, off-street parking and garden
- Household composition and length of residence
- Affluence and lifestyle indicators
Every record resolves to the Unique Property Reference Number, the official key for each UK address, so property, household, business and mover data line up exactly. You can read more about that join on how it works.
Two other differences matter just as much. First, the data is continuously refreshed rather than reclassified on a fixed cycle, so a retrofit or a change of tenure can show up long before the next census. Second, every attribute carries its source, its date and a confidence level, so you can see how current a record is rather than trusting a black box.
In practice, that means you can build an audience from the facts that matter to your offer. A solar installer can focus on owner-occupied houses rather than flats or rentals. An insurer can look at value bands and property age. A broadband provider can target specific streets and property types rather than whole postcodes.
Putting it into practice
You do not need to rip up your existing approach overnight. A sensible route is to test address-level targeting alongside what you already do and let the results decide.
A quick checklist:
- Check what you are using. Find out what your current segmentation is built on and when it was last refreshed.
- List the facts that matter to your offer. Tenure, property type, value, energy rating, household make-up. Be specific.
- Remove the obvious non-fits first. Excluding addresses that cannot buy is often the fastest win.
- Profile your own customers at the address. Appending property and household data to your existing file shows who really buys from you. Our data enrichment service does this in one step.
- Run a split test. Mail one cell selected by segment and one selected by address-level attributes, and compare response and cost per sale.
- Keep it fresh. Use data that is refreshed continuously, not a file that was accurate once.
If direct mail is your main channel, EasyDM lets you build that kind of property-led audience without the usual overhead.
The bottom line
Segments answer the question "what was this area like?" Your campaign needs to know "what is true at this address now?" The closer your data gets to the second question, the less you spend on doors that were never going to open.
Want to see what address-level targeting looks like for your market? Explore PropertyDB or speak to our team for a free count by area, property type or any of the 100+ variables.