Graham Smith: Online targeting using geodemographics
15 September 2016

As Barry Leventhal highlighted in his keynote post here in January, geodemographics remains highly relevant in the world of big data. Consumer brands now typically have masses of data on the transactions and purchase behaviours of their customers, but the key drivers and predictors of those purchasing decisions, which include a combination of life stage and affluence, continue to present a challenge in terms of data collection.

How effectively can transactional data on its own support the cultivation of existing customers, let alone prospecting for new ones? What can customer spend tell us about customer loyalty, or the frequently very basic demographic information captured against customers, which is often limited to age and gender, tell us about the individual’s circumstances that may drive future purchases?

The value of established geodemographic classifications is increasingly being recognised for online advertising, as linkages are made between online and off-line data sources, to broaden consumer insight and increase targeting precision.

According to leading online agency Eyeota, sociodemographic segments - including geodemographics - dominate advertising spend in a number of sectors including retail, finance, automotive, services and utilities, accounting for 50% of all target segments bought in Europe last year.

A brief explanation of online targeting

For those unfamiliar with online advertising it can appear a complex world, with publishers at one end selling space across a range of sites, be it banner advertising or social media networks, and advertisers at the other end - usually agencies representing their end clients who are the ‘brands’ (retailers, banks, consumer goods companies and so on).

Audience segments or target groups may be generated using DMPs (Data Management Platforms), which are essentially data warehouses that manage cookie IDs - a cookie is a small file that is downloaded to your computer when you visit a website and that can be used to remember your activity on the site.

These audience segments enable specific user groups to be targeted, based on purchase behaviours, browsing history, interests, demographics, locational attributes, or other variables. Some of this information will come from the cookies themselves as they track customer journeys online, and some from the ability to connect cookie IDs to off-line attributes of an individual.

In turn, DSPs (Demand-side Platforms) allow real-time bidding for advertising space, or ad ‘impressions’, which are available through market places called ad exchanges. This process is increasingly automated, based on the information provided from a DMP, which allows advertisers to assess the value and relevance of an advertising space in relation to the target audience within milliseconds and place an appropriate bid relative to expected return.

In the split second it takes for someone to open a web page an entire bidding process has taken place, and the advertiser who places the most value on that particular individual viewing their ad has purchased and placed an advert for them to see on that page.

Bringing in geodemographics

The adoption of geodemographic classifications allows a common language to be used for targeting, as key groups understood off-line and across a wide range of situations can be reached online too, providing for a consistent multi-channel strategy.

Location-based variables may also be taken into account: an example being in advertising online grocery shopping only to people who live within an area that is covered for home delivery.

So how is all this information joined up, and how do advertisers reach the “Executive Wealth”, the “Career Climbers” and other groups they recognise?

The first link in the chain is for the geodemographic provider to work with a partner who can ‘onboard’ their data into the digital world by linking with DMPs (Data Management Platforms) and DSPs (Demand-side Platforms).

The data will have an address element enabling it to be joined to other information held by the DMPs/DSPs, and in turn to cookie IDs. Audiences made up of specific geodemographic groups can then be selected because the individual’s cookie is already linked to the associated classification.

Social media and apps

Social media sites are also using geodemographics for targeting, although they generally operate in a very different way. Facebook, for example, operates in what’s termed a “walled garden” with no exchange of information outside the Facebook site itself. There are no cookies involved. Facebook already holds information about its users such as age, date of birth, and email address. This information can be linked to geodemographic classifications directly. Brands can upload customer email addresses to Facebook to present adverts to their customers, or to build lookalike audiences based on geodemographics, as well as other variables.

Certain geodemographic groups may be supressed to ensure vulnerable groups are not targeted. Adverts for pay day loans are a case in point - although this may also mean those groups cannot be offered support services by organisations that might seek to help them.

Moving the focus to mobile, smartphone apps have contributed towards the development of online geodemographic targeting. Although users of an app may not be identified individually, the wifi- or GPS-based location of a mobile device can be used to track patterns. If a device is regularly accessing apps in the morning, evening and at weekends from the same location, the chances are the user of the device lives at that location. By identifying the nearest postcode to that location a reasonable estimate can be made of the geodemographic type of the device user. The device ID can also be linked to an email address opening up opportunities to provide a consistent message across multiple devices.

Summary

To conclude, despite seeming a world away from its traditional roots, online targeting can and does benefit greatly from the long-established and proven benefits of geodemographics. This author believes this is likely only to increase as individual data protection concerns and limitations of use of personal data mean that modelled data segmentations have an increasingly important role to play.

Graham Smith is an Associate Director at CACI, and manages the Data Services technical team as part of the company’s “Data Lab”; a team of statisticians and analysts who are responsible for creating, developing and maintaining CACI’s data products.

Any views or opinions presented are solely those of the author and do not necessarily represent those of the MRS Census and Geodemographic Group unless otherwise specifically stated.