The improved availability and precision of mobile and social media data offers an opportunity to examine the true impact of social advertising on bricks-and-mortar sales in a robust way, and confirms its positive influence on purchase decisions.
Models of marketing effectiveness typically aim to evaluate the impact of an independent variable (something that can be changed, such as advertising or price) on a dependent variable (things like brand perception scores or sales). The problem is that, up until this point, most marketing effectiveness research outside of performance media has been fundamentally flawed by failing to control what statisticians refer to as ‘confounding variables’. These are factors that are often outside the researcher’s control and can have an impact on both independent and dependent variables. These can be incredibly diverse and include anything that has an influence that can’t be controlled, such as the weather, a major news event, distribution, or competitor activity.
This presents a problem for marketers. Without explicitly knowing it, agency planners have become adept at identifying confounding variables and using them to tear market research apart (usually if it gives an answer they don’t want). They can find a multitude of confounding variables that somehow prove that the media plan or creative execution must have been successful. While this may provide short-term vindication, it ultimately undermines the confidence that brands have in what really is driving their marketing successes (or failures).
One area that has been hard to measure and become a holy grail of measurement over the past decade is proving the link between investment in digital media and sales through non-digital channels (or, in the real world, ‘shops’).
Social media and its impact on driving offline store visits has probably been the most discussed topic in this area. Through its very nature, social is often ‘always on’ or running alongside other elements within a campaign. This makes it difficult for even the highly skilled practice of econometric modelling to separate social from other channels, and, instead, it is often attributed as part of the base sales.
An alternative approach for advertisers is to test the impact of social by asking consumers what influences their purchase decisions. Unfortunately, as sophisticated as the human brain is, the memory is an unreliable measurement tool. There is, after all, an extensive and damning back catalogue of research, built up over the past 40 years, that quantifies consumers’ inability to identify what drives their purchase decisions and general behaviour.
Impact is also often assessed using a test and control region, but this introduces a whole plethora of confounding variables, from store location and transport networks to the weather and digital usage. This represents a real issue when the key to successful measurement is to keep everything the same and only change one variable, in this case the exposure of a group of people to advertising.
However, an answer is now here. The improved availability of mobile data from providers such as EE offers an opportunity to examine the impact of social advertising on sales in a robust way. It is now possible to reduce, or completely remove, the chance that confounding variables emerge to undermine confidence in the results.
Empowered by the opportunities that this data-led environment provides, we set off with an ambition to control the different variables and measure the link between social investment and footfall.
It’s often said that simplicity lies at the heart of genius, and the wealth of data available allowed us to implement a deceptively simple methodology. Facebook has access to more than 10 million users’ mobile phone numbers through its smartphone app, which can be cross-referenced with EE’s database to provide a truly trackable solution. With the amount of information that each user is prepared to post on their profiles, access to this mobile database would provide the depth of demographic information required to identify an individual and if they had been exposed to advertising.
Armed with this data, EE would then create a geo-fenced area that would log if those who were served ads physically entered a predetermined zone. However, there were still limitations in how accurately we could track each mobile signal, which meant a large area such as a supermarket, DIY warehouse or cinema was required to monitor the number of people entering a retail destination. With its huge outlets, this made IKEA the ideal partner for the study to accurately register people who had entered the area and visited the store
We used Facebook to create two identical groups. The first ‘test’ group consisted of all those people whose mobile phone number ended with an odd number and IKEA’s ads were only to be served to this group. The second ‘control’ group (all those with a mobile phone number ending with an even number) would receive no advertising.
By analysing mobile behaviour and logging the visitors to the area, we were able to identify exactly who had visited the store and whether they were part of the test or control group. During a pivotal sales period for IKEA, we worked with fellow Denstu Aegis Network agency iProspect to log people’s actual behaviour, rather than their claimed behaviour, by recording if those who visited the store had received advertising.
The results conclusively proved that Facebook advertising significantly influenced consumers’ shopping behaviour. Geo-targeted Facebook ads successfully attracted additional visitors to IKEA Cardiff by driving a 3 1 % uplift in store visits among 22- to 25-year-olds and an 11% increase across all age groups. These additional visitors delivered an ROI of 6:1 against the media investment with Facebook during the campaign.
The campaign has wide-reaching and profound implications for how advertisers should measure the impact of digital, and particularly social media, on bricks-and-mortar sales. It also highlights some of the basic ways in which advertisers should be using paid-for media on Facebook and then testing the impact on offline sales.
Using a platform such as Facebook is about being more precise with your advertising (as long as the cost differential isn’t too much). There are at least five independent variables that could be tested using this methodology alongside the vast data available through Facebook.
- Socio-demographical: This study highlighted that younger people were significantly more responsive to advertising on Facebook. Digging deeper into the mobile data and Facebook data may well help to further refine this audience. It could be that fans of a particular genre of films or music are more likely to buy a product, or fans of a certain brand on Facebook will have greater propensity to act upon a targeted, commercial message.
- Brand affinity: Facebook provides several ways to target people who have previously interacted with brands or its competitors. By building lookalikes of the people who are fans of your brand page, you can target a new group of potential customers with appropriate messaging. This may provide smaller groups, but allows you to build scale and use Facebook as an extension of an existing eCRM programme.
- Product need: Facebook recently launched a fairly audacious partnership categories programme which uses third-party data about people’s offline behaviour and then makes it available to target people in Facebook. This opens up a whole range of possibilities such as identifying people whose home insurance is due for renewal that month and then serving the right messages to them.
- Temporal targeting: Many brands find themselves embroiled in a constant struggle to be front of mind all the time – a desire driven by the hope that they can serve key messages at the precise moment when a purchase decision is made. Measuring the times that people visit a retail store can help pinpoint the triggers for a purchase decision based on the time and day. For example, the weekend after The Great British Bake Off‘s return to TV screens may see a spike in kitchen equipment sales and represent a ripe time to deliver social media messages.
- Geographical targeting: When it comes to retail outlets, location is a critical variable as exposure to a piece of advertising is unlikely to inspire someone to drive the length of the country to buy something. As such, it is vital to determine just how far most people will travel and by doing this, you can create a ‘geographical footprint’ that identifies hotspots for effective advertising. This could mean identifying the distance people are willing to travel against product category, or surrounding locations that drive the largest increase in footfall.
The benefit of this approach is that it allows you to test just one variable and create two identical groups based simply on their phone number. The next step here is to rapidly test this approach plan and then add scale from the insights and lessons learnt, while other tests run in the background. Building on previous learnings and conducting these rapid test-and-learn initiatives means we should certainly start to understand what at least 50% of our advertising budget is doing.
Ian Edwards is Head of Strategy at Vizeum UK
This article was originally published in the December 2014 edition of AdMap. You can read it online here.