Monday, 16 May 2011

In search of a fourth dimension in segmentation

In 2008 McCain Foods, an FMCG brand, decided to embark on an eCRM (Electronic Customer Relationship Marketing) programme. The original premise was that the company wanted to establish whether, by engaging consumers using email, the effectiveness of digital could be measured in terms the advertising world understands well - brand consideration and brand preference. The eCRM programme scored some quick wins through segmentation, including an increase in brand consideration of 11%. But what really started to attract attention was the methodology used to establish and track value in terms of purchase frequency and sales revenues, through comparing like-for-like segmentation with supermarket data.

The original brief for McCain Foods' eCRM account was simple: establish whether or not digital could be used as a channel for changing brand consideration. The context was straightforward too: in an advertising ecosystem where TV is the principal means of affecting brand preference, and in which TV audience reach is becoming fragmented and diffuse, is there a new way of using digital to shift consumers' perception?

One suspects it is a common question. The McCain brief came a decade after the more basic questions about whether digital could offer a new channel for effecting sales were answered by the likes of Amazon, and five years after eCRM strategies for online retailer brands like Virgin (and in particular Virgin Holidays) demonstrated that relevant contact using pertinent, timely emails brought cross- and up-sell efficiency to the new digital retail channels. A whole universe of retailers have, and do, use eCRM to drive sales through their online stores and optimise the ongoing contact with their customers.

It's actually very easy to see the effects of an eCRM programme when you can track a customer from first moment of truth, say an eyeball on a cookie-dropping interactive ad, or a click on an Adword all the way through to lifetime value via a checkout mechanism. There is little mileage in labouring the arguments for segmentation per se (other than the observation that a three audience segmentation served by three separate email campaigns increases the effective return by 50%). That given, typically in the progression of an eCRM programme there is a straightforward series of steps that happens:

  • Demographic segmentation
  • Customer Touchpoint Planning (also known as Customer Touchpoint Management or Customer Journey Planning) for each of these segments
  • Test emails are sent
  • Results tell us something of the recipients' behaviour (both interms of reaction to calls to action as well as, further down the line, transactional and value-oriented behaviour)

The results deliver the next dimension of segmentation: behavioural, which attempts to infer likely behaviour from previous behaviour to inform messaging relevance. A new set of tailored emails is sent and the programme is refined. At this point planning insights need to be brought to bear to hypothesise motivational drivers.

Take working mum Sue. If we don't understand her motivations and drivers, we may look simply to her working mum status, infer she lacks the time to cook a proper meal, and promote the speed at which Oven Fries can be prepared. Actually, if we think more about motivation rather than circumstance, she doesn't have time to create original things to do with her young children. If she has a certain background (something we can easily establish) we may find she is motivated primarily by her kids' welfare, reinforced by her lack of time. This may offer some direction for us when we create messaging that will appeal to her directly and quickly.

These "What" and "Why" segmentation layers are key to what happens in subsequent phases. Without this "3D" approach to segmentation, all that is happening is a tightening of efficiency. And actually just the first two make for a pretty well oiled machine. But it does not help us identify motivation, and an individual's motivation may have a substantial effect on what will or won't encourage them to engage or buy. Adding the motivation element adds effectiveness.

McCain Foods took this approach. The segmentation they had already performed on their base of several hundred thousand email addresses was broad, seven in total, and attitudinal in nature. At one end there were brand resistors, at the other brand advocates. Resistors divided into two broad camps: those that were resistant to the frozen potato products category in general, and those who were consumers of the product but were resistant to the brand on the basis of price (or perceived price) differences and were likely to opt for supermarket own-brand rather than branded products.

The segmentation was tweaked to match exactly the then-commercially available datasets available from Dunnhumby describing Tesco (the UK's largest supermarket chain) shopper behaviour.

Data, as is usual for most brands, had been acquired more or less by accretion; competition entries, some bought data, some from third party partners, some from website opt-ins, voucher promotions, field marketing activities and so on.

All the data was piled into a generic segment, and as segment affinity was identified through behaviour (which links they clicked) or responses to questions (outbound questionnaires, additional promotions), records were assigned to the appropriate segment.

The usual persona development was then performed on each segment, creating a set of usable pen portraits that allowed creative to be briefed efficiently. This persona work was validated using highly tailored surveys sent to samples from each of the seven segments, and in turn this enabled the creation of subsets. At this stage the base consisted of seven broad attitudinal segments, split into demographic subsets. The persona development allowed the creation of some notional motivational drivers.

So how does this help? Well, let's take Sue Example. Sue is a category resistor. Sue went to college, and has a young child. So far so demographic. Sue is well educated, is unlikely to believe everything she is told by advertising, and she probably thinks frozen fries aren't good for us. Sue is primarily concerned about the wellbeing of her child. McCain's principal products are made of potato and sunflower oil, and nothing else. But saying "they are good for your child" over and over again is unlikely to change Sue's attitude.

But we know she has somehow ended up on the database, and if we can find a creative way of engaging her then we may have a chance. The creative execution used for Sue in the email campaign that ran for ten months used motivation as a starting point for the solution. In month one Sue received an email introducing Farmer John and his red tractor for printing out and colouring in. In month two Farmer John illustrated how a field is ploughed. In month three how potatoes are planted. In month four, the lifecycle of a potato. In month five, the painting-by-numbers included harvesting. And so on. By the end, Sue knew that if it was McCain, it was good for you. She had had some safe, fun, kitchen-table activity for her child. And her attitudes had changed.

In fact, over the first ten months of the activity, engagement with brand resistors (measured simply by open and click-through rates) went up from 14% to 63%.

Using a brand tracking agency to measure national brand consideration and brand preference scores, the campaign was benchmarked so that the effects of the eCRM activity could be established. An assumption was made that the effects of TV, in-store and outdoor advertising would be felt by consumers both in and not in the eCRM database and could therefore be factored out. That said, clearly those in the programme would naturally index higher than the national average simply by dint of being exposed to the brand's messages (let alone because it was partly a self-selecting audience) - 61% versus 20%. By the end of the first year this had changed markedly: while the national score had fallen to 12%, the brand consideration score in the eCRM base had risen to 64%. This measure, the de facto standard for measuring the effectiveness of television advertising, showed that eCRM did indeed have a place in the arsenal of high level marketing.

Taking the next logical step in measurement would, for an online retail brand, be easy. Measuring the correlation between changes in brand consideration or preference and actual sales would be straightforward given a directly auditable transactional process. For an FMCG brand whose sales are entirely through third parties (and actually, it doesn't matter whether the product is ultimately online or in physical stores; any sales chain where the sale is owned by a third party vendor has the same challenges) there are some reasonably robust estimation models for making assumptions about how well marketing drives revenues. However, direct attribution remains a real challenge. McCain Foods was in the very fortunate position of being able to benchmark changes in the attitudes and behaviour, including sales behaviour, against data available commercially from its largest third party vendor, Tesco. The segmentation exactly matched. And purchase data was available through the Tesco Clubcard shopper database, which collected information about every Clubcard holder's shopping basket at every checkout in the country. McCain could see, in detail, purchase volumes by product, by segment, by location and by value.

Address data was not available, so individual records could not be extracted and compared to the base, however McCain already had a regular programme of surveys called 'Golden Questions', which allowed interrogation of shopping habits. Normalising the response data gave a viable means of benchmarking changes over time, attributable to the motivation-based activity of the eCRM programme.

In the six months of the primary study, benchmarked against the national Tesco data, average purchase frequency went up by 3%. Matched against average transaction value, and combined with a 25% growth of the identified "brand engager" segment, sales revenue for the highest-responding segment rose by 38% - a number in seven figures.

The development of the eCRM programme became much more important to the brand, and in 2010 a website was launched to support the activity. Based on behavioural targeting principles, the site serves content based on segmentation inferred from the first few clicks on the site. So, if a visitor is known (and identified using an opt-in Cookie) the content they are served will be based entirely on the predefined customer journey (or touchpoint plan) through the comms programme, based again on tapping into motivation-based insights. If a visitor is not yet known, then the opening interfaces are designed to offer choices that require visitors to effectively self-select into tentative demographic (kids, no kids, partnered, likely to be single) and behavioural (shops for household, influences family activities) segments. Using a content management system that is integrated with the content management and eCRM broadcast and analytics suite, each visitor gets content that is appropriate to the (inferred) segmentation.

As soon as possible, visitors are encouraged to opt in to on-site tracking so the final segmentation can be nailed down and the visitor brought into the on- and off-site eCRM campaign. The driver for this is a promotion mechanic called Spud Shillings, tokens that are awarded for various activities including viewing recipe videos, downloading meal planners, sharing content using social networks including Facebook, and uploading content (which is encouraged within segments that are likely to respond using a simple model based on Forrester's Social Technographics profiling). These tokens can be redeemed for product discounts or for third party promotions, for example free cinema tickets, kids' activities or sporting events. Again, activities and selections are written back to the database to inform much more granular segmentation and so that we can analyse the effects of particular calls to action and the usefulness of trade partnerships for future planning.

So what does the future hold? The next step will be the incorporation of social media tools into the site, at a level significantly higher than simple 'share this' or 'like'-type functions. The recent launch of Google Analytics beta tools for tracking on-page activities at event level, so that clicks on a video play button or expansions of a rollover can be tracked, have provided the brand with the means to build analytics into the site to a previously impossible degree. Coupled with Facebook's API, which allows developers to bring the entire Facebook suite of tools (including share with specific friends, messaging, live chat, content distribution) onto a brand-hosted site, this means we can start to add a fourth dimension to the segmentation: advocacy. This will revolutionise what McCain and other brands involved in the testing of the tools can do with eCRM. No longer are we limited to the attributes of the customer, or the motivations behind their behaviour. The door has been opened to a deeper understanding of how specific individuals interact with their own networks, and this profound extra dimension could help grow customer bases exponentially amongst audiences we could never otherwise identify or, if truth be told, credibly or relevantly reach.