FEATURE1 March 2009

The data matrix

Every time consumers use the web a trail of valuable data is left behind. Researchers would do well to take a closer look at this information. Anna Cliffe, head of insight at Brahm, outlines the true path to enlightenment

Desk research has long been the forgotten ugly sister of the research world, the work that no researcher wants to do. Where’s the fun in sitting in front of a computer reading reports when you can be out and about conducting focus groups with real life consumers? Fundamentally, most researchers would rather be creating their own data than analysing someone else’s, and many research agencies are not properly resourced to deliver desk research properly anyway.

In recent years, however, there has been a quiet revolution in ‘desk research’. While the research world has been looking the other way, the growth of the internet into mainstream usage has meant a mass of new data being generated as consumers interact on the web.

What I’m not talking about here is using the web to conduct research or to create data. As a data collection tool the web has, of course, been embraced whole heartedly to conduct research, both quantitatively and qualitatively. Rather, I’m talking about the data that is created by consumers as a by-product of how they access and behave on the web.

As we all know, consumers leave trails across the cloud of data every time they go online, every time they shop, or search, or talk to a friend on Facebook. Consumers search for products and services on Google, write in blogs and forums, chat on social networks… and data is created.

Some of this data is freely available and some can be bought from specialist suppliers. It’s the sort of data which online media planners and e-marketeers will use as a matter of course, but as yet seems to be less widely used by the research community. But we’re interested in consumer data, aren’t we – no matter how it’s created? Shouldn’t we be having a look at this?

Data, data, data…

There are a number of key sources the researcher can investigate that provide data on customer behaviour online. These include:

  • search data – the terms consumers type into search engines
  • site analytics – the volumes of visitors to websites and how consumers move around them, and
  • the written word online – for example, forums, blogs, customer review sites like dooyoo and social networks.

Understanding how these data sources work and what they can provide to the researcher can be a useful addition to our armoury.

“Looking at search can also unearth useful data about how topics are being searched for – very useful if a brand wants to understand how to position itself online or where it needs to advertise to improve awareness”

How consumers search
Search data is freely available with Google Insights for Search and Google Trends, and can be analysed in a number of ways. Volumes of search for a particular brand can provide a measure of brand awareness or saliency, for example, and can be a useful addition to campaign tracking (even if the campaign itself is running offline). For example, if a client is launching a new product, it would make sense to track whether volumes of search related to that product rise during the campaign period. Google Trends will also show you search data in relation to key news stories to provide further context around why volumes are changing.

Looking at search can also unearth useful data about how topics are being searched for – very useful if a brand wants to understand how to position itself online or where it needs to advertise to improve awareness. Search results can even give an insight into the consumer mindset at that particular point. For example, in a recent project we conducted for a financial services client, we wanted to understand how customers would search online for conveyancing services. Looking at the types of search helped us to understand that customers were interested in the price of conveyancing, and also who was providing conveyancing locally. However, it also helped us to understand that customers had a desire to understand more about the actual process of conveyancing.

Of course, all this information could have been gained by carrying out research with consumers directly, so where’s the advantage in looking at online search volumes? It goes without saying that the depth of insight isn’t going to compare to that delivered by qualitative research. But as an addition to the researcher’s toolkit, this is an exercise that can be carried out relatively quickly and cheaply and can provide that initial spark of insight which can help to steer a campaign appropriately.

Understanding site visitors
Unlike search and trend data, site analytics data is not so freely available but, it can be bought from a variety of companies. This sort of data tells us the volumes of visitors to a particular site, and can also be available attached to demographic information if bought from a company like Hitwise. For example, it can be a ready source of information about who is visiting a site in terms of age and gender.

Analytical data can also be used to show how and where customers are interacting with a website, for example, how long they spend on the home page and what the hottest parts of the site are. Interestingly, it can also tell us when customers are leaving a website and the conversion rates for particular pages. Many researchers, of course, do bring this sort of data into a typical web usability study.

Blah, blah, blogs…
The last of the key data areas is more qualitative in nature and here we could include blog entries, forum threads and entries on review or social networking sites. Of course many research projects now use online means to capture consumer thought – be it through online diaries, asking respondents to keep blogs or by setting panellists tasks on a forum. But the web is also full of publicly displayed commentary and opinion on every single topic imaginable – including many of the brands, products and services researchers may be working on. Gathering this data in any sort of representative way, and analysing it, is certainly a challenge for researchers.

Some companies now offer web trawling and scraping services so that mentions of a brand or product online can be collected and logged. The absolute volumes of mentions can be tracked and even their ‘sentiment’ defined as positive or negative. This is an automatic process, and skill and care are needed to analyse the results. We also need to consider if this a relevant addition – it won’t be for all brands. But again, it can be a useful addition to the tool box of measures available.

Other free analysis tools are appearing all the time – Monitter (monitter.com), a site which shows Twitter output topic by topic is one example. Understanding how we can use these tools and develop them to deliver real insight will be a key issue.

Facing up to the challenges
Of course, despite all the excellent resources available, using digital data is not always straightforward and there are some challenges to face.

For example, there is a real skill to search term analysis. In a recent project for a private healthcare provider we wanted to understand how customers were searching for orthopaedic information services online. Now ‘orthopaedic’ isn’t the easiest word to spell, and in the real world customers would search for these services using a wide variety of spellings, misspellings and of course different manifestations of the same thing (e.g. ‘knee pain’). Knowing how to pull together a representative list involved both the experience of an online media planner and the technical tools which are available via Google.

“Despite all the excellent resources available, using digital data is not always straightforward and there are some challenges to face… there is a real skill to search term analysis”

Also be mindful that the demographics provided by search data must, at times, be taken with a pinch of salt, particularly geographical location. UK search data can be filtered to provide volumes from the regions, but the mechanism by which this is defined isn’t always accurate. And it’s difficult to get hold of search data by good demographic splits such as age or gender. The major challenge, however, for researchers will be integrating this sort of data intelligently into our other methods.

Client silos… or how data gets stuck
So why haven’t more researchers embraced web data? A key issue is certainly the point mentioned earlier: data likes this falls into the unfashionable realms of ‘desk research’. Many researchers also, quite rightly, have concerns about the accuracy of what is available. But a further issue here is, I believe, the way client companies are organised internally.

Typically, a client organisation may manage digital marketing activity (the website, e-CRM campaigns) separately from other marketing activity and certainly market research. Because digital activity is so inherently measurable and trackable, digital teams become experts in analysing the resulting data, much like off-line DM teams are. The data becomes a product of that digital team, stays there, and isn’t always accessible to the research team. I think it’s up to researchers to embrace this data, understand it and use this knowledge as a springboard to take our broad range of research methods back to digital marketeers.

This data is available… so let’s use it
Clearly this kind of data is not going to replace the production of primary data, of asking consumers questions directly and of measuring their response. This is a different sort of data, the pulse of the internet which ebbs and flows as consumers interact with sites and brands. Nevertheless, it’s a source of data which researchers ignore at their peril – we can learn a lot from digital marketers and online media planners in this area. As researchers, whether on the agency or clientside, getting hold of and analysing this data should be a key aim and it should become a normal part of an analysis of marketing activity.

The Yorkshire Building Society takes a new route to insight

Yorkshire Building Society wanted to launch a new online service to customers and came to Brahm for a research strategy which would help them develop a marketing positioning, name and content strategy.

Rather than conduct traditional qualitative research, we suggested looking at the available search volumes data and to use this as a springboard for understanding how customers were currently searching for the service online.

Brahm began by developing a list of target keywords for search analysis, and then establishing volumes using Google. We identified that the top 20% of keywords were delivering around 80% of the search volumes, so concentrated on these. We categorised the keywords in groups, or customer ‘mindsets’ – these included, for example, customers searching on the basis of ‘price’ or ‘a local service’. Exploring these ‘mindsets’ formed the basis for the overall content and naming strategy.

“Using this approach meant we were very focused on the actual behaviour of customers online, could make an assessment of their search behaviour and
could tailor our approach accordingly.” said Nick Gander, Online Acquisition Manager, Yorkshire Building Society.