What you see and what you get

Getting the big picture is vital for any marketing director who wants to demonstate the effectiveness of what the department is doing.

Boards care little about the detail of each campaign and everything about how many sales it will lead to.

Which gives marketing something of a problem, since all of the measures it receives relate to the micro-level of specific activity. From clickthroughs and site hits, coupon and phone responses, opportunities to see and awareness, all media research has been based around demonstrating the impact of a specific ad, position and impact.

So where should marketing directors begin if they want to understand how those results add up to an overall indicator of marketing effectiveness? As budget squeezes start to make themselves felt, a raft of new techniques, models and data sources is starting to emerge.

Econometric modelling is one approach that is enjoying a revival, albeit in a new, enhanced guise. “Optimisation is the reason we launched,” says Jo Young, managing consultant at Marketing Reality. “There are lots of different channels and ways to spend your marketing budget. We are digging into how they impact on each other.”

Her consultancy combines econometric modelling, which unpicks multivariate interdependencies, with customer behavioural analysis. That means blending a technique previously used to justify TV advertising with one used to show the results of direct marketing.

“We’ve gone way beyond that by looking at data in time series and including factors that have not been looked at before, because marketers need to understand what could have influenced sales,” says Young. To leverage this, marketers need to be able to extract consistent data sets, preferably over a two-year time period, on both marketing activity and sales.

“You need to open your eyes to what might have influenced an outcome,” says Young. “Where we have done it, it has surprised people – but it has been helpful to them to make spending decisions.”

Getting to grip with these interactions can identify when the wrong decision has been made, as with the retailer which pulled its TV advertising at Christmas and saw sales plummet, for example. By linking media activity to customer behaviour, Young says marketers can also see how specific target groups are influenced. “We can create different models for each segment,” she says.

In most marketing departments – and especially across agencies and media buyers – disparate channels have always been looked at as if they stand alone. With the rise of customer insight departments, the ultimate target of all that marketing has emerged in a unified view of the customer.

Digital marketing is threatening to become a 300lb gorilla. Whatever your problem, the answer is search

Recognising the singularity of each customer is forcing the pace of unified media analytics. As Carlos Soares, head of customer insight at British Gas, says: “The challenge we face is making sure that any money we spend is not thrown into a black hole, but into something the business can see. If we are currently spending £1,000 to get 80 sales, could we spend £800 and get the same thing?”

As a former monopoly, his company has data on every household in the UK. As well as detailed transactional data, it captures feedback from direct marketing, TV advertising and its field sales force. Each of these had been modelled and tracked within the company, but often within separate spreadsheets with pockets of individuals carrying out localised modelling.

“We had to say we need to put that information into a new technology and put demands on our analysts to come up with something,” says Soares. That something turned out to be EMB Marketing Science’s Media Optimizer.

“We used to do econometric modelling, but we weren’t able to bring different channels together and understand how media impact on each other. If we were spending above the line, what would that do to an email campaign, for example? We couldn’t see it in hard numbers, but we knew if we pulled out, it was to the detriment of sales. This takes us to that next step,” says Soares.

With top-down buy-in, the system is being applied to marketing planning for the first half of this year in the belief that it will prove what British Gas gets for its marketing investment. “The big thing is that you can’t argue with hard facts. One of the big things here is input-driven decisions – this is one of the pieces of input,” says Soares.

In the current climate, having evidence for decisions could be critical to avoid making fundamental errors. One of the most obvious issues is when a large budget item, such as TV advertising, is automatically targeted for cutbacks without considering what impact that will have.

As Ian Liddicoat, director of EMB Marketing Sciences, points out, “this system allows clients to run multiple media planning scenarios in a web tool and display a curve for variations of the media mix.” By considering the effect on both volume and cost of sale for each approach, marketing gets more effective, not just more efficient.

“Our models are focused on the future,” notes Liddicoat. His company has an actuarial background and is therefore very used to building models that not only fit historical data, but are also highly predictive of future outcomes. “If a client can save 15 to 30 per cent of their media spend, that is a sizeable sum of money, often tens of millions,” he says.

Aligning data sets to allow for crosschannel comparisons is no easy task. As Richard Davies, managing director of Acquity, points out: “For TV, you have got a six-week AB deadline and you get the research six days after, whereas online you can test and evaluate in one day.”

By combining the 46 million consumer records on his agency’s Continuum database with client data sets, any sale can be matched back to a targeted message via the postcode. “It is not aggregated data, it is at individual level,” points out Davies. “Clients have traditionally aggregated data, so if they got 50 people responding, they did not know who they were.”

Digital channels are transforming the way response and sales can be tracked. They also threaten to become the new 300lb gorilla. Until the late 1980s, whatever your marketing need, the answer was always a 30-second TV ad. Now it is search, not least because, in linear media analytics, the last click wins.

Yet Google’s own research has shown that 67 per cent of offline marketing activity drives people online. For some clients, 90 to 95 per cent of sales happen online, but might not be won without that offline support.

“Nobody starts their journey with search,” points out Davies. Taking the last click model in isolation distorts the effectiveness of this channel. As he points out, a home insurance client may pay £32 for a category search. If it only converts 50 per cent of those clicks, the cost is really £64 and if conversion runs at 10 per cent, it costs them £320 per sale.

When his agency looked at the effectiveness of Hiscox Insurance’s media spending, it found that with two million possible targets, category search was proving too expensive. Instead, it built a plan that involved targeted outdoor, print ads in The Times and Daily Telegraph and no TV ads at all. By spending on brand, rather than category search, the company was winning clicks for 20 to 30 pence each and getting 2.5 times more searches than in the prevous year.

What all of these approaches are deploying is a much more data-rich method of understanding how inputs have led to outputs. Transactional databases are providing the most complete picture of who has purchased, while marketing plans and media research give a sketchier indication of where inputs were applied.

As analysts know, the more data is available, the more accurate the prediction and the closer the curve lies to the optimal result. While customer insight departments may dream of having a perfect view of their marketplace, something like it is actually being developed in the United States.

“We are aiming to be advertisers’ single source for market research and media information,” says Bill Harvey, president of TRA. A veteran of a similar attempt called Scan America 15 years ago, he says the new business is “different to every single source that has ever existed.”

Previous attempts involved large-scale consumer research panels which kept TV diaries (allowing their exposure to advertising to be tracked) and scanned the barcodes of all the products they purchased. TRA does not require any of that effort by individuals and is also much larger in scale.

“In the US, 86 per cent of all transactions are accompanied by frequent shopper cards,” says Harvey. His company has negotiated acccess to this data on 54 million households every week, each of which typically buys 100 packaged goods. The company also has settop boxes in 370,000 households (with a target of reaching 1 million) providing second-by-second data on TV viewing.

Using an analytical database system from Kognitio, TRA is now able to deliver to its advertising clients a web-based, self-service tool which links advertising input to sales output. “It is what marketers have wanted all along,” says Harvey. “If you ask an agency or highly-paid analyst to do that and all they can do is retrieve page reports from other tracking systems, their job is in jeopardy.”

To preserve privacy, clients can not look at data below 30 households in granularity. Media that are not tracked by TRA, such as in-store marketing or direct mail, can be mapped against household geography and store distribution to incorporate their contribution. The company is also in discussion with ISPs about including Net data, or writing APIs to allow behavioural tracking data to be ported into TRA.

That could help digital marketing analytical tools which have a strong grip on what is happening online, but are blind to offline influences. “Marketers need to look at an holistic view of all their marketing activity and how it has impacted,” acknowledges Craig Whiston, client services manager at Coremetrics.

Because of the way his company’s system builds a LiveProfile of historical online behaviour, each consumer can be followed across the Web. Once that cookie trail has been associated with demographic data, it unlocks all of the personal information captured by customer databases.

It is a long way from the world in which marketing threw everything at the wall and hoped some it would stick. Boards are demanding a much clearer picture of what they are getting for their money. While the outline is still open to interpretation, at least the shape of things is finally being filled in.


  • The UK has 14 national and 90 regional daily newspapers, plus more than 3,500 consumer magazines
  • There are 283 TV channels and 285 radio stations
  • During the week, an average adult spends 24 per cent of their awake time watching TV, 13 per cent listening to the radio, 7 per cent using the Internet and 3 per cent reading a newspaper or magazine. In homes with an Internet connection, usage rises to 10 per cent (source: IPA TouchPoints)
  • During the weekend, an average adult spends 29 per cent of their awake time watching TV, 11 per cent listening to the radio, 4 per cent using the Internet and 4 per cent reading a newspaper or magazine (source: IPA TouchPoints)


    Leave a comment