DIRECT MARKETING: Too clever by half

Direct marketers are beginning to realise shiny new pieces of technology have their limitations and cannot replace traditional methods of analysis, writes James Curtis

Direct marketers are beginning to realise shiny new pieces of technology

have their limitations and cannot replace traditional methods of

analysis, writes James Curtis

There is no doubt that technology is driving direct marketing forward.

The ability to gather and analyse vast quantities of data is the bedrock

of all targeted marketing, allowing marketers to understand and predict

how their customers behave.

Technological developments have made it affordable for businesses to

capture and store data on customers and, as the volume of information

increases, marketers look more and more to technology to provide answers

to their questions.

As a result, the language of direct marketing is now peppered with terms

more suited to the bridge of the Starship Enterprise than marketing

departments. Massively parallel processing, data warehousing and neural

networks are some of the buzz words of high-tech marketing.

But has this head-long rush into high-tech happened too fast? There is a

growing consensus that direct marketing is concentrating too much on

buying new toys and not enough on using them effectively. As Barry Hill,

senior vice-president of product development at American Express says:

‘The technology is there before the marketing imagination has arrived.’

A classic example is neural networks. This is the technology that

supposedly imitates human thought by spotting patterns in data: they can

range unsupervised through information and draw conclusions. The

attraction in direct marketing is that they are good at identifying

trends in customer databases - usually a complex mix of demographics,

purchasing behaviour, lifestyles and mailing history. This can then be

used to forecast behaviour, analyse response or look for clusters.

For a while, neural networks were held up as the next-generation

modelling technique, an unbeatably powerful analysis tool. However, they

have failed to live up to their promise, with many businesses finding

that they get better results using traditional analysis techniques, like


The number of companies actively using them in direct marketing is very

small, although they are widely accepted in financial services as a risk

assessment and credit-scoring tool.

The problem many companies have found is that unless their data is very

well organised, clean and properly prepared, neural networks and other

leading-edge analysis techniques do not bring anything new to the party.

The boring fact of life is that the majority of time should be spent

preparing data for analysis, not analysing it. Edwina Dunn, managing

director of Dunn Hunby Associates, contends that 90% of companies have

poorly prepared raw data. While this is the case, investing in sexy new

technology is a waste of time. Dunn argues: ‘People are focusing on the

technique because it is easier to debate whether you should use this

system or that. But they are not answering the really important

question: have they got all the data?’

The same is true of data-mining techniques, such as MPP and PC-based

versions like Rapidus and Viper. These allow incredibly rapid access to

huge databases, but unless the marketer knows what he is looking for and

what to do with the information once he has found it, they are not worth

the expense, and, in the case of MPP, this is no small investment. Neal

Muranyi, marketing director of the Database Group, says: ‘You can buy

all these gizmos and still not get a result. You can get swamped with

information and not know what to do with it.’

The key requirement when considering new technology is to ask whether it

will improve understanding of customer behaviour and allow the company

to communicate with them more efficiently. If, instead of bringing

greater wisdom, new technology is clouding marketers’ understanding of

what they do, then there is a problem.

This is one of the oft-quoted problems with neural networks, in that it

is often hard to understand how the system arrived at a particular

conclusion. Tony Masters, director of the Computing Group, says

marketing people often ‘can’t see what is going on in the neural network

model. They don’t understand why some prospects are better than others

and so can’t adjust their marketing strategy to suit’.

The benefit of using traditional techniques is that the marketer can see

how every conclusion is arrived at and so gets a more thorough

understanding of the logic behind trends in the data. In a strange

admission for an IT consultant, Dunn admits that ‘properly-trained

humans are more expert than systems’.

An over-reliance on IT can make marketers forget the most basic

principles of what they are doing. It is perfectly possible to get lost

in the science of your technique but miss the point of why you are doing

it. Keeping a keen eye on what you want to achieve from the database is

key to managing it properly. Richard Knee, commercial projects manager

for Heinz, which has invested extensively in its database management

system, says that ‘if people understood what they wanted before going in

they could identify their target database very quickly’.

Profiling need not be a mind-blowingly high-tech exercise. An example is

the Pareto principle which states 20% of customers account for 80% of

profits. Getting some of the customers in the middle of the database

into the top 20% is often a question of overlaying the profiles of the

most profitable onto the rest of the database. Chris Morris, database

director at NDL International, says: ‘This can be done just as

effectively with standard techniques as with neural networks. You can

leverage profitability without getting anally retentive about what

package you use.’

Some companies, however, are making imaginative use of neural networks

and other hi-tech analysis tools without allowing it to drive their

strategy. American Express is taking targeted marketing one step further

by customising the promotions in its account statements to match the

requirements of individual customers. This information is taken from the

company’s European data centre in Brighton and analysed to identify

trends in customers’ buying patterns. An extra page is being included in

British card holder’s statements, containing offers designed to match

the spending patterns of the customer.

More than 80 UK retailers are participating in the scheme, which will be

extended later in the year to include money-off coupons for stores local

to the customer. The results so far are impressive: response rates for a

non-customised car rental offer were 0.06% but rise to 9.32% with the

customised promotion.

Barry Hill, at Amex, says neural technologies ‘are playing a role but

are not the main thrust’ of the programme, which draws on ‘a salad of

techniques’ for analysing the data. ‘We are quite happy to experiment -

whatever works is great. There is rarely a consensus of what is the most

effective tool.’

But Amex still employs traditional analysis techniques. Hill explains

that the company uses neural networks ‘upstream’ in the analysis

process: ‘If we have 12,000 variables we may use it to cut it down to

800 and then use classical methods to look at the rest.’

While accepting that many marketing people do not know how to use

technology, Hill argues they should be more relaxed about allowing

computers to do the spadework instead of getting paranoid.

‘People need to be more relaxed about technical invasions and adjust

their roles accordingly. They need to spend less time on methodology and

more on creativity,’ he says.

The example of Amex shows that neural and other high-tech analysis

techniques can be used successfully as long as the user does not allow

it to complicate their objectives. Neural packages designed specifically

for marketers will help simplify the process, but in the mean time,

direct marketers could benefit from doing the simple things better.