Jed Mooney is the managing director of Datahold, one of the UK’s leading database management companies.  Based in London and offshore in the Philippines, Datahold’s clients range from small start-up businesses to FTSE 100 corporations.  Below Jed describes why you must rid your mail file of unprofitable names and, crucially, how to do so.

For data professionals, the biggest task we face is convincing catalogue marketers that the capture of raw data is far more important than any subsequent modelling, profiling or purging.  Indeed, any data bureau worthy of the description will tell you that raw data quality must be placed first in importance.  So, let’s be clear, before you even begin thinking about ridding your mail file of unprofitable names, make sure the raw data being refined is grade A quality.

If you haven’t done so already then, you must start paying huge attention to the procurement and capture of your mail file data.  Ensure that it is accurately recorded and consistently captured over time with a set number of variables and parameters. If property can be defined by the phrase ‘location, location, location’, any mail file should be measured by the maxim ‘quality, quality, quality’.  It is only once the quality of the data can be verified that any subsequent profiling or mail file purging can be deemed accurate and worthwhile.

Why? If the raw data is poor, any subsequent activity using that data will also be poor.  34 per cent of companies don’t even validate any of the information they collect* (research by QAS).  Yet these very same companies will change marketing strategies based on such information.  It’s madness!  Put simply, if your database is already poor by qualitative standards, ridding that file of unprofitable names will not make a huge difference to either responses or ROI.  It will be a waste of time.

Let me give you an example of how good quality data makes a difference: imagine a customer, Mrs Irene Thomas OBE, Rose Cottage, 33 the Brambles, Bromley, Kent, BA1 3FF.  Now an off-the-shelf database solution using the Royal Mail’s PAF might log Mrs Thomas’ address as 33 The Brambles and might leave out Rose Cottage as superfluous information, in addition her honorifics, in this case the OBE might also be dropped.  That’s all well and good, after all the mail will still reach her, but research has shown that people’s preferred name and address, when mailed, will generate higher responses because it is how they want to be referred to.  So, if you want to create a good quality mail file, make sure that the original data is first class and that standardisation is not removing any important elements.

OK, so moving on, let’s imagine a situation where we’ve got a good quality mail file of 200,000 postal and email addresses.  How do we ‘rid it’ of unprofitable names?  The first stage is to take out all those names that are duplicated on the database.  De-duplication is the process whereby people who are registered on a database twice are identified and the two sets of data are ‘de-duped’ to create a single address record.  This can substantially reduce a mail file size but the advantage is that direct mail wastage is reduced and targeting is improved by ensuring people are not mailed twice, to their obvious annoyance.

De-duplication will typically be on at least two levels, individual and household. There is virtually no instances where a dedupe should only be at individual level. Being able to identify respective family members is vital for the accuracy of subsequent processing and analysis.

The registration of people on a mail file twice or sometimes three times over is extremely common, especially amongst customers who make multiple purchases and move house frequently. Indeed, the average person spends no more than seven years in any one address.  It’s not surprising therefore that many FTSE 100 companies have multiple entries for one single person on their databases with de-duplication an ongoing standard database practice.

However, be warned! The big temptation when de-duping is either to drop important information to make it ‘fit’ into a standardised field format or even worse, to simply add the two sets of individual data together with no thought behind the process.  The latter, in particular, will make a database far too unwieldy and subsequent analysis can become problematic if not downright impossible.   There is a popular misconception that keeping every piece of data on a person is good – it is not!  A good data bureau can advise on how best to consolidate and refine your data properly as part of the ongoing de-duplication process.

Once the data has been de-duped and consolidated, you can start looking at the various suppression files in order to further streamline your data.  The first category of people we want to remove from our mail file are those that have (a) moved to an unknown address (goneaway) (b) died or (c) simply don’t want to be contacted.

There are over 30 different suppression files available to UK marketers such as Royal Mail’s National Change of Address (NCOA) file to name but one well-known example. Choosing the best suppression files is more a function of their reputation in the market place allied to what exactly you are looking to suppress.  To that end, a good data bureau will put forward their recommendations and ensure that they are all fully accredited with the relevant trade bodies.

As a general rule NCOA, or a similar file that has forwarding address, should be run first. The majority of companies unknowingly have duplicated customers at their old and new address. By identifying these records first and consolidating information accordingly makes subsequent processing far more efficient than would otherwise be achievable.

Now, having refined your database by de-duping, consolidating and suppressing that data, you can look at segmenting the data into viable categories.  The three criteria of recency, frequency and value should be the benchmark parameters that form the foundation of the segmentation because they divide the data according to the profitability of a customer.  Yes, categories such as age, gender, geography, socio-economic status are important but they are dwarfed by data that places value on a customer.

For example, imagine two people, living in the same city.  Mr A is 45, lives in an affluent suburb of Bristol, and can be classified at the top end of any socio-economic benchmark criteria.  Mr B is the same age, lives in the same street.  and has an identical socio-economic grading.  Using age, gender, geography and socio-economic status, there is absolutely no difference between either person.

However, using profitability criteria such as recency, frequency and value, it might be that Mr A spent £30 nine months ago on one single product.  By contrast, Mr B might have spent £50 each and every month on a variety of products, ranking him as a far more recent, frequent and high spending customer.

For that reason, when segmenting and analysing your mail file, it is crucial that you know the profitability of each customer.  This will make getting rid of unprofitable names far easier.  After all, when you know that one person ordered an accessory worth just £4.50 over three years ago, you can probably omit that record when mailing a high-net value product offering three years hence.

By using common sense and basic elementary mathematics, you can segment out of your mail file any customer whose frequency, recency and value precludes him or her from the probability of responding to your offer affirmatively.  Set a benchmark criteria, for example, recency (nine months or less); frequency (more than once) and value (£above average) and watch your mail file bring in a far enhanced response and ROI.

There is software available, particularly good database management software that allows you to access your data online and build basic models that take into account these basic variables.  More detailed analysis, however, will probably need the assistance of a data professional who will be able to build more sophisticated models – if you need them - after all, keeping things clear and simple is the most important factor in analysing and modelling data.

Once you have a cut off point for any particular campaign, you should have a clear idea that the remaining names on your mail file are profitable.  Choose your product offering carefully and test a sample of the mail file to ensure that your basic assumptions are proved correct.  You’ll find that if the quality of the raw data is good, your basic assumptions are more likely to be correct too.  That is why having good quality data is so utterly important.

On that note, please be aware that with a well managed mail file, good segmentation and modelling will bring in a good uplift of a few percentage points.  A poorer managed database might see a dramatic uplift simply because nobody had done any cleaning and analysis previously.  It’s a bit like a house clean – if it’s already clean then another clean will simply maintain it.  If it’s dirty, even a mild clean will give spectacular results.  So don’t be disappointed if you hear tales of a badly managed mail file subsequently seeing far superior uplift whilst your mail file sees only a mild improvement.

Database marketing is becoming more important, not least because the volume of data and the accompanying computing power is ever increasing.  To that end, I’ll leave you with two key points: (1) Make sure your original data is of a very high quality and (2) keep your segmentation and analysis clear and simple.  Stick to those two points and you can’t go wrong.

Contributor: Jed Mooney, Managing Director Datahold
Source: Catalogue & E-business magazine
PubSection: Good Data Housekeeping
Publication Date: May 2008