Marketing data (MD) often fails to achieve the planned or desired results. So non-marketers see its use to generate profit with some scepticism... Marketing data quality must underpin the analysis and support any subsequent marketing actions. Let’s explore how this can be achieved.

The aim of this article is to illustrate a simple approach (used prior to marketing analysis) to address this crucial Data quality issue. This could save you time and money in the long run, and if done periodically enables you to avoid the replication of the data quality before any marketing analysis, and improving the profitability of your marketing efforts.

You can overcome many data quality issues, if you know what you are up against! – before waiting for an interaction with your Customer to expose data quality issues, risking the potential negative impact on development activity.

Business Data Coverage

The first step is to: Confirm that the extracted MD set meets business expectation.

The way to do this is to undertake a set of “yearly” analysis and compare this against what is expected from the business MI for the same period, i.e. the correct Business Data Coverage. As a minimum check the variables suggested below per year (calendar year or business see example table 1 below:

Table 1

Year 1

It is important to include industry specific indicators, for example, Net Receivables for Credit Cards; Number or Annual Services for Automotive, number and value of Claims for Insurance, etc.

Note: If The Business Data Coverage variance is less than 5%, that is exceptionally good. Our experience is that for most companies at least one field will be more that 30% different, and we have experienced up to a 90% variance where there is data missing. If the variance is more than 5% the MD should be re-extracted.

Valid Data Population Analysis

Step 2:

The next step is to assess the correctness of each individual field. This can be achieved by building a descending hierarchy of fields, as per the examples in the table 2 below, to undertake the validation analysis:

Table 2

The examples used above have been simplified but illustrate the types of validation rules that can be built from the front-end system data capture rules, internal system rules and common sense; they don’t have to be embedded into any system.

Note: This analysis normally takes several iterations, because the data has been collected over several years, system formats have changed and the validation rules are the current ones people remember. Often the results are unexpected and can shake the heart of your core operational databases, for example the idea that all customer id’s or account id’s are in the correct format. In our experience this is rarely true. Similarly inception dates can be a minefield; this is an area where system loads/migrations have contributed to the errors, and e.g. all dates being set to the date of load.

Finally some of the valid field populations can be surprisingly low, e.g. valid Telephone number 40%, valid email address 5%.

Often the original extracts will be have to be redone after this analysis due to the lessons learned and the Business Data Coverage will be much closer.

Drop Down Reports

Whilst data quality at field level is useful, to really understand your business data quality you need to undertake this analysis at a record level.

Step 3

Take the previous table 2 and now run it at record level, from the top data field, then the top and second data field and so on.

Table 3:

Often for field 19 in the hierarchy, the percentage is between 60% to 85%, although in an intermediated business 40% is not unusual.

The 3 steps of data quality analysis

The 3 steps of data quality analysis will give you a good insight to your MD quality.

It will identify both where you have problems and usually the necessary remedial action – both IT lead, and changes to front-end systems.

These steps will give you a fully auditable, correct MD to be used in analysis. This can be used, with confidence to gain the insight to inform and plan activity to drive profit. If done of a partial set this will allow you to do the analysis based upon correct data and widen it to the full customer base.

Finally you can undertake customer development activity with confidence, knowing that each customer interaction will not risk exposing your Customers to the reality of your data quality.

By Martin Hickley,
Customer Essential
Email: Martin.Hickley@customeressential.com
Web: http://www.customeressential.com

About Customer Essential

Customer Essential provides experienced and practical consulting to drive more profitable customer management activity.