Back in 2001 I built my first Data Warehouse (DW). It was delivered using SQL Server 7, VB Script, Data Transformation Services (DTS) as the Extract Transform and Load (ETL) tool and NADIS Onlooker delivering name and address matching.
As this was my first DW, I assumed that it was quite normal to deliver as Single View of a Customer (SVC) with cleansed addresses tagged with a Delivery Point Identifier (DPID) matched against the Australia Post, Postal Address File (PAF).
After successfully delivering the financial services DW, I boldly embarked on a career in Business Intelligence (BI) consulting. It is now fifteen years later and I am yet to build another DW that tackles the SVC goal in such a complete manner. As you can imagine, I am surprised and honestly, a little disappointed.
Now, why is this? Have clients given up on this goal? Have source systems like the ubiquitous Customer Relationship Management system negated the need? Maybe data quality has improved so greatly that there are no such things as duplicate customers in the data source(s) or has Master Data Management delivered on its promise?
Frankly, my experience is that none of these are true. I still hear marketing and sales begging for a more personalised relationship with their customer and complaining about the quality of the data that is captured and the reluctance of IT to fix the issue. This is all made more confusing by not being able to find anyone who is going to own this issue.
This could be just me, but my suspicion is that BI, as an industry, has failed to make a SVC the norm, instead of the exception.
So let’s recap on why we want to do this:
- It ensures that mail outs reach the right person, the cost/lost opportunity of not doing this is huge.
- The same client doesn’t receive multiple unrelated interactions (i.e. a phone call, email, mail out, etc.) and they don’t end up frustrated with a perceived ‘scatter gun’ approach.
- It helps ensure compliance with the Privacy Standards. For example, if there are duplicate records for a given customer record, what happens if only one has the ‘do not mail’ flag ticked and you mail the other one?
- The true risk profile of a customer can only be assessed once you have a single view.
- Accurate segmentation of the customers is possible based on their value. This means that a valued customer will receive the appropriate level of service, and as a consequence, be retained.
- Cross-sell or up-sell opportunities become possible.
- You don’t commit the cardinal sin of mailing a customer whose ‘deceased’ flag was only set in one of the records against their name.
- You know how many distinct customers there are and consequently what the market share is.
These are just some of the advantages, as you can see, they are real and tangible. So why doesn’t it happen more frequently? Here are my theories:
- Name and address matching is still considered as complex since the concepts (e.g. ‘what makes a match’) are hard to communicate to non-technical sales staff.
- Having a SVC has not been given a dollar value. It falls into an invisible ‘lost opportunity’ that no one sees.
- Traditionally, the software required to deliver an SVC has been costly. This is certainly no longer the case.
- The cost of sourcing a PAF file is considered prohibitive.
- When road maps are developed, delivering an SVC is always scheduled to happen once the operational reporting, dashboard, KPI, ad-hoc reporting, etc. needs have been addressed. Consequently it never quite happens.
- There often isn’t a mature Information Management strategy to support an SVC. This includes:
- Data Owners/Stewards
- Data Quality
So let’s assume that an organisation has been able to get this type of initiative running. Why doesn’t it deliver:
- A business process to deal with ‘near matches’ isn’t in place. These are customers that have fallen outside of acceptable tolerances for a true match. As a result they are not matched manually and the original issues continue.
- Calculating a customer’s lifetime value (CLV) can be complex, this impacts segmentation.
- Matching rules are not maintained overtime.
- Addresses are not corrected if mail is returned due to an invalid postal address.
- Ownership of the client is lost, so address changes are not received from the client. This happens in mortgages where the mortgage broker is not updated on client detail changes once the bank has written the loan:
- The staff who capture customer details are not:
- Incentivised to get the data right at the point of sale. This can be a carrot, stick or both.
- Educated on the consequences of getting it wrong.
- Supported by making the source systems easier to use and therefore capture the customer data correctly.
So this blog is a call to action, unless of course we have given up and it’s all too hard!!