If you are embarking on a Single Customer View, Customer Data Integration, Master Data Management, Data warehousing, Customer Relationship Marketing or any project that requires you to identify duplicate customers within you enterprise data, you may well need to engage the services of a specialist data matching vendor.
As the customer you will and should want to know what it is you are being sold. What does it do? How does it work? How does it compare to other vendor products? What will you get at the end of the day? These questions can lead you into a maelstrom of complexity, technical jargon and convoluted logic. Heuristic matching, deterministic matching, the Fellegi-Sunter linkage model, fuzzy logic, RSwoosh, DSwoosh…..Using a computer to identify duplicated customer data is very complicated.
I would not discourage anyone from delving into the finer points of Customer Data Matching – I love it!
However in the interest of actually getting something done, there is one aspect of the vendors offering that it is worth focussing on. That is – what tuning facility do they offer? Regardless of the underlying match logic, you are going to want to see (before you buy and certainly before you implement) what the vendor software can achieve with your data. This will no doubt involve some sort of test and you must be able to see and understand what has happened. You need to see real live examples of matches found within you own real live data. You also need to see those matches that they fail to find (and there will be some!). If you don’t like what you see can you change the rules and then see and understand the impact of that change. You need to be able to see the impact of changes in both qualitative and quantitative terms. It is a given within the world of data matching that changing something within the processing logic can have unexpected results. It is therefore vital that you as the customer are happy that you can see and understand the impact that the vendor solution is having on your data. So when the vendor says they use advanced fuzzy match techniques you need to see examples of where these techniques have been applied to your data – are they correct?
Be warned – fuzzy matching has the potential to do more harm than good.
And remember, a vendor is there to sell you what they have to offer, not necessarily what you need.