Matching rule for duplicate customer records

Hi Dominic, I've got a data question rather than a fault. Some customers are showing up twice in the brain, once from the routing source and once from the helpdesk source, with slightly different names and spellings. Nothing is broken, the connectors load cleanly, but the account managers are seeing two records for one site and it is confusing the picture. Before I change anything, I'd like to agree a matching rule with you so that the merge is done once and done properly. I don't want to delete anything I can't restore. Can you tell me what you'd need from me to look at it? Imran

Hi Imran, Thanks for flagging it. I've taken ownership of this, and I've copied Marcus so he is in the loop. Here's what we know: the connectors are healthy and the records are loading, so this is about how they are matched, not about a failure on our side. What we don't yet know is how widespread it is. Could you send me a few examples? For each, the customer id from both sources and the name as it appears in each. That will let me see what the records have in common. My first thought is that company number is the most reliable thing to match on, as names and spellings drift. I'd rather not guess at a second step until I've seen your examples. Next update by 16:00 tomorrow. Thanks, Dominic

Hi Dominic, Sorry, it took me longer than I said to get this together. I wanted to test it first. I went through the examples and the pattern is clear. Where both records have a company number, it matches every time. A few records have no company number, mostly sites, and for those the site postcode is the next most reliable thing. So the rule I'd suggest is: - match on company number first - where there is no company number, match on site postcode - anything that still doesn't match stays separate and goes on a list for me to check by hand I tried it on a sample and it gave the results I expected. Which customer id is the survivor in each case? I'd suggest the routing record, since the account managers know it, but I'm happy to follow your view. If you agree, I'd like to merge the duplicates. Imran

Hi Imran, Agreed on the rule: company number first, then site postcode, and anything left over stays separate for you to check by hand. I think that is sensible, and I'd keep the routing record as the survivor, as you suggest. Here's what is confirmed: the duplicates are merged and the account managers should now see one record per site. I kept the list of the records before the merge, so we can reverse anything that looks wrong. What I haven't confirmed is the leftover list. That's yours to review, and if any of those look like true matches, send me the ids and I'll fold them in. If anything looks odd on your side in the next week, tell me and I'll pick it up myself. Marcus is copied so he can mention it in the next review with Nicola. Thanks, Dominic