Yesterday, actually it was a normal day for me, I again experienced why marketing automation won’t work without data quality measures. I attended a webinar by a French marketing automation company who had a really nice tool to track customers on the website and collect leads for further processing. They have put much effort on a rule based engine to segment leads and a state of the art backend to have a nice working environment. I then asked: What happens if a person mistyped his e-mail address? What happens if a person set fills out his name in lowercase? What happens if the person has a typo in the postal address? First there was no answer. But then the webinar leader said: Why should someone do that? The answer is easy: Because we are human! There is a certain percentage of people who are not 100% concentrated when filling out a form. Maybe because they are using their smartphone where a typo can happen very easily, or they simply don’t know their correct e-mail address (I have often seen Austrian or German e-mail addresses with @gmail.at or @gmail.de “ but, as we all know, there is only gmail.com).
So this sophisticated marketing automation tool had no functions to validate or correct the data a lead / customer puts in the form. Garbage in “ garbage out. This people will never receive information they requested “ or have their name spelled wrong. What would you think about a company which addresses you with Mrs. john dOe .
Don’t get me wrong. Marketing automation is a great thing “ but most of the companies forget about the basics we once had: A responsibility for the data, a penchant for correct data. All we now do is collecting masses of data, and let the automatism do the job.
After that, I helped our team to correct and enrich a client’s customer database. We had 248.246 datasets of customer card information. With our automatic and manual process we saw very clearly how the data quality increases by every step. /data.mill showed us that only 27,81% of the postal addresses were correct and 71,36% could be corrected automatically. In 0,83% of this dataset there was no hit: Meaning our /data.mill could not find any clue. After looking to that data in detail, we saw that there is no chance to get this data corrected. But hey “ increasing the number of correct postal addresses from 69.029 to 246.182 is a great achievement!
After the postal addresses have been corrected we looked at the names of the customer database. It was the typical mess. Mixed up first- and lastnames, company names as first name, wrong gender code, and incredible academic titles (the funniest one was Autolakira, which describes a car body painter in German dialect). I then thought back to the marketing automation tool earlier. What would be the salutation line for this customer: Dear Mr. Autolakira Johnny Carwash & Tuning Inc. Sounds funny, but is certainly not a way to talk to a customer.
What also was very interesting, that we found 1.718 names, which could not be processed in /data.mill. We looked at the names to get more detail and of course had some funny ones like Mr. Gunsen Roses, but also a big number of names like fdjkslhjakasd. I don’t know why people type in names like this “ are there any psychiatrist out there who have an idea? “ but it was interesting that these strings have a one thing in common. Starting with the letters d,f,j,k,s, – so all letters in the middle of the keyboard. We are now looking at these datasets to find patterns and learn more about this phenomenon.
Back to the facts: We corrected 16,94% of first names and 17,82% of the last names (mostly put the first letter to Uppercase). 1,75% of the datasets had the wrong gender. So back to the first question: Why do people do that? I don’t know “ because we are human I suppose “ but what I know is, that you need to improve the data quality if working with marketing automation or you won’t be able to communicate with your client or prospect in a respectful dialogue.
