A few months ago I ran into an issue that could scuttle many of the marketing automation initiatives undertaken by a multitude of companies. While performing some genealogical research I found out my information had been cross-referenced with someone who was not me and had no relation to me. In other words, I had encountered the bane of all marketing: dirty data.
Who's That Knocking at My Door?
Even when it seems innocuous, there may be far reaching ramifications of dirty data. For example, most genealogical databases are based on some type of public records. What if those records are incorrect? Could that affect a credit rating or even a job application? At the very least, might I end up with some unwanted pseudo-relatives at my door?
Dirty data -- data that is incorrect, incomplete or out of context -- is especially dangerous to an otherwise well thought out digital and social marketing campaign. When data is in this state it can force the marketing automation software to perform incorrectly and inefficiently.
Worse, it undermines the analytics at the heart of modern marketing and causes marketing professionals to make poor decisions. It is worse than no data, since it seems good but it's not. Not only are bad decisions made and misguided actions taken, but it isn't always obvious that that is what is happening.
Consider the effects of bad customer data on marketing automation. If the primary customer information is not correct, then all marketing efforts are wasted in a flood of bounced emails, texts send to the wrong people, and incorrect social media ad placement. Coupons end up with people not in your target market. All of that adds up to lost opportunities and waste.
The situation gets decidedly worse if good (and expensive) data is combined with the dirty data. The good data, from well managed CRM systems, for example, is now contaminated by the dirty data, rendering an analysis of it useless and any decisions based on that analysis completely wrong. Consequently, money targeted at customer engagement merely delivers annoyance and confusion.
Clean it Up
Imagine a multichannel marketing campaign for a new type of soap. The company has developed a soap for men that removes the worst grease, is not full of perfume and doesn't need water. It's marketed primarily to hobbyist mechanics, do it yourselfers and outdoorsmen.
The company combines data from current customers and those who have visited the website. This data is combined with purchased demographic data, an email list and social media data from a different source, and analyzed to create profiles of potential customers. Our intrepid marketing pros have set up a series of ads and coupons to be distributed through social media, websites and email targeted at this profile.
Unfortunately, the social media data is bad and contains wrong social media handles, many of the emails are old and non-functional, and the demographic information is incomplete. The profile's model has holes and plain wrong contact information. The marketing automation software does what it is supposed to and happily sends emails to wrong or inactive addresses, places ads in front of 14-year-old female Justin Bieber fans, and sends half the coupons expected. It's a disaster.
There are ways to avoid this problem. The good news is that many of the social and digital marketing products on the market can detect bad emails and handles and alert the people managing the campaign. It also helps to have a reputable information provider that will take financial responsibility when the information is not up to expected quality.
Another good practice is to test the data with a small version of the campaign and see if an acceptable number of responses is reached. Finally, there are software products that analyze and detect bad data. IBM DataWorks has a data cleansing capability as does Informatica Data Quality.
In the end, the only effective way to have a successful data-driven marketing campaign is to be on the alert for dirty data. It is becoming fashionable to trust data over instincts. When there is complete control over the data used, perhaps -- just perhaps -- that may be valid. Otherwise, know that dirty data is out there, and remember to not only trust, but verify. Just don't trust too much.