As most of us know, Superman was severely weakened by exposure to kryptonite and could not perform his normal super hero feats. In Defending the Caveman, Rob Becker wrote that a woman asking a man to go shopping at the mall was a form of kryptonite for most men. In both cases, removing kryptonite resulted in immediate recovery.
Are you applying mysterious, green deficiencies into your lead acquisition efforts?
Here are a few "kryptonite" factors that can derail the application of AI analytical tools.
Lack of individual data
Many marketers continue to rely on readily available data such as geodemographics (where the consumer lives) or household data to define the appropriate personalized response to an individual consumer. Referring back to Defending the Caveman, men and women rarely respond to stimuli in the same way.
If you are not collecting data points for the individual consumer through a first or third-party source, you need to start envisioning what transaction or survey responses can I input to my analytical views.
Hail household case study
My wife and I have been married 48 years in 2019. We often complete sentences for each other—we are a family unit. Yet, our shopping behavior is not synchronized. In almost every circumstance when she says “yes,” I will respond/buy negatively. She loves to shop for clothing, household goods. I love to shop for cars and travel. I answer surveys; she will not answer the phone or your email. We are two unique shoppers.
If you apply only household data, our address or home IP address only, then you will find yourself in the traditional advertising saga — half of my advertising is wasted. I just do not know which half — to Mike or to Judy.
Due to our shift in digital marketing, many of us have decided that direct mail methods are no longer necessary.
I will contest that stance. If you standardize using a CASS software and then apply NCOA, PCOA and various suppression files, you will discover that your ability to correctly integrate data enhances.
In almost every case, we study: 10-20 percent of the normal prospect data being utilized is not correct. Thus, you have mistargeted—Mike Hail is not receiving the offer designed for him. In certain cases, like loyalty files, the percent soars into the 30-40 percent range.
Here is a short list of most observed discrepancies:
Duplicate records — Both hyphenated and foreign names lead the way here
Old or not current address — NCOA is not capturing half of the movers
Address is not standardized — I run CASS on every file that possesses a name and mail address—Best Practice!
Undesirable prospect—suppression condition: death, nursing home, prison
Most of us were trained to apply some form of match code to merge/purge records to integrate our consumer data. In my case, I have worked with several defined record connectors —Social Security Number, Drivers License Number, Telephone number.
In the omnichannel marketing environment today, the identity resolution tools must recognize online and offline feeds—postal, email, Google, and Facebook. After all, the consumer is expecting, perhaps demanding, a personalized response.
A traditional match code processes fail us. Instead, we use a Persistent Identification Number process. At DDG, we generate a PIN for the individual, the household, and for the address. In addition, we can include an email, a phone number, and date of birth into the PIN creation. All of this allows us:
To identify the buying unit that matches our product or service — individual, household, address
To personalize our reply for that enhanced customer experience
To integrate online and offline data — one account or multiple accounts, how do you view the total consumption of a household, what is the total annual spend for this person?
To respond in real time -- immediate application of a score and the correct segmentation message
My astute colleagues at DDG will write more about Best practices for Sampling, errors to avoid in statistical processes. But as a data dog for 40 years I wanted to note the data related mistakes I witness that act like Kryptonite and make even the strongest AI/machine learning fail.