Notes from the Engine Room: CRM Data
Taking things for granted lately?
Food on the table, clean running water, roads to travel on?
What about clean data for your CRM system? Is this an assumption that you’ll regret having made?
And here’s one more question. Do you capture data defects and keep them from entering the system?
Where Data Defects Hide
It became rather obvious that sales to China, Chine, Cina, Kina, Chuna all meant the same country of China. It’s not because I speak French, Italian, Danish and figured out a misspelling when someone “fat-fingered” a “u” for an “i” on the keyboard. This represents just one example of one country of sale. Just think of all the countries to correct!
In addition, to country, SalesForce provides data on state, region, district, city, county, etc. As these apply to sales, they also apply to Company, Product, Contacts, Leads, etc. you easily come up with thousands of records that need mending.
Another data issue to fix occurs with two different systems for date. Is it day, month, year or month, day, year? What does 04/03/07 really mean? Is it 3 Apr 2007 or 4 Mar 2007?
Lastly, what do you do with duplicates? Are they really duplicates or because one data element is different, can they be mistakenly entered by two different people at two different times?
How they enter the system
You guessed it. People make mistakes when they’re in a hurry, on the go, using different devices, or just plain making quick assumptions to get the job done.
But there’s more. The old company determined it best to open the data feed from SAP to SalesForce using the features of both systems. This clever technique eliminated the need to employ a human to enter the data or review the flood gate. What happened became another entry for data errors.When a batch job failed and it re-ran, it sometimes created duplicate records that passed through.
How they propagate
One method of propagation of errors lies in the above example when systems transfer data. Consider another method of propagation when one piece of data feeds into another. One of the beauties of SalesForce lies in charting and graphing sums of the data, such as all the sales to China, or all the sales in Asia, or all the sales to XYZ Company. What happens to the decisions managers make when they only receive partial data?
Who enters them
It’s rather obvious that people do. Or is it management who allows the people to make the entries?Who could blame the sales rep for being in a hurry at an airport, train station, or “fat-fingering” a text on their smart phone or device? Is it reasonable to hire a data entry clerk for all system entries? What data issues does this present?
The damage they cause
The foreshadowing above suggests managers make the wrong or uninformed decisions by assuming clean data. That comes at a cost, as well as the cost for keeping the data clean. Think of the salary and benefits paid to a Data Quality Manager vs. the cost of poor business decisions. Furthermore, human behavior intervenes and says that “if I can’t trust this piece of data, then what can I trust about the rest of the system?” Now users question the validity of the costly system you spent millions of dollars, euros, rupees to implement and maintain.
How to prevent them from happening
SalesForce uses “pick lists” as one of the methods of controlling data. If the only given spelling of China in the pick list is China, all the data becomes clean. This solves a lot of issues and funnels key data into one field for the summations for management decisions mentioned above. This improves the quality of the data for key performance indicators. However, sales reps need to enter data on Contacts and Leads. Having a Data Administrator to create pick lists, customize the system with changing data needs, and show users the data that needs correction becomes a cost of maintaining the system. The discipline of the organization makes this much easier and speaks to the governance by management.
Lastly, current technology makes the life of the Data Administrator much easier with computers and conditional logic. Consider using the computer to run through the data for missing data elements, or data not contained in the pick list and reporting them. The Data Administrator then can go back to the source and make changes to pick lists to keep the system current. This is much easier than looking for the needle in the haystack.