18 Jun Ai Can’t Outsmart Fake CRM Data
CRM data is made up primarily of data entered by humans. Of that data, much of it is entered by sales professionals. Sales people have incentive compensation plans that are related to the efforts they are reporting data on in the CRM. These compensation plans naturally create bias and, in many cases, the resulting CRM data reported by sales people is in turn very biased; some might even call it fake data. Because of this, companies who are experimenting with Ai related initiatives from data stored in their CRM can often experience unexpected results. Let’s dig in further.
Sales people have quotas and commission plans. These two are inextricably linked and inform how a sales professional behaves, how they decide to share data and insights, and how they expose their sales activities and outcomes.
Consequently, CRM has two key resulting challenges; incomplete data, and inaccurate data. Both of these conditions starve machine learning models of what they need to see proper patterns.
- Incomplete data is related to sales people deciding to withhold activity or interaction/engagement data from the CRM system. This happens when the sales person believes entering data is a waste of their time or that sharing data could be risky to their personal objectives such as retiring their sales quota or achieving their personal commission objectives. If taking the time to share this data will be favorable to their objectives, they will share the data. This almost always occurs because of a positive sales outcome. Even in these situations, there can be incomplete data related to cycle timing and sequencing data associated with the sales professional’s sales pursuit. Most often this occurs because the data being shared is only shared after much time has passed or key events have taken place.
- Inaccurate “fake” data has two forms; intentional and unintentional. Unintentionally inaccurate data usually occurs when data entry compliance is audited in some way and sales reps are alerted to their lapse in data entry. They then scramble and enter data in a back-dated fashion, trying to remember or transpose cryptic notes they may have written down or attempting to do it from memory. Often this data entry approach leaves out critical data and usually includes positive bias in terms of contextual insights that surround the back-dated entry of data. Intentionally inaccurate data occurs when sales reps deliberately input vague or misleading data because they want to shield themselves from any negative consequence associated with sharing data that may indicate poor performance on their part.
In these cases, Ai and associated machine learning models will not have fact-based data to learn from. For example, if a machine learning model only has contextually positive outcome data to learn from, the model will produce unexpectedly positive results because it cannot take into account all the incomplete or inaccurate data to balance its interpretation of the overall data set.
So, does that mean that Ai projects within the CRM world are doomed? Not so fast. We believe that Ai should be used in pragmatic ways to help not only correct the fake data dilemma, but to also drive powerful augmentation of customer-facing professionals including sales and customer experience representatives.
It’s a well-known fact – sales people hate having to enter data as a part of their sales pursuits. They don’t enter data about their activities, and they don’t enter actionable data memorializing their conversations and engagements with customers and prospects. We believe that using Ai to solve these challenges is the starting point for the next generation of professional selling to take place. An example of pragmatic use of Ai and machine learning:
- Minimize data entry first, by using NLP machine learning models and Ai to collect conversations and sift through them for the data that Reps hate to document or share in the first place.
- Keywords and key phrases
- Action items
- Sentiment and context
- Activity meta data
- Outreach activity tracking
- Engagement measures and metrics
Ai and Machine Learning can be put to work solving the first order problem of CRM: inaccurate and incomplete data. Then downstream from these first order data collection Ai initiatives, CRM becomes invaluable to your predictive and prescriptive Ai initiatives – those that have been envisioned and those that haven’t even been thought of yet.