2969_SM_SalesCompensation_blog (1)-1PKS Prakash and Arun Shastri co-wrote this blog post with Chad Albrecht.

In our previous blogs, we discussed how AI and machine learning can impact sales compensation plan design. Now, let’s consider the three ways that machine learning can potentially impact quota setting.

 1. Accuracy: When it comes to quota setting, what does it mean to be accurate? 

  • Have outliers been appropriately weighed?
  • Have systematic biases that could lead to under-forecasting or over-forecasting been eliminated?
  • What degree of accuracy do we desire? (Does a greater level of accuracy at a national level guarantee a higher level of accuracy at a sales territory level?)
  • Does the forecast consider the stage of product lifecycle (Can we reach the same level of accuracy for a newly-launched product as a mature product?)

For each of the questions raised above, traditional methods when combined with machine learning produce increasingly better answers, which translates into more accurate territory-level quotas. While better accuracy is always a good outcome, companies should only consider factors outside of a salesperson’s effort and ability in generating the forecast (territory potential, for example). Otherwise, top performers may be unfairly burdened by the machine learning model.

 2. Agility: There are two types of agility: speed with which quotas can be generated, and the ability to sense and quickly respond to changes in process or data during the year. Leveraging machine learning accomplishes both through:

  • Improved data quality and management: State-of-the-art approaches stitch together various data (for example, automating data lakes).
  • Technology: Data ingestion, modeling, dissemination and consumption can be automated if they’re supported by the right technology.
  • Algorithms: Model management can ensure that algorithms run correctly, the QC process is always on and metrics and SLAs are consistently delivered.

The machine learning approach helps organizations improve accuracy and minimize human effort involved in optimizing quota setting. Appropriate time and energy can then be devoted to communication of these goals and change management. The system can also measure the health of quotas in real time by monitoring any events that affect forecasts, such as a new product launch. Organizations also should be open to creating shorter term quotas to adjust to market realities, or change quotas based on performance in the marketplace. Even if updating quotas more frequently is warranted, which is not always the case, the sales organization may resent it (If I’m killing it, they’re going to raise my quota, whereas if my coworker is dogging it, his “penalty’” is a reduced quota, the salesperson thinks.)

  1. Understandability: With sophisticated algorithms, more data, and enhanced computing power, machine learning models are increasingly more accurate in setting quotas. However, there’s an inherent tradeoff between accuracy and understandability - the greater the accuracy, the greater the likelihood that the model is unable to tell you why. However, great strides are being made to enhance the interpretability of these machine learning models. If you have a greater sense for why the quotas are what they are, should you therefore pay on actions? Organizations must communicate insights generated as part of the quota setting process to the sales organization in order to remove the “black box” perception associated with quota setting. This will in turn enhance adoption.

 

Advances in machine learning have impacted goal setting in meaningful ways, and organizations must step up to take advantage of them.


RELATED CONTENT 

BLOG POST: Artificial Intelligence's Impact on Sales and Sales Compensation

BLOG POST: How AI Will (Or Won't) Shake Up Sales Comp 


 

 

Topics: sales compensation, sales compensation programs, artificial intelligence, machine learning