What is Revenue Science?

The deluge of customer data available to revenue teams presents enormous opportunities. However, traditional machine learning methods fail to support business decision making for two reasons: They are focused on prediction vs. optimization, and they typically rely on understandings based on correlation rather than causation.

Revenue Science is a quantum leap in machine learning over customer data. It relies on two breakthrough methods in data science: Causal Machine Learning and Prescriptive Analytics.

What is Revenue Science?

The deluge of customer data available to revenue teams presents enormous opportunities. However, traditional machine learning methods fail to support business decision making for two reasons: they are focused on prediction vs. optimization, and they typically rely on understandings based on correlation rather than causation.

Revenue Science is a quantum leap in machine learning over customer data. It relies on two breakthrough methods in data science: Causal Machine Learning and Prescriptive Analytics.

Machine Learning Must Do More

Fact: Existing analytics and machine learning solutions are NOT designed to support decision-making. 

They are descriptive – providing an understanding of what happened

They are predictive – estimating what will happen. 

Descriptive and predictive are not enough. What is needed is an equally precise understanding of how marketing and sales actions can be optimized to deliver results. This is the essence of Prescriptive Analytics and lies at the heart of Revenue Science.

Improving Outcomes vs. Simply Predicting
What Will Happen Next

A key aspect of Revenue Science is its focus on outcomes. What is an Outcome and how does it differ from a Prediction?

What are Predictions?

A prediction is an estimation of a future KPI, under the assumption that a company’s sales and marketing activities going forward will be similar to the past. This is how most data science works. 

What are Outcomes?

An Outcome is an estimation of how a KPI will change, given changes in marketing and sales strategy. Outcomes can be optimized, predictions cannot.

Using predictive methods to improve outcomes can lead to suboptimal results. For example, many enterprises manage churn by building a predictive model to calculate customers’ propensity to churn. Those customers that have a high chance of churning are then targeted for outreach. The challenge is that companies do not want to predict churn, they want to improve retention, and those customers that have a high probability to churn are not the same accounts that are likely to change their behavior because you sent an offer.

What Should We Do Next?

See how causal machine learning & prescriptive analytics will give you better retention results. (Hover over the images below to see the answers)

Customer A

50%
CHURN PROBABILITY

Customer A

Potential to reduce to 20% churn
Great target for retention programs

Customer B

90%
CHURN PROBABILITY

Customer B

Potential to reduce to 85% churn
Poor target for retention programs

Outcome-based approaches result in a 3x improvement in retention versus “predict then act” methods

Causal Machine Learning

Understand the cause of any outcome

The applications of causal machine learning on customer data are broad and can be employed to drive revenue growth and operational efficiency across marketing, sales and product experience.

How do we know if a coupon increased the buyer’s propensity to buy? Maybe they wanted to buy anyway. Causal Machine Learning can tell the difference.

Optimizing KPIs requires an understanding of causality – of how sales and marketing actions affect business outcomes. Most machine learning algorithms are designed to identify correlations, not causation, and can lead to biased or even misleading estimates of causal effects.

Paid Search – Increased Budget May NOT Lead to Increased Revenue​

A classic example is paid search advertising – as the amount of spend on paid search increases, so does the amount of revenue attributed to traffic from the placements. However, in many cases this traffic would have converted anyway via organic listings, resulting in an inflated understanding of ROI.

Causal Machine Learning is a class of methods that are specifically designed to leverage data to reliably establish a relationship between cause and effect. In the paid search example, it helps you understand to what extent advertising causes traffic vs. a situation where people who are already likely to buy are also clicking on your ad.

Revenue Science utilizes Causal Machine Learning to go beyond predictions to augment human decision-making and drive profitable growth.

How do we know if a coupon increased the buyer’s propensity to buy? Maybe they wanted to buy anyway. Causal Machine Learning can tell the difference.

Optimizing KPIs requires an understanding of causality – of how sales and marketing actions affect business outcomes. Most machine learning algorithms are designed to identify correlations, not causation, and can lead to biased or even misleading estimates of causal effects.

Paid Search – Increased Budget May NOT Lead to Increased Revenue​

A classic example is paid search advertising – as the amount of spend on paid search increases, so does the amount of revenue attributed to traffic from the placements. However, in many cases this traffic would have converted anyway via organic listings, resulting in an inflated understanding of ROI.

Causal Machine Learning is a class of methods that are specifically designed to leverage data to reliably establish a relationship between cause and effect. In the paid search example, it helps you understand to what extent advertising causes traffic vs. a situation where people who are already likely to buy are also clicking on your ad.

Revenue Science utilizes Causal Machine Learning to go beyond predictions to augment human decision-making and drive profitable growth.

Prescriptive Analytics

What Should We Do Next?

Once the effects of actions have been reliably estimated, it is then possible to implement a Prescriptive model that can determine what the ideal course of action is for a particular customer scenario. 

One of the best applications of prescriptive analytics is Uplift Modeling. For example in direct marketing, estimating the causal impact of treatment at the individual customer level allows optimal target audience selection: Target only people for which the predicted effect is positive. In the case of churn, this translates into only targeting those customers who will respond favorably to an offer; or in the case of sales outreach, only contacting prospects whose likelihood of buying increases because of the outreach. A similar approach can be applied to prescribe optimal actions for a wide variety of customer scenarios.

By enabling prescriptive analytics with causal machine learning, you can activate actions to meet your company’s growth goals

image-05-05

Schedule your Demo today!

Schedule your Demo today!