Prescriptive Action

The Drivetrain
Approach.

Instead of simply forecasting the future, actively shape it. With the Drivetrain framework, your Data Products are designed to maximize key business drivers rather than providing passive observations.

View the Framework Why it Matters

The Drivetrain Approach for Data Products

The Four Steps to Actionable Data

A common reason for the failure of data initiatives is that they begin by focusing on data and then try to find a problem to solve. In contrast, the Drivetrain Approach starts by defining the objective first.

Define Objective

What specific business result are we aiming for? It must be quantifiable.

Example: Maximize long-term revenue from existing customers.

Identify Levers

What factors are within the business's direct control to influence the objective?

Example: Which promotional offer to send, or which channel to use.

Collect Data

What information connects the levers to our desired objective?

Example: Historical purchase history, demographic data, past campaign responses.

Build Models

What is the statistical relationship between the levers and data that influences the objective?

Example: A machine learning algorithm that forecasts the likelihood of a user converting with Offer A compared to Offer B.

The Final Output

Mathematical Optimization

The last phase of the Drivetrain involves utilizing predictive models and executing an optimization algorithm to recommend the best course of action. exact lever settings The data product provides clear instructions to help you achieve your goals.

Why Predictive Models Aren't Enough

Picture creating a predictive model that accurately identifies customers with a 90% likelihood of churning. This model is not only highly accurate but also mathematically robust. But it's fundamentally useless on its own.

The reason is that providing a 10% discount to a customer who is 90% likely to leave may not influence their decision, but rather result in a loss of profit. Instead of focusing on predicting customer churn, the Drivetrain approach concentrates on identifying the most effective strategy (discount, phone call, feature unlock) to increase the chances of retaining a particular user.

From Output to Outcome

Designing Data Products should start with the desired business outcome, rather than being based on the existing data.

The Flawed Approach

We possess a large amount of customer data. Our plan is to create a neural network for behavior prediction and integrate it into a dashboard for the marketing team to analyze and utilize.