Predictive analytics shows the company’s development and provides opportunities for continuous learning and improvement.
Predictive analytics shows the company’s development and provides opportunities for continuous learning and improvement. If a business decides to “play by the numbers” then predictive analytics is the ultimate cheat sheet. Especially if your business model is built around repeatability. This way you are already measuring metrics across the entire customer lifecycle from acquisition to repeat usage and revenue. By using predictive analytics, you can lower your cost of acquiring users and helping you with ensuring sticky customers, and increasing your revenue.
The Business Insights screen purpose is to allow the decision makers to “play by the number” and understand what is the potential of the different strategies of predictions and their impact on their customers while utilizing their different customer engagement toolkit.
The numbers reflect the prediction as if it ran 30 days before the actual prediction date. It allows the system to hide a month of data and then validate and estimate the predictions’ accuracy and impact on the business.
“List to rank” - 1,148,312 people purchased at least 5 products in the last 30 days. (counting from today - 30 days)
“Examples” - 1,372 people purchased at least 5 products in the last 30 days and then purchased more than 20 products of this type during the past 30 days
The prediction ran on this date: 31 July 2018
Business properties provides the end use to play with the different engagement tools (using the engagement cost filed) and the conversion revenue, in order to be able to estimate the ROI of the prediction and its potential effect. In this case the focus is set to: Target top: 5% which is Population: 57,410 out of the total 1,148,312 people | The engagement cost is : $2 | The conversion revenue: $900
By utilizing the numbers which are set in the business properties, we can calculate the estimated ROI based on the accuracy of the prediction, marked with Endor AI Engine: ROI = (true positives * conversion revenue) - (population to target * engagement cost)
As there is no model that we initially measure the prediction quality by, the calculation is based on the baseline of the random guess which takes the total true positives and divides it by the population to rank (or a subset of the list to rank, by the same percentage that is defined in the business properties…Making it easy to understand what you would have got without any model)
The end user can switch the toggle to “With AI”, by doing that he will be able to input the performance of his internal model and measure it vs the Endor AI engine performance.
On the right side, the system calculates the baseline, taking all the people that behaved in the desired manner in the hidden 30 days from the system, showing the prediction difficulty. The lower the baseline, the harder to target the right audience without utilizing any predictive model. In addition to the baseline the system shows what would be the cost of engagement if no model is used.
Download the full report as CSV, will return the report of all the “list to rank” and their probability of behaving in the desired manner.