Hello everyone.
Last time, we introduced how lead scoring can "predict when customers will buy" and "resolve resource shortages in the sales field."
This time, I would like to talk about lead scoring and synchronization with the sales department.

What does your sales team expect from lead scoring?

The sales department will want to know the number of inquiries from exhibitions, direct mail, website inquiries, etc., but they will also want to know which inquiries (projects) have a high probability of success.
When scoring leads, it is becoming increasingly important to build a predictive model that calculates probability and accuracy based on various data obtained from inquiries and past performance, etc.
Marketers have countless worries, such as predictive model algorithms, data normalization, etc.

Integration with Azure Machine Learning

I don't think it's easy to build a predictive model.
For this reason, machine learning has become an available tool.
For example,Microsoft Azure Machine Learning
Azure Machine Learning is a powerful cloud-based predictive analytics service that enables you to rapidly create and deploy predictive models as analytics solutions. Azure Machine Learning not only provides predictive analytics tools for your models, but also enables you to deploy predictive models as ready-to-use web services. Azure Machine Learning also enables you to rapidly create, test, operationalize, and manage predictive models in the cloud, without having to purchase hardware or manually manage virtual machines.

Trying out Azure Machine Learning

We will create a model on Azure Machine Learning that predicts whether the winning prediction for an inquiry is correct.
The conclusion of the forecast is that if the company's financial situation shows a good year-on-year profit growth rate, there is a high possibility of winning the inquiry.

1. Prepare past data
I imagine that the results of the inquiry have something to do with WinLoss and the company's financial situation.

2. Select the verification task and the items to analyze
Add a validation task to Azure Machine Learning and set the items you want to use for analysis.

3. Create a model flow.
The data is ready for the validation task.
Connect each one with a flow.

4. Prepare the validation data you want to predict.
Although the results of the win are predictable, we have prepared data that resembles this, hoping that it can be derived from past data. Because the number is small, the results may be extreme.

5. Test the model and check the training data and prediction data.

The results of the validation task show that the results are weighted heavily in terms of year-over-year profit growth.
A possible scenario is that, based on past performance, if the year-on-year profit growth rate is greater than 100, there is a high possibility of a win.

From the verification data we want to predict, we can see that the data for which the "Person in Charge" is "Tokumoto" and the year-on-year profit growth rate is positive has a high predicted value of 0.88. Conversely, the data for which the "Person in Charge" is "Komiya" and the year-on-year profit growth rate is negative has a low predicted value of -0.09 (negative values are likely due to a lack of sample data).

This was an extreme example, but to increase the amount of data or to evaluate from multiple angles, we need to look at it from different perspectives, but we will stop here for the sample.

Summary

Dynamics CRM will also be offering a standard feature that works with Azure Machine Learning in the future. We created a model on Azure Machine Learning to predict wins in inquiries, and the provided feature will enable users to link it with Dynamics CRM data and check the predicted values on screen.
When salespeople can use these predictions as a guide, it can be said that lead scoring is synchronized with the sales department.
I look forward to seeing how CRM and machine learning are linked together in the future, and I'll be keeping a close eye on this.