Machine learning is used in Lead Intelligence scenario. Using it, the scores for open leads are calculated and displayed to users along with several other pieces of information on a side panel.
How do I use it in my work?
Identify open leads which have a high potential of conversion, with machine learning. The score would indicate the likelihood of conversion of lead to opportunity. Sorting the leads in descending order for the score, the leads could be ranked in the order of likeliness of conversion. Such leads can be pursued proactively by the sales team for further processing.
Thus, lead intelligence supports the sales team in improving the lead conversion key figures.
Lead score is based on a range from 0-100. There are three buckets for the scores – Very likely to close, likely to close and less likely to close
Currently the legend is defined as follows:
0-50 = RED = LESS LIKELY
51-75 = YELLOW = LIKELY
75-99 = GREEN = VERY LIKELY
Later it will be dynamic based on customer data.
There are various other pieces of information made available on the side panel for the lead which would be of interest when assessing the lead. For example: Activity engagement which shows number of emails, phone calls, tasks etc executed for the lead. It also shows how many days the lead has been in the status it has been assigned currently. The side panel also shows the last status the lead was in and gives an indication if the current status was a positive development or negative.
These details shown on the side panel are out of the box and currently cannot be changed.
How do I configure it to use?
Configuration steps are same as for Opportunities etc. The only change is that the model is created for Lead Scoring scenario.
You need to have an Enterprise Edition license for the C4C product.
Since the machine learning uses the historical data from leads to train, it helps to have a year worth of leads in the system. Hence, if you are in the process of implementing C4C then wait for a few months and then activate the service. Around 5000 or more leads would be a minimum starting point.
It also helps to have the leads maintained with data entered in as many standard fields as possible.
The data should also be evenly distributed in the dataset. For example, exhaustive status schema with statuses documenting all possible lifecycle statuses for the lead and similar number of leads in each status. Hence, if lead is entered in the system only as a formality when they are accepted then the model will not be able to predict for the other statuses of the lead.
Process of Implementation:
The steps for implementing it are identical for all delivered machine learning scenarios in C4C. Following are details - courtesy blog by Ralf Kammerer:
Check the customer data for the points detailed in the prerequisites section above.
Currently these checks are manual. In future releases, there would be reports provided to help you check the readiness of the system by yourself for those scenarios.
Activate Lead Scoring
You need to decide which tenant to use to train the model and which to activate it in. Since it is important to have real time leads’ data for training the model, it is important to do it on the tenant which has the most accurate data.
The recommendation is to activate it directly in the production tenant since there is usually the most accurate data. Since the feature is only a ‘read’ feature – there is no risk of harming production data. Also, the feature can only be made available to some test users at the beginning.
Usually there is no sufficient data in the test tenant, so it makes no sense to create the prediction model for testing there.
Copy of production tenant
Another option is to test the feature in a copy of the production tenant with the ‘productive’ data records.
An additional tenant is required. It may lead to additional cost depending on your license.
Testing of the dynamics in the model is not as good as in the production tenant since there is no constant change of data.
Model has to be recreated in production and might lead to other results than in test due to different data after some time.
The feature has to be activated by SAP.
An incident should be created for component LOD-CRM-ML in the tenant where it should be activated.
Please request the activation of the Machine Learning scenario ‘Lead Scoring’.
In a future release, it would be possible to switch the feature on via scoping.
A new business role should be created in order to assign the authorization to see the lead intelligence side pane related in the lead workcenter.
The business role should be created by the customer/implementation partner and it must be specified in the incident to request the activation of deal intelligence by SAP.
SAP will enhance this role in order to show the new features on the UI.
The role will then have to be assigned to the relevant users by the customer.
This procedure will be replaced in a future release and the authorization would be granted by the Administrator just like any other authorization in C4C.
Train machine learning model
Before you can make use of the lead intelligence feature, you must create your customer specific trained model which is used to predict the scores for your leads. To train your custom model go to the admin workcenter and choose the view ‘prediction services’.
Select the Lead scoring scenario from the list of scenarios available and use action Add Model ( This list of the scenarios is delivered out of the box for machine learning and with new releases in future more scenarios will be added.) to create a new model.
Once created use the action Train to train the model. It may take some time for this activity to complete depending on the volume of the leads in the system. The historical lead data is collected and sent to the Leonardo machine learning component and your customer specific model is trained/created based on your historic data. Only the open leads are scored.
Once the training is successfully completed validate model training results. The SAP Leonardo component returns an accuracy which tells you how good your model is after the training.
Generally, a model with an accuracy below 50 is not good enough to have a satisfying prediction result. However, evaluate this in your specific project scenario.
It is necessary to activate the model once the validation has been done and the accuracy is acceptable to for your scenario.
Once the model is activated, all the leads in open status get a score. The first scoring is done in batch and after that updates are done daily after the end of the business day mostly.New leads created after the activation are scored in the daily update run. Check the scoring updates the next day for the leads changed on the previous day.
For the first test, you can validate if the score matches your own expectation for a lead.
As a second step, you can change leads and see if the scores update (which happens only overnight) matches your expectation. Currently it is not possible to change a lead and see the effect on the score prediction right away.
Once testing is complete you can make the feature available for all relevant users.
If you did the model training etc in the productive tenant- just assign the business role for Lead Scoring which was created before, to the relevant users.
If you did the model training etc in a test tenant: Repeat steps described in section ‘How do I configure it to use?’ in your production tenant.
Only one model can be active in a tenant at a time. The active model cannot be deleted. You can create a new model and train it at any point of time. If you set it to active, it will deactivate the old model.
Also refer to the Administrator guide and User Stories for latest details on how to setup the functionality and how to use it respectively.