Skip to main content

Symon.AI help center

Lead Scoring

Abstract

Use data on past wins and losses to determine the likelihood of a new lead becoming a win or a loss.

Knowing which leads are likely to close is an AI use case with high potential. By using data on past wins and losses, Symon.AI can determine the likelihood of a new lead becoming a win or a loss.

Legend

The tools in this blueprint are color-coded based on their function:

  • Yellow: These tools contain the input data or results that are ready for export. These are what you’ll be mapping to in Incentives for the Symon.AI Calc object. The Lead Scores (Data) tool contains the incoming data on previous leads.

  • Pink: These tools sort and provide the results for the exports.

  • Purple: The Predictor tool is a Super Tool in Symon.AI. In this blueprint, it predicts the win/loss rate.

  • Red: This tool prepares the data for the predictor.

  • Turquoise: You can change these tools to fit your data into the blueprint or the output you want.

  • Lime Green: These are the output tools which contain results that are ready for export.

Pipe inputs

This blueprint uses the Adapt tool to map the columns in your data. The required columns are:

  • Prospect: Product information.

  • Converted: Identifies Yes or No or Open data. This is the target for the Predictor tool.

You can look at the sample data to see how the tool maps columns. Whatever column names you use, you can use the Adapt tool to change them to names the pipe recognizes. In the sample, the remapping looks like this:

  • ProspectID to Prospect

If you have a different column name you want to use for conversion data, for example "Won" you should change it like this:

  • Won to Converted

How to interpret the results

In this Blueprint, to generate results, you must export the data. Go to the Exports page and create a new export. For more information on exports, see Exporting data . We'll explain this in the How it Works section. It’s important to read the information in that section since we’ll be training a model to predict values.

Exports

There are two outputs that this blueprint generates:

  1. Ranked Leads List (Export): This output contains the ranked lists of leads likely to convert and requires the export.

  2. Lead Count by Last Activity (Export): This output gives us the amount of leads based on Last Notable Activity.

Using these lists, you can make visualizations or begin making calls from the top of the list down.

How it works

This Blueprint uses historical conversion data to determine the likelihood of open leads to convert. This Blueprint requires the use of an export and will be described in this section.

If you have already built and trained your model, skip directly to the Export section.

Build - Training the model

The sample data contains information on the lead, a ProspectID and a conversion result. In this example, the conversion result includes Yes, No and Open. Due to the conversions provided by the sample data, we remove those that are currently in Open status. We do this by using the Filter out rows that are still Open (Filter) tool to filter rows that are still Open. This allows to train the model since we want to predict based on historical results.

This Blueprint uses the Predictor tool in this data set to build and train the model. But before we make any predictions, we must exclude some columns. For the sample data, we exclude the ProspectID, and Status columns. We don't want Symon.AI to train the model using these columns because they could be randomly-generated or they can be built chronologically. To put it another way, we're confident these columns are not factors in whether the deal was a loss or a win. We've also excluded the Status column. Although it seems like this column gives us a lot of information on the data, it's actually a proxy for the results. If you look at the example, you'll notice leads where Converted is No, the Status is updated to Lost to Others, while for Yes the Status is updated to Won. Having this proxy for the results actually skews the results. In data science, we call this overfitting. Overfitting occurs when we train a model with known results. Later, when we introduce unseen (new) data, we end up with a less accurate model. So it's important we exclude this information from the Predictor.

We call this type of work Feature Engineering. This is any work we do to transform our data to better fit the problem we're trying to solve. This makes the model better at predicting results for unseen (new) data.

Tip

You can test this yourself by removing Status from the Exclude Column list and press Update. Notice how the accuracy score jumps to 100%. The only "important" column is Status. These types of results generally mean there's a problem with the model.

Now that we've done some Feature Engineering for our Predictor tool, here are our next steps:

  • Use the Probability Filter (Filter) tool to choose the prediction we want to look at (Yes) and the likelihood threshold we're interested in (0.75).

  • Sort the results to see the leads with the highest probability of converting.

We'll use the remaining branches to get the exports we want.

Train the model

Now that we have built the pipe we want, we can train the model. Press the Update button and Symon.AI outputs the results. The sample data has an accuracy score of 83%. Now that we've trained the model, we can use what the model learned from the historical data to predict on new data. Symon.AI will maintain what it has learned going forward.

Export: Get actionable results

New data can come in two forms:

  1. It comes in the same file: Like our Sample Data, this is one data set where the open leads are included in the same file as closed leads.

  2. It comes in a separate file: Leads that we would like to predict on are found in another file with the same columns and information as the trained data.

Export: Using the same file

  1. To get results, go to the Export page and click Create export. In the Export section we need to deselect Filter out rows that are still Open (Filter). To do so, click on the tool in the Export view and press Deselect once it appears. For this Export, we've Deselected that tool to allow Symon.AI to predict on the new data. Once Symon.AI completes the new export, you can download the data from the Export tools.

  2. Click Create export.

  3. Click Export now.

  4. Search for and select the Lead scoring pipe. Click Next.

  5. Optionally add an Export name. Click Export now.

Export: Using new files

  • If the data you would like to predict on is in a separate data set from the one you used to train your data, first ensure that the two data sets have the same columns originally used to train the model. Once you have done this, complete the same steps as you would for an export but select the new data import in the Export section.

    Tip

    Be aware of the amount of time since you've last trained your model. Let's say that you trained your model on January 1st, 2020. On January 5th your boss provides you with a list of leads that he would like to have scored. You ensure that the leads have the same columns you used to train the model and are now ready to score them. In this case, you can upload the list of new leads and go directly to the Export tab. You can export the data through your trained Symon.AI Blueprint and get the answer to your boss. No retraining required.

    It is now April 1st, 2020 and your boss asks for a report on a list of new leads. Once again you ensure that you have all the necessary information and go to export your data. However, it has been 3 months. Many leads have been won and lost since you last trained the model on January 1st. This new information helps Symon.AI learn more and be more accurate. Upload your data up to April 1st in the Lead Scores (Data) and press update to retrain the model. Follow this up with the export command to get results on the leads your boss wants you to score to get the most up-to-date score.