cf6e74f923461f259f8545a45b947aa7b1c66503

4 Simple Ways to Evaluate if your Scoring Model Works

Blog Post created by cf6e74f923461f259f8545a45b947aa7b1c66503 on Dec 9, 2015

After building a scoring model in Marketo and letting it run for a few months, you will want to make sure it is working properly so that you know that people are getting points for the right activities. We'd then recommend revisiting this every six months or so. Here are four fairly straightforward, if sometimes time-consuming, things you can do to validate your scoring model.

 

1. Make Sure All Campaigns Are Running
The first step is to simply scan your campaigns and make sure that they are all processing records.  You can know whether the campaign is processing data if there are members of the campaign in the campaign overview. 

 

2. Evaluate Your Scoring Footprint

The scoring footprint looks at the impact that your scoring model has on your database.  In order to understand your scoring footprint, follow these steps:

    1. Create a smart list: lead score is greater than 0
    2. Look at the number of records that has been pulled (e.g. 52,041)
    3. Next, go to the lead database and use the pie chart to get the number of known records in the database (e.g. 103,766)
    4. Use a calculator (http://www.percentagecalculator.net/) to calculate what percent of your database has a positive score (e.g. 50.2%)

 

These numbers should be taken every 6 months and kept in Excel.  Over time, you should see a trending increase in the percentage.  If you have a trending decrease, look into what the cause of the decrease is.

Repeat these steps for the behavior score and demographic score as well.

 

3. Evaluate High Value Attributes and Activities

Use the following steps to look into which attributes or activities are most closely associated with buying behavior:

    1. Create a smart list in the Lead Database called “Became an MQL,” use the filter Revenue Stage was Changed to MQL
    2. Create a Leads by Campaign report (assuming you still have these in your reports, they're a bit of a legacy report) locally within the scoring program, the report should have the following attributes:
    3. Under campaigns, select the scoring model program
    4. Lead created at: All Time
    5. Custom Columns: Became an MQL

For each scoring campaign, this report shows the number of members in the campaign and how many of those members became an MQL.Use Excel to calculate the percentages.  Next, evaluate which activities / attributes are most notable.

 

In this example, notice that Searched for a Keyword has the highest rate of bringing in qualified leads, so we would want to make sure that this campaign has a high number of points and may want to look into investing in SEO to increase keyword searches.  Adjust your model to reflect your findings.

 

4. Check Your Values

For demographic scoring, we can use Marketo to understand which values have the highest indication of buying behavior. For each demographic campaign, do the following:

Create a smart list, in this example we will use job title.  The smart list should use the attributes Job Title is not empty and Has Opportunity is true.

Export the list into Excel

Copy all of the job titles from the list in Excel

Paste them into some tool like Wordle (http://www.wordle.net/), which is a third party tool that creates a visual diagram for which values appear most frequently in a list.

 

It's important to not that statistically, you also need to compare how frequently those values appear in your database as a whole. For example, if 50% of your opportunities are created by Managers, but that's also 50% of your overall database, this doesn't tell us much. If however, your database has 20% managers, then there's something more interesting there.

Outcomes