%20(1).png)
If you’ve been following along with best practices in the PLG space, you’ve probably heard of PQL Scoring. PQL Scores help product-led companies monetize their user base by identifying which accounts have the highest likelihood to convert. This is very similar to MQLs, but rather than using engagement metrics, you use product usage metrics to build a score.
We wrote an intro article that describes the difference between a PQL vs MQL, so that’s a good starting place if you’re not familiar, or you can dive into the top 3 PQL examples that are commonly used by product led companies. Last but not least, here’s a breakdown of why we decided to launch our machine learning PQL scoring and how companies benefit from it. But since you’re still here, let’s give away the secret sauce and let you know how our PQL scoring works at Correlated.
Benefits of PQL scoring
The state of the art PQL score today is something like the following: marketing or growth teams work with data teams to identify what they think are key indicators that someone is a good prospect using either statistical analysis or machine learning. In some cases, they may even be leveraging their own domain knowledge to formulate hypotheses about what indicators make up a PQL score.
Data teams then gather the data and surface a list of prospects in the form of a dashboard or excel spreadsheet. These lists are typically delivered (at most) weekly. GTM teams then have to parse through the list and decide what to do. The best PQL Scores convert at a higher rate than MQLs and cold outreach because prospects have gotten the chance to use your product and are showing the right buying signals. In fact, they’ve shown to convert 5x higher than MQLs.
Hopefully we’ve convinced you that PQLs are great! However, the existing implementations of PQL Scoring come with several caveats, outlined below.
Learn & share with other product-led leaders
Join our bi-weekly Product Led Revenue call that features experts from leading PLG companies. See you there!
Pitfalls of PQL scoring
Single PQL Scores measure likelihood to achieve only one goal, like conversion
However, product-led GTM motions are relevant throughout the entire lifecycle of a user’s journey through your product. One PQL Score does not fit all. So why don’t more people build PQL scores for all lifecycle stages (think conversion, expansion, and even churn)? Because the act of building out a single PQL Score is difficult, requires cross-team collaboration, and needs to be operationalized. Imagine the effort to build out multiple PQL Scores that need to be iterated on, tested, and improved?!
PQL Scores are only truly impactful if they are easy to act on
We touched on this briefly above, but PQL Scores need to be operationalized. What this means is that they need to be sent to the right teams or funneled into the right playbooks to nudge users in the right direction. Even if teams come up with ways to define PQL Scores, they still have to figure out how to act on them appropriately in a scalable way.
PQL Scores can be a blackbox if not implemented correctly
Let’s say you’ve figured out what makes up your PQL Score
- PQL Scores often require data analyst involvement. They are delivered to GTM teams as spreadsheets and reports and are difficult to operationalize and keep up to date.
- PQL Scores can be blackbox if not implemented correctly. What does it mean if someone is flagged as “hot”? What is the appropriate follow-up?
How Correlated’s PQL Scoring works
Correlated’s approach to PQL Scoring tackles the above problems, introducing a new and opinionated way of leveraging PQL Scoring to drive better business results. That’s a big claim, so how exactly does it work? We’ll walk through exactly what it takes to get going with our current beta version of our PQL Scoring product.
Step 1: Give us that data!
Our PQL Scoring Engine has the same data requirements as Correlated itself. At a bare minimum, you need:
- All your accounts in a table (or other supported data source)
- All your users in a table (or other supported data source)
- Product events (either as raw events or rolled up as metrics)
It’s also beneficial to provide firmographic data points about your accounts and users in the above tables, or in any additional tables you want to layer on.
You also have to make sure you are including the data that represents the “goal” you’re trying to predict. Here are some examples of what I mean:
- Conversion Goal: this could be sign up for paid events or a data point that tags paid users
- Expansion Goal: this could be an upgrade event or a data point that tags expanded users
- Activation Goal: this could be a combination of product events (like used Feature A 10 times and Feature B 1 time)
- Retention Goal: this could be represented by a lack of events over a period time (no sign-ins in the last 30 days) or tagged as a data point when users downgrade
Now that you’ve got your data in order, we can move on! Note that Correlated makes it really easy to add in as many data sources as you want over time, so you can keep adding data as you collect new data types.
Step 2: Tell us more about your customer lifecycle
Before Correlated can help you score PQLs, you have to tell us a bit more about your customer lifecycle. What makes Correlated’s scoring model really unique is that you can build models for each stage in your customer lifecycle trained on your own data. This means that you can have a score for Conversion and Expansion, and those scores and how they are defined are unique to your business!
Correlated makes it very easy to set this up with a dropdown interface that allows you to define the “goal” you want accounts or users to achieve.
For example, if you want to predict which Accounts convert, your goal might be Account Type = paid. If you want to predict users who expand, your goal might be users who clicked “Upgrade Plan”. Once you tell us how you define your customer lifecycle stages, you’re all set!
Step 3: Take a break! Correlated will generate PQL Scores. Is it magic? No, but it is machine learning.
Now that you’ve told us what you care about, Correlated will do the rest. To peel back the covers a bit on what’s going on, here are the steps that Correlated does that your data team doesn’t need to do anymore!
Preparing Data:
Machine learning models accept very specific data formats and perform better when data is prepped. One common example is turning product events into a format machine learning models can understand. Correlated automatically does this, and this is just one example of the many ways Correlated preps data to ensure the model gets clean data to train on.
Building features:
After data is prepared, it needs to be turned into “features”. Essentially, features are tagged to entities in the model and are used to predict outcomes. Correlated does all the work of turning literally all the data you pipe into Correlated into features.
Labeling Training Data:
Since we’re trying to predict an outcome (did someone convert or not), most machine learning models require training data that contains positive and negative examples. Correlated will build out these labeled training data sets based on the goals that you specified in Step 2.
Training the model:
Correlated will now go ahead and train a machine learning model with your unique data. Because our models are built to be delivered as a SaaS solution, we’re able to build multiple models with different goals.
Interpreting the model:
Once the model is trained, Correlated goes an extra step to ensure that the models we build are not blackbox. We’ll tell you the top inputs that are correlated with your goals. You can use these “correlations” to build your own Playbooks or leverage the insights to optimize your customer lifecycle.
Scoring Accounts and Users:
Now that the data is cleaned and prepped, the model is trained, and you have some understanding of what’s going on in the model, Correlated scores your Accounts and Users (depending on which model you’re using). The score is a predictor of how likely an Account or User is to achieve an outcome. You can use this score to prioritize who to talk to and you can leverage our Playbooks to automate workflows.
Step 4: Operationalize your PQL Scores
Now that you’ve achieved a PQL Score using Correlated, you can easily deliver the highest scores to your sales team, automatically! Correlated comes with robust Playbook capabilities, including downstream support for Salesforce, Hubspot, Outreach, and Salesloft.
Imagine this scenario:
- The top PQL Scores for enterprise accounts trigger a Salesforce task to be created assigned to the account Owner.
- The top PQL Scores for all other accounts trigger a Slack notification in an SMB channel and automatically placed into an Outreach Sequence
- The middle PQL Scores are surfaced in an Account List view for SDRs or AEs to prospect and are also placed into Hubspot nurture campaigns to encourage more engagement and product usage
- The lowest PQL Scores are ignored, but targeted with CSM outreach if they trigger churn alerts
You can see how Correlated’s PQL Scoring Engine makes it possible to get a PQL Score faster by leveraging machine learning. It also allows you to test and iterate, learning as you go! Finally, Correlated’s PQL Scoring Engine fits in seamlessly with the entire GTM motion, making it easy to operationalize the scores you’re generating.
Try Correlated’s PQL scoring
If you’re interested in trying out our PQL Scoring Engine, it is currently available in Beta. You can sign up for a free account and contact us via Intercom to request access.
.jpg)
