If you’ve been following the evolution of product-led growth, you’ve probably heard of PQLs. PQL stands for “Product Qualified Lead”. They are similar to MQLs (Marketing Qualified Leads) in that they are “qualified” leads, but PQLs are qualified when users do certain things in the product, whereas MQLs are qualified when prospects engage in certain ways with the company’s marketing assets.
Who uses Product Qualified Leads?
You need PQLs if you’re selling into prospects who are already using your product. As they use your product (either in free trial, free tier, or a lower tier that you can upsell), they are constantly indicating their intent. You can use this to prioritize who to reach out to without being annoying.
According to OpenView’s PQL report, the best PLG companies employ PQLs, so if you’re looking to improve your product-led motion, you should look into doing so as well.
Examples of PQLs
The three most common categories of PQLs we see are the following:
- Condition-based: if a certain set of conditions become true, a lead becomes a PQL
- Score-based: conditions are weighted, and if a threshold is crossed, a lead becomes a PQL
- Machine Learning based: a model predicts the likelihood to convert and assigns a score. If the threshold is crossed, a lead becomes a PQL
In this blog post, we’ll dive into each of these categories and discuss how you can go about defining each type of PQL, and how you should think about measuring the success of your PQLs.
1. Condition-based PQLs
If you’re just getting started with PQLs, we highly recommend that you start with condition-based PQLs for the following reasons:
- It’s less of a black box as you are purposefully defining the conditions that make up a PQL
- You can incorporate your existing data model into your PQL definitions
- You’ll be able to test and iterate since you built the conditions in the first place
- You can link PQLs to relevant playbooks based on the conditions you defined
Let’s walk through some tangible examples.
Example: Conversion Triggers
Let’s say that you’re trying to find users in your self-serve funnel that are likely to convert. A good place to start is to work backwards from conversion and identify “triggers” that happen before conversion. For example, in Correlated, a conversion occurs when accounts invite more users to the platform. However, before that, users have to achieve some level of success in the platform.
A typical customer journey through Correlated:
=> Sign up
=> Import Data
=> Create a Workflow
=> Invite more users (conversion event)
So, we could build a PQL that scores accounts higher if they create more Workflows. Now, you can test and iterate to see if accounts who create more Workflows do indeed convert more. You can also automate Workflows to focus on driving accounts to adopt Workflows, since that’s the main condition defining your PQL. You can also use a combination of different conversion triggers as your PQL, so this is a very simplified example.
Example: Hand Raises
This is an obvious way to generate PQLs. In product, you can build ways for users to upsell themselves. For example, you might gate a “pro” feature, but count someone as a PQL if they land on the feature and click the “upgrade” button. You might lower the sensitivity of your PQL to include all users who land on the “pro” feature. You might also include a “schedule a demo” or “contact sales” in the product to capture users interested in various features that they can’t access today.
If you’re able to link the hand raise directly to what the users were doing in product when they reached out, you’ll be able to have a much more informed conversation with the user about what they want to achieve.
2. Score-based PQLs
Score-based PQLs layer on top of Condition-based PQLs to prioritize your leads.
Many of the examples outlined above are considered “binary”. A user either satisfied the conditions, or they didn’t. However, what if you don’t have enough leads? Or what if a user really satisfies one condition out of three, but doesn’t satisfy any of the other two? Are you missing out on leads?
This is where score-based PQLs are highly useful. Rather than defining a lead as a “yes” or “no” concept, a PQL is scored on a range. Higher PQLs are more likely to convert, Mid-Range PQLs still might convert but are less likely to, and Lower PQLs are less likely to convert. The benefit of using scores is that it helps prioritize leads and leaves room for your sales reps to make a judgment call on whether or not to reach out to PQLs with lower scores. You can also use scoring to segment out your customers and do different things for them. For example, customers who aren’t a “High” PQL can be put into nurture funnels rather than being immediately sent to AEs.
So how can you go about building a score? Let’s run through an example.
Example: Activity Health
Another common way to define PQLs is to measure Activity Health. Now, this one gets a bit more complicated as it requires you to do a bit of math to define what “healthy” means.
Correlated provides this out of the box if you import event data into Correlated. We grade accounts and users against other accounts and users on a scale from Poor to Excellent. If you’re looking to do this on your own, you can start a bit simpler. Perhaps your PQL is if someone has above average activity (total count of usage events). Or, you could weight different activity types (for example, advanced features are weighted more) to get to a full score.
If you choose to weight activity types, you can build playbooks around driving more usage for those activity types.
3. Machine Learning PQLs
You’re probably thinking, do companies really use machine learning for PQLs in B2B SaaS? And the answer is a resounding yes! However, this is of course limited to companies more advanced in their PLG journey who have the resources to build out data science teams to do this.
Typically, building out a machine learning PQL involves the following steps:
- Gathering relevant data
- Formatting the data in a way that the ML model can understand
- Training the ML model
- Testing the ML model for accuracy
- Using the ML model to predict outcomes
Machine Learning PQLs are powerful, but they also come with some serious caveats. Machine Learning models tend to be more black box, so it’s harder to figure out exactly why a PQL is good. Also, Machine Learning models are difficult to maintain over time, which means that it’s harder to test and iterate continuously unless you truly have the resources to do so.
So what can you do if you want to try using machine learning to build out PQLs?
There are great off the shelf tools that can get you started, mostly in Python. Some cloud vendors like Google also have great tools for machine learning.
For a less technical route, Correlated just launched a closed beta for PQLs generated by machine learning that you can test out. Simply create an account and let us know you’re interested in testing it out! We’ll be sharing more information about this feature soon.
PQLs are a marathon, not a sprint
At the end of the day, you can see how PQLs can be as simple as one trigger condition, or as complex as full-on machine learning models. Our advice is to start from somewhere, knowing that as your product and business grows, your PQL definitions will grow with it.
If you’d like some support along the way, Correlated is here to help you uncover PQLs and implement playbooks on those leads for conversion and expansion. You can create a free account here.