Don’t Rely on Your CRM’s Lead Scoring Model . Build Your Own.

William D'Souza
6 min readJan 11, 2024
Photo by JOSHUA COLEMAN on Unsplash

In high velocity sales environments, ensuring that your teams are using their time appropriately to communicate with the right leads is always of top concern. We always want to ensure that sales teams are hitting their goals and looking and we always look towards revenue and sales operations teams to establish the right processes technically so that the sales agents can utilize their skills on dealing with clients; where it matters most.

Many teams have different approaches to help these teams, they create KPI’s and send data to their CRM platform so that sales agents can utilize it. They create sequences that are (hopefully) backed by data but sequences take long to establish and aren’t useful long term. They work for a short period of time until things change (and they always will) and by the time most teams establish their sequences, things have changed.

Sequences do solve a lot of problems and if backed by data correctly, they can be made so that creating sequences aren’t as difficult as they typically are right now.

Effective teams utilize some sort of lead scoring model, and some get sold on their lead scoring models built on “intelligent machine learning” from CRM platforms… they flat out don’t perform well out of box. It’s not that the engine running these models are bad, it’s that the data going into it is either limited or garbage.

If you don’t want to read the rest of the article, the basic summary is this:

Ditch the platform’s lead scoring tool. Build your own.

Why You Need a Built in House Lead Scoring Model

Most teams may feel that they have a good sense of their customers, not acknowledging that customers themselves will evolve and change, along with the industries they operate it. Having a good gut feeling of who to target is one thing, but knowing who to contact and why is another.

While you can go with personal industry experience or your gut feeling, this is most probably how you will become outdated. While these tools may be scary to implement… “AI is taking over and all our jobs!”… its important to realize that lead scoring models are tools to operationalize sales. They won’t replace the skill set involved in making a sale, but can be used to automate a self led sale with some creative thinking.

Overall Benefits of a Lead Scoring Model

Lets outline a few benefits of having a lead scoring model in the first place:

1. Prioritization

  • You will have quantifiable metrics on each lead that comes in, which creates empowerment within your organization to operationalize better
  • It will give your various sales and sales related teams the ability to create rules for prioritization

2. Time to Cash Flow

  • Decrease your time to conversion, inadvertently increasing your time to cash flow
  • Convert leads faster by contacting them when they are most engaged rather then waiting days to target them or flat out miss them

3. Expanded Automation

  • Focus on automated product marketing or traditional marketing initiatives to get leads more interested or self convert themselves
  • Use promotional material at the right time to increase likelihood to convert

4. Always Updated KYC (Know Your Customer)

  • KYC’s are an important initiative (and regulation in many situations) that are done once and never updated, but they can hold a lot of value if you keep up with its purpose… knowing your customer…
  • Don’t wait for long periods of times to understand why something isn’t working when it used to so you can stay up to date with less effort

What is Wrong With Your CRM’s Lead Scoring?

Like we said before, its not that these CRM’s have some horrendous engines. Some engines are better then others but its more of an issue of input data and goodness of fit.

Their models can learn overtime if you set up your CRM properly, but most sales operations team set up their CRM to optimize for it operationally, not realizing that the CRM can be an analytical power house. While what they are doing is important, these professionals are more times then none not data experts, and lack those fundamental skills.

Garbage in. Garbage out.

With all this in mind, it makes sense to create an application decoupled from the sales operations that can act and be maintained independently. This allows dedication to maintaining your CRM operationally without the pressure of adhering to complete analytical standards (you should always have data hygiene standards though). Decoupling it satisfies both operational teams and data teams, you get to put food on everyone’ s plate.

Some other benefits?

  • Include external data that the CRM may not contain, increasing your models efficacy
  • Ability to bind context across varying domains in your organization, because your whole organization plays a role in converting customers
  • More transparency into understanding scores in your model along with flexibility of managing a model on your own

How to Get Started on Building a Lead Scoring Model

Photo by Isaac Smith on Unsplash

While this won’t be a step by step guide on how to build a model in itself, we will touch on a few points sequentially to help get you started. Building models are unique and context matters quite a lot; no two models are constructed the same.

Its also important to note that you want to set checkpoints in a project to qualify it. These projects aren’t always going to succeed for a number of reasons and you need to know when to quit while your ahead to save on costs. However, if successful, the benefits should outweigh the investment.

  1. Collect feedback from your sales and sales related teams on what their frustrations are, what they find both working and not working
  2. Understand what data sources you have available to you that are logically related to your initial customer journey
  3. Run analyses on your data to ensure that the data you have is viable as input data for your model, visualize and understand your data
  4. Run through a model development cycle with proper testing to further validate and tune your model
  5. If and when you are happy with your model and it’s metrics, run a pilot program to ensure your model is working at intended
  6. Train your teams! This is the most undervalued step and the ones that can have severe consequences to your model… yes you can actually break your model and its intended purpose overtime (a topic for another day…)
  7. If all is well, its time to put it into production, but don’t forget to set up proper monitoring and constantly run tests on your model to compare it to the actuals that you have, as this will ensure the models effectiveness

While this may seem like generic steps, its important to outline these steps properly so you don’t have negative consequences when building this model. Thinking the model should be able to function versus it actually functioning is two different things.

The steps are meant to outline that importance a few things; context, validation, time investment, and viability over the models lifespan. When you put out a model like this, you may start to notice that everyone starts to find a use case on how they can use it, and you will have to manage that in itself.

Time to Get Started?

While this may seem like a daunting task that will take time and effort, it is a investment that we have seen do wonders for companies. Most companies that have a properly functioning lead scoring model understand that time is a valuable asset, and an application that focuses on quantifying a lead’s worth, comes with numerous benefits.

Most companies we know don’t even actually rely on their CRM’s lead scoring tool, they establish a single proxy to help quantify a lead and build operations around it, and its usually done extremely naive to the point where you may as flip a coin. The ones that rely on the CRM’s model may or may not be benefiting from it, it all depends on the data inside the CRM.

At Kizmet Solutions, we have implemented multiple lead scoring models in a variety of situations. We can help evaluate your data first to ensure if its something to proceed with. If not, we would be happy to compile some recommendations to get you on the right track.

--

--