Propensity Modelling

Sort and filter lists of data by it's likelihood of performing a given action – whether that's buying, leaving a review, or something else.

Wouldn't it be great if you could pick out the contacts or companies from a list that are most likely to perform a specific action and focus on them? Imagine if you could pick out the 4 companies that were actually going to convert this month from your prospect lists, and focus fully on them rather than the whole of the 1,000 that might be on there. What if you could pick out the existing customers that were showing the most signs that they were going to cancel soon, to show them some of the love that would help reduce churn? Propensity modelling can help in those situations and any others where there is a list and an action involved.

What is propensity modelling?

Propensity modelling is the process of predicting the chances that a given entity will perform a given action. Whether a prospect will convert, or a customer will cancel, or any number of other potential scenarios.

Once these likelihoods are computed, they can be used to help guide precious resources in the right direction. Sales resources can be targeted at the prospects that are most likely to convert, customer service teams can be pointed at customers who might need a little more attention, and so on. Through focussing those resources in the right direction, you can achieve higher efficiency levels with the added bonus of spending less time having conversations that waste the time of both parties.

How does propensity modelling work?

For any given entity (e.g. prospect) and action (e.g. conversion) pairing, we work together to find all available measurable aspects. Once we know all of the individual aspects that we can measure about an entity, and how that affects their likelihood of taking the action, we can run a list of entities through the model and rank them by their propensity score.

As a simplified example: Let's imagine you're looking at companies on a prospects list, and you know that certain features about companies make them more likely to purchase from you. You attach a weighting to each of those features, based on how important they are relevant to the rest:

  1. Revenue between £500,000 and £3,000,000 (1 point)
  2. Revenue growth between 10% and 25% year on year (3 points)
  3. Between 20-50 employees (2 points)
  4. Increasing number of jobs listed on online job boards over the past 3 months (3 points)
  5. Increasing average review rating on Google Reviews over the past 3 months (1 point)

A company that fulfills numbers 1, 3 and 5 on that list would get a score of 40% by getting 4 out of the available 10 points. A company that fulfills numbers 2 and 4 would get a score of 60% and as such would be a better prospect to focus on.

Frequently Asked Questions

Do I need vast amounts of data to model from?

No – in fact a set of 'best guesses' can work as an initial model to work from. You likely have more relevant data than you think, though.

Won't this decrese the size of my lists?

Nothing will be deleted from your lists, but propensity modelling allows you to focus on those most likely to take a specific action. That means, for instance, that you focus your marketing activity on the prospects that are most likely to convert.

What actions can I create models for?

Almost literally anything – if we can define and measure a set of features for a given list and action, than we can create a model to work from.

Let's re-focus your lists.

If you're ready to explore the possibility of ensuring that your lists are ordered by something more useful than their place in the alphabet or the last time they were called, we're ready to chat.