Forum Academy Marketplace Showcase Pricing Features

Likelihood of success or failure (Predictor)

Good Morning Team Buble,

comparethemarket.com allow their users to input personal information and answer a number of key questions. In return, they can browse different mortgage options and most importantly the % chance of success if they were to apply.

This little tool allows those shopping for a mortgage to protect their credit score as they can make more informed decisions on which mortgages they are likely to be offered should they take the plunge and submit an application.

I want to build a similar predictor and integrate with bubble.

First things first,

Does anyone know how these work? I think the basic idea is that there are a series of factors that constitute ‘success’ and these are weighted against the data provided by the user.

Would appreciate any input.

Hope this grabs some of your interest.

Mike Lewis

I’ve done something similar to build a “matching system” that connects two users together. It’s tricky, but doable in Bubble. The key thing to workout is actually NOT Bubble related, but rather the logic you want to use to make the recommendations. I highly suggest having a clearly worded understanding of how you want it to work. If you can explain the concepts to a non-technical person of how your predictor will work, that’s a great starting point for bringing it over to Bubble.

Chances are that it is a rules engine that is implemented in the comparison site. With the rules being periodically updated by the provider.

Often these are “black boxes” so that CTM has no access to the business rules, they just input and gather the results.

That is how things tend to work in General Insurance, but this may not be as sensitive.

It is also possible that it is a call out to the provider via Webservice. But for something that is not likely to change that often, it may not be worthwhile,

The difference is that the “rules” are owned by the provider, and thus either need to be maintainable by them, or easily changed.

Thank you @NigelG and @andrewgassen for your responses. Let’s look at a relatively straightforward example and I will do my best to articulate what needs to be done.

User wants to know how likely he is to win a pub quiz.

Factors that could affect performance in a pub quiz:

  1. Number of pub quizzes previously completed
  2. Age
  3. Years spent in Education
  4. Books read per year

Each of the three factors might hold an original value of 0.25 or 25%

The four factors are scored depending on the answer to the question these scores are then weighted depending on how important the factor is compared to the others. The total of the weighted scores would be a value between 0-1 with 1 being 100% chance of success and 0 being 0% chance of success.

In your opinion, is this the correct starting point?

Did you ever come right with this app? Looking to potentially building something myself.