Join the Community

01 Community General

01 Community General

Upgrading Scoring Mechanisms with Insights from the Clay Community

·

·

We are looking to upgrade our various Scoring mechanisms and wondering how people have found success doing this within the Clay community. I totally understand how we would use Clay to implement a determined scoring model, but frankly we could use some autonomous / data-backed assistance in determining the model and making it "self-healing" over time. Problems we're trying to solve:

ICP: Are our "fit" definitions correct or do we have an incorrect opinion of our market?
Signals: Aggregating various 3rd party (6sense, G2, etc.) and 1st party (website, inbound touches, etc.) into a single score
Inbound: Prioritizing the right inbound leads for our Sales Development team
Churn Risk
Opportunity Likelihood to close
Account Likelihood for Cross-sell / PLG expansion

Are people building all of those scoring models in Clay or using 3rd party tools (Forwrd, MadKudu, RevSure, etc.)? How are you doing the analysis to determine the models up front?

3 comments

Jani V.
·
·
my 2 cents - without being a developer and seeing the actual algorithm of the "self-healing" 3rd party tools, but having worked with a handful of clients who used these, they eventually discontinued using them, and opted for a Clay-powered model that they have more control over and visibility Alternatively, you can push the MadKudu, RevSure, etc. scores and segment data into Clay, and try to combine them with Clay-monitored signals, but this can definitely become a cumbersome exercise
Collin R.
·
·
Am I missing a capability within Clay to use ML models to DERIVE an algorithm? Also, I understand being able to RUN a static algorithm once determined, but seems like it wouldn't be able to handle a dynamic "self-updating" algorithm? Looking for tools to help with clustering analyses, dynamic ICP modeling, opportunity scoring, etc.
michael v.
·
·
Hey Collin I built an account scoring model in Clay at my last company. Here's how I did it:
Exporting a list of all customers from CRM including ARR
Ran enrichment in Clay to gather data points like "PLG vs SLG", "OSS Core to Business Model", Popularity of Open Source Repos, etc.
Then I exported the fully enriched list to ChatGPT
I spent about 2.5 days just playing with the data. ChatGPT is really good at helping to prompt you for different slices and ways to explore
I tried running a k-means and linear regression, both bombed and didn't turn up anything good
I found the data points that had the most impact on deal size and time-to-close
I built the scoring model in Sheets then copied to a Clay Score
Iterated on the point system until I got known "great" , "good", and "ok" accounts scoring correctly
After building, I realized I should have built 3 scoring models specific to Commercial vs Enterprise Vs Renewals
There's no easy button on this. Great scoring comes from deep understanding of the market and customers. Clay helps make getting accurate data easier. Hope that helps!