When I joined PredictHQ recently, I had so many questions. We’re around 100 people, and a lot of the internal knowledge is still pretty tribal. I’ve asked, “Hey, can you show me this analysis? Have you ever done that?” And the answer has mostly been, “We’ve never had the bandwidth” or “We’ve never had the opportunity or the know-how.” In the cases where that work was happening, it was happening in a silo, so other stakeholders didn’t have any visibility or knowledge about past results. There wasn’t really a culture of analysis ingrained across the org yet. So I decided that, as a starting point, I’d shove some data into ChatGPT and see what it said. The whole process took me maybe 20 minutes, when in the past it would have been days of analysis that still needed to be validated.

Disclaimer: It's important to go over the content you're running through ChatGPT and clean anything that isn't public domain that might be sensitive. Make sure to only run anonymous content, remove financial information if needed, and so on.

Since we don’t currently track the reasons for closed-won opportunities in Salesforce, I decided to start with closed-loss ones, which we do track. As long as you have Salesforce access and your team record loss reasons, you’ll be able to generate a simple report like the one I used here.

I know that some people will say, “Well, you have to clean the data, unify fields properly” and so on. And I agree that ideally you’d go through a rigorous cleaning process every time you want to answer a question in a data-driven manner, but I think it’s just as important to understand how to do more with less—and quickly. Not everyone has the resources and the time or bandwidth of public companies, so it's crucial to find faster, simpler proxies for directional signal.

I think for small orgs in particular, it’s useful to know whether it’s even worth it to pursue a more thorough investigation of the data. When you don’t have the resources to do extensive research, it can be so easy to lose that consumer-centric focus, and having a tool that unlocks all those insights living inside the CRM, chat tools, emails, and even more quantitative data can help make that process available to orgs of any size.

This was actually the first prompt that I used, and I was pleasantly surprised by the results for the most part.

I’m using the paid version of the product—I would recommend anyone who’s interested in ChatGPT do that—but I still had to split the data up into two batches, which introduced a couple of issues that were easy enough to overcome.

The first was that one time it returned percentages and the other it returned actuals, so I had to reconcile those.

The second issue was that the buckets or categories that it grouped the data into were slightly different, but they were all similar enough that I could easily merge them by hand.

While working for SolarWinds and Heap in my previous roles, I'd regularly done rigorous win and loss analysis. With product companies, you really see a lot of information around whether the product is missing capabilities or other competitive disadvantages that you might have against your space. So I knew that those were the kinds of things that I needed to validate here to make sure that we were looking at the data the right way and there weren’t any anomalies. When I took those results back to leadership, they found that everything was pretty much in line with what they had in mind—there was nothing weird or irregular. That was a real unlock for me: like, wow, you can easily quantify things and have a better understanding of how to approach discussions about where we should go next, where we should invest, and where we might be able to drive growth. As a one-person army here, the value of being able to get that directional signal by myself in a matter of minutes was huge.

Shelly Eisen-Livneh

I used ChatGPT to run this quick and dirty loss analysis. It helped give the team a baseline of directional signal and set us up to start building a more robust culture of analytics.

ChatGPT Loss Analysis at PredictHQ

Loss Analysis H1 2023

Raw SFDC data

ChatGPT Summary and Quantification

Loss Summary

Sign up to see more