The Scoring Model Was Wrong And It Put The Worst Lead First

I analyzed 2,163 companies in a specific US sector, trying to determine which ones should be approached first. Two key factors mattered: their ability to pay and the severity of the problems they faced. I measured these factors using public data such as domain age, presence of sitemaps, last publication dates, mail providers, and registry filings.

Initially, I encountered issues with the data. For instance, I found that 24 percent of the companies were unmeasurable, despite having archives going back more than a decade. I reran the measurement process and the coverage improved significantly, from 24 percent to 88 percent. This highlighted the importance of accurately measuring data, as not measured or existing data can create misleading results.

Next, I used a weighted sum to score the companies based on their ability to pay, the severity of their problems, and additional supporting signals. Surprisingly, a business with lots of money and no problems was ranked at the top of the sales list because it scored high on one axis (ability to pay) and had a low score on the other (severity of problem). This clearly showed that using a sum to combine the two factors was incorrect, as both had to be present for a meaningful sales opportunity.

After fixing these issues, I discovered that 1,680 of the 1,912 measured companies had no sitemap, only 76 had a blog, and only 205 showed any publication date at all. Moreover, 1,293 companies used Google Workspace or Microsoft 365, indicating that they were real operating firms. Most companies in this sector were not publishing anything, a finding that only became apparent once the measurement layer accurately reflected the data.