Predictive Scoring Failures and Lessons Learned

By July 23, 2018
Share this:

Predictive scoring is a tool that marketers and sales associates often use to analyze whether a company is a good fit for the product or service that their business has to offer. Most algorithms are set up to analyze this data and to consider both positive and negative results in their formula.

While this sounds ideal at first, it can actually cause some serious issues for clients that are using this algorithm. Scoring methodology can put you in a bad place if it’s not effectively implemented. The software could give you false positives and tell you that companies that aren’t actually a good fit for your business are the kinds of businesses you want to pursue. Or worse, it could give you negative results, telling you your ideal companies aren’t a good fit after all.

In some cases, it’s a matter of the algorithm not quite understanding nuances and complexities of what it’s actually scoring. In others, it’s a case of the company not understanding what kinds of interactions should be scored and what should be ignored.

What Failed with Predictive Scoring?

Let’s say, for example, that Gold Standard Fittings made a huge sale to Coca-Cola. The algorithm would count that sale as a positive and conclude that beverage companies are probably a good fit for GSF. It would rank all future beverage company prospects high because of this initial success.

Later, GSF tries to build a relationship with Pepsi, but they can’t quite make the sale. Maybe the timing is wrong, maybe there’s a purchasing freeze, maybe it’s Pepsi’s peak season and they’re focused on other matters. However, our example algorithm counts this as a negative score, and lowers its Fit ranking for beverage companies. As a result, GSF misses out on pursuing Royal Crown and Faygo as possible clients.

Sales fall through for a myriad of reasons. The fact that GSF was able to get Pepsi into the sales funnel at all shows that their product would work for them, but maybe Pepsi just had a better relationship with a competitor. Or they didn’t have it in their budget. Or the purchasing manager moved on. It could be any number of reasons, but the non-sale is counted as a negative and throws off the entire algorithm, and as a result, the company’s entire sales funnel.

When negatives are introduced into an algorithm, it can mis-analyze opportunities, and prospects that are a good fit for the company may be getting ranked lower as possibilities when it wasn’t the product or service that caused the sale the fail.

We’ve Seen This Happen Up Close

We had this issue at EverString when we first started out. We would introduce both sales and missed sales as positive and negative scores, respectively, and based our algorithm on those scores.

Except we started hearing from several users that customers who had ranked high were dropping out of their sales funnel, and customers who were buying were being ranked low, like 20 out of 100. With a score this low, our users were unable to pursue those customers because they weren’t even on the radar. They would just show up and make a purchase without any kind of warning.

Lessons Learned

However, as we’ve grown, listened to our customers, and revised our formula, we’ve removed negative scores from our algorithm. Now, we analyze the companies that our clients bring in and close sales with, as well as checking the ones that enter their sales funnel on their own. The second group may be ranked differently, but by removing negatives from the equation, we are able to truly identify the best leads for our clients.

Understanding how data is collected and then placed into an algorithm is critically important for scoring properly in order to target the best companies and potential leads for clients. This is something that remains an issue for many of our competitors — they haven’t managed to figure out how to deal with “negative” sales events.

As we strive to consistently identify the best leads for our clients, we have analyzed our algorithm and made sure that the scoring process is clear and effective. We’re able to deliver more accurate scores to our clients about their prospects, and they can properly focus their energy and attention to the right areas.

If you would like to learn more about how EverString uses its predictive scoring algorithm to help you to find the right kind of customers and prospects for your sales funnel, request a demo of the product.

Share this:

Related Posts

s