Why Snyk chose to build a new lead qualification system over buying

Houston, we have a data and reasoning problem.

That’s the conclusion reached by Randall Degges, VP of AI engineering and developer relations at AI security platform Snyk, after he was approached by his company’s demand generation team over an issue with their lead qualification system. Although Snyk’s tech stack had applications like Salesforce, MadKudu, and Marketo to help organize leads, there were discrepancies in how high-quality leads were classified.

“For a few months this year, all of the marketing leads that we were taking, if those leads did not directly connect to an account that was already in Salesforce, we were basically just rejecting them and ignoring them completely,” Degges said, adding the company had ignored almost 8,000 leads that should have been classified and placed in Snyk’s Salesforce system.

To build or buy? That is the question. Degges was tasked with finding a solution to the inconsistency, which he blamed on how applications integrated with one another. From his perspective, building a piece of software from scratch was the obvious route to take.

“To my knowledge, they’re not tools that can do this easily today that you can just get off the shelf,” Degges said.

Cody Pierce, co-founder and CEO of AI-native browser security platform Neon Cyber, told IT Brew that build-versus-buy discussions have been going on since “the dawn of computers,” but AI is making the former a more appealing and accessible option for companies.

“The barrier to actually build something that kind of works is pretty low now, and on the pro side of building versus buying, companies know their business the best,” Pierce said, adding that software bought off the shelf is often one-size-fits-all.

How it works. It took about three days for Degges to build out Snyk’s new lead generation system, which leverages AI to score certain characteristics, like job title and company of employment, to determine if someone ultimately makes a good candidate for Snyk products.

“It takes all this information, and it eventually takes all of those data points and all the scores that we associate with each of those things, and then it runs it through a final AI reasoning layer,” he said.

The company spoke with internal sales development representatives (SDRs) and other staffers who would typically look at leads to benchmark some of the results from the newly built system by making sure leads were appropriately categorized.

“We did this enough times to get confidence that the results we’re getting matches what the human SDR would want to do,” Degges said. “We’re trying to get as close to human intuition as possible and have a high level of accuracy.”

Leading by example. Since building out Snyk’s new lead qualification system, Degges said the company has seen large efficiency gains. He estimates employees save about 1,300 hours a year on the manual work associated with lead generation.

Degges has been incorporating team feedback into the software as needed. He plans to continue to build out the system so that it can serve more than one purpose.

“We’re in the process now of finding ways to incorporate it into SDR daily workflows when they’re doing those types of outbound research projects and things like that,” Degges said. “So, we’ll be able to leverage it in different ways, not just to solve this one-off problem.”

Companies debating whether to build or buy software to address company problems should assess the costs associated with the former. “Not just like the one-off cost too, but the maintenance cost, the infrastructure cost, everything,” Degges said.

Pierce added that IT pros should make their build-buy decision after asking themselves the following questions:

• What would happen if the software goes down?
• Will the tool be internal or customer-facing?
• How many people will rely on it on a daily basis?

“Those three questions probably get you in a good spot to think about it from your business,” Pierce said. “Because it’s going to tell you what the impact [is] if you’re not doing maintenance, if you’re not scaling it out, and if you’re not improving it over time.”