One side effect of 2020’s lockdowns was the collapse of many predictive models.

At the time, Guan Wang was in charge of customer success (renewals, operations, and strategy) at a SaaS provider of B2B travel services. Wang worked with the data science team to build a model to predict the likelihood of any given customer renewing their account. «A key factor in the model was ‘user adoption’ and we had to pivot to ‘onboarding,’ or the volume of new users signing in to use the SaaS platform at that account,» he said.

When pandemic lockdowns hit, user adoption essentially went to zero. And so did the predictive value of the model. 

“We shared all the possible factors and data points with the data science team so they could build a new model to come up with recommendations for the customer success team,” Wang said. “It was a very challenging time, and a big shift to figure out the best [account] health indicators” as the pandemic continued to unfold.

Clearly, the pandemic was an unusual, radical shock to the world, and thus the economy. But buying behavior was already changing quickly, and continues to do so as the world shifts around us. From increasing extreme weather events to the ever-evolving social media landscape, businesses face an array of factors that, gradually or abruptly, can require a refactoring of models and an urgent shift toward more real-time data.

Mastering the Art of the Quick Change

Wang is now Snowflake’s Global Director of Marketing Intelligence—a data consumer at a company totally focused on data. “Our team mission is to enable Snowflake marketing to become the industry’s most insights-driven team,” he said. 

These days, according to Wang, there’s no shortage of data for marketing departments to work with, in both B2B and B2C. “Ad channels, social channels, content marketing, sales events—we can use lots of channels to engage and to collect signals about the customer,” he noted. However, many organizations don’t have the right processes in place to wrangle all that data effectively. That’s the first hurdle.

“Companies need to get all that data in one place, where you can see a full picture of the customer, and understand how to keep them engaged throughout the sales process,” said Wang. 

The second hurdle is the rate of change. On-again, off-again pandemic lockdowns certainly had a negative effect on many industries, from travel and ride-sharing to manufacturing. But some companies and industries faced the opposite effect, experiencing a sudden business boom—think of video conferencing, food delivery, and language apps, for example—which is preferable but still comes with its own challenges. A McKinsey survey conducted in June 2020 found 75% of consumers had already “tried a new shopping behavior” in the first few months of the pandemic.

However, there was already plenty of change in the air before the pandemic hit. We can’t anticipate what new technology will emerge that will influence future consumer behavior. Consider social media platforms. According to a 2019 survey centered on creating brand loyalty among younger generations by creative agency Composed, 34% of millennials use Instagram to discover new brands, and of course Instagram didn’t exist when members of that generation were born. For Gen Z, that number jumps to 58%. Gen Z is also heavily influenced by TikTok, which of course didn’t exist when Gen Z was born.

In addition, millennials and Gen Z buyers have high expectations for corporate social responsibility that preceding generations did not widely display; Deloitte research found that 37% of millennials have retreated from companies they didn’t believe were ethical.

Wang pointed out a B2B wrinkle that further compounds the challenge: very long sales cycles.

“A customer might be qualified, but then moving to close may take 6 or even 12 months. So  how can I keep that customer engaged with key marketing and sales signals?” he said. This brings us back to the necessity of getting all the data in one place. Wang believes marketers need to extract better data insights that incorporate not only traditional marketing segmentation, but also an understanding of the best “next step” recommendation based on each customer action.

It all adds up to this: marketers have to make sure predictive models are working with data that accurately reflects what’s happening on the ground, in the moment.

Nearing Real Time

All of this inexorably points marketers towards real-time data. Wang said it’s an important goal, although many marketers will need to take some interim steps on that road. 

A 2021 study by InsideSales examined lead responses at more than 400 companies, covering more than 5.7 million leads. The study found that conversions were eight times higher when the seller responds to a new lead within five minutes. However, less than 1% of first response calls occur in that time window. B2B buyers might cut you a bit more slack; 2020 Foundry, formerly IDG Communications’ research on millennial-generation tech buyers, found that only 16% expected follow-up within the first three hours. 

“For companies to figure out how to engage in the first five minutes, it’s really hard-core science. And potentially huge engineering costs, too,” said Wang.” Getting all the data real time is really difficult. Think about linking Google ads and Facebook ads into a centralized platform. A lot of data pipeline is needed to bring it all together.”

“Near real time” is the interim step that most companies will set their sights on. Wang said many data suppliers still only update the data once per day. As a data consumer, he is pressing to have all data updated every two hours, which is how Snowflake’s marketing group defines near real time.

The ultimate vision for marketers, Wang believes, is a full visualization of tomorrow’s extremely complex, fragmented buyer’s journey. Right now, he said, there is “no easy way to visualize or map this nonlinear journey.”

“It’s really complex, almost like a neural network, and it’s never been developed by any company—even a company like Google, they’re building amazing AI and machine learning models, but still no one has modeled the whole network.”

In the long term, Wang has an optimistic outlook for marketers: “In the next 5 to 10 years we’ll have a better understanding of it,” he said.

In the meantime, busting data silos, moving toward real-time data, and keeping predictive models current are valuable things for marketers to focus on.