Statistical Data Analysis

Natero: Machine learning for SaaS customer retention

The world of Artificial Intelligence and machine learning is one that seems to get muddled. On the one hand, there is all the science fiction potential of truly smart machines, on the other, there is the gradual automation of complex data analysis enabled by machine learning. And while the latter tends to be party to all the more bleak predictions about job losses, this isn’t necessarily the case, as the scale of possible number crunching is way beyond human capability, and normally requires the help of human workers.

In fact, this is precisely what I think is interesting about machine learning. Forget all all the big scary headline grabbing stuff. Think instead about all those very prosaic areas it is working away to assist with in the background, which will subtly change the way we work and do business forever.

Darktrace is one interesting example of this because it uses machine learning to profile normal employee behaviour so it can easily spot malicious intruders in the enterprise. And now Silicon- Valley-based Natero claims to offer something similar for the world of SaaS customer retention. As the press release boasts it is “the only customer success platform to merge machine learning for predicting behaviour and big data analytics for deep customer insights”.

So, what does this really mean in practice? Well, as Craig Soules, the Founder and CEO – and formerly Principal Researcher at HP Labs – explains over the phone: as SaaS has taken over as the dominant revenue model “people have to be happy in the service”.

This means gut instincts for what is working with a customer only go so far. If an account manager is handling 200 separate customers it will be impossible for that individual to really tell what is happening with all of them. And while there is a lot of data available on customers – from billing details to subscription preferences – much of it is siloed and the majority of account handlers are not too comfortable handling data, anyway. 

The solution that Natero has come up with to counteract this comprises of gathering each individual client’s customer data together and then building a bespoke machine learning algorithm based on the data. It then provides easy-to-view intelligence graphs and notifications on customers that might be about to leave and customers that might be ripe to invest in more services. The upshot, as Soules sees it, is that businesses will be able to use these insights to keep customers happier and in the long run build better products.  

The data utilised at present doesn’t include external factors, like social media, explains Soules because existing customer “data is so rich that we don’t need them”.

Natero raised $3.3 million in Series A funding from Merus Capital, Salesforce Ventures, Y Combinator and Andreessen Horowitz Seed in 2014 and officially launched last November with its first product launch at the start of May. It already has SaaS firms such as Freshdesk and Rainforest QA under its belt and at the moment is “focused on sales and growth”. 

The direction the company takes will depend on how it progresses over the next six months, Soules explains. He adds he would certainly consider raising more funding if need be and may also look to produce some out-of-the-box solutions, where relatively common datasets could be grouped together. This could be helpful as some clients start with no usage data which means it is a couple of months before there is enough intelligence available to build a solid solution.

Machine learning is slowly taking off across numerous areas from automated virtual assistants to the programs scanning YouTube cat videos to learn about their behaviour. So, why is nobody else applying it in this particular way, especially as SaaS is growing at such a phenomenal rate?

“You need detailed usage information,” says Soules and you need to be able to scale the data. Many people in the field simply do not come from that kind of data background, they are more focused on the process sides.   

“We wouldn’t have tried if we hadn’t seen it done in other areas,” he clarifies.

At present there are two other companies in the space. These are Gamesight and Totango – both are slightly larger than Natero. But Soules explains that the main difference between these and his organisation is, although they have been running for longer and do provide limited data modelling, they do not use machine learning.

The whole area around analytics and data “can get muddied” he says.

This area of confusion around data is part of the reason the idea for the company came about in the first place. Soules tells us when he was working at HP he noticed a disconnect in the data processing tools available to developers – who are comfortable with complex tables – and those suitable to business users. The latter want usable insights and easy-to-follow charts, graphs and notifications – not databases. 

Customer retention was not the only idea Soules investigated before the launch though, he also looked into eCommerce and the Internet of Things, but concluded this space needed a solution. “I saw a lot of bad business software,” he says before going onto explain he wanted to be instrumental in helping to improve the overall quality of business software moving forward.

The thing that surprises people most about all this is “the fact that it works,” says Soules.

“Machine learning is not the end,” he concludes, it simply takes the complex analysis away, and “enables people to focus on things they’re good at” like true account management.  


Further reading:

Can ‘good’ machine learning take on global cybercrime?

Office 2021: Why robots won’t end drudgery or steal our jobs

What will health tech mean for ordinary people in 2026?

How will tech have transformed our lives by 2026?


« Typical 24: Steve Miller, nfrastructure Inc.


Gamification & a drive to change millennials' view of insurance »