Consumer intelligence startup Streetbees plans to create the world's largest data store about human behaviour by collecting granular consumer insights through an app powered by emerging data science techniques.

The London-based company has already developed a global network of more than one million consumers who provide photos, videos and text about experiences and products, which corporations can use to gain deep insights into their market and understand tastes in everything from tea to universities. Now it wants to scale that network by using machine learning.

© Streetbees
© Streetbees

Streetbees provides users - known as "bees" - with a chat-style mobile app that they use to complete tasks that request information around specific services and products. They then share images and words about what they're doing and why, which Streetbees converts into insights, using natural language processing and machine learning techniques.

Traditional statistical approaches can only group answers into clusters that are identified in advance. Machine learning can work out the best way to divide them as it analyses the data.

Tasks can either be voluntary or paid, with the fee determined by location and complexity. Writing down some thoughts on your mobile network might earn a few pounds, while taking a one minute video of making tea could make you a tenner.

Scaling the network

Streetbees is now developing the next generation of its consumer research network, using applied AI to create an automated knowledge system that CEO Tugce Bulut claims "will in time hold even more information about how we behave than Facebook or Google."

Adding contextual information to the demographic groups traditionally used in the market research industry gives the company a unique position in a sector that generated $44.5 billion (£34 billion) in 2016, as the company's new chief technology officer explains.

"By combining those two things together, we can understand a lot more about why people are doing things and for what purpose, and we can ask them clarifying questions that get behind it," says Sam Lowe, Streetbees' first CTO. "You can get a lot more information with that than you can by just asking multiple choice questions that you've assigned in advance."

"It's only really been possible to do this at scale with the combination of mobile apps and AI. Before that, you could only get that level of insight if you kind of spoke to the people verbally and had a conversation with them, and that really limited how many people you could reach.

"It provides a lot more insight because you don't have to decide in advance how you're going to analyse the different groups of people, and what they think, and what they like. You can let the machine learning that we use draw that out from all the different dimensions, and tell us which ones are the most important."

Streetbees was founded in 2015 by current chief operating officer Oliver May and CEO Tugce Bulut, who came up with the idea when she struggled to find quality data during her career in strategy consultancy.

The company raised $12 million (£9.1 million) in Series A funding, led by Atomico, which adds to the $5.1 million (£3.9 million) in seed funding it raised the previous year, and will be used to develop the product and expand the business.

It now operates in 87 countries and has worked with 9 out of the world’s 10 largest consumer goods companies, including Pepsi, Unilever and L’Oreal.

How Streetbees understands behaviour

Streetbees learns about context through a technique called 'demand spaces', which divides consumers into different clusters based on the purpose and context around their actions.

This technique recently identified a new segment of consumers dubbed "one-handed gamers" who are male video gamers aged 20-30 who frequently want snacks they can eat in one hand so they can keep playing. 

It also discovered a subset of this group comprised primarily of female compulsive phone users who crave healthy snacks while they chat.

"I think intuitively we can understand where that's come from, but it's the kind of thing that's almost impossible to spot using traditional demographics, because those people are hidden in amongst all the other people in the same social age group who don't behave like that," says Lowe.

"The combination of the context and purpose together with their demographics gives a segment like that, which brands can then tailor products exactly to, and cater for better."

Developments in AI

Lowe previously worked as CTO and head of AI at medical chatbot company Your.MD, and says that a background in AI an essential requirement for his new job due to the speed of developments in data science.

He's particularly excited about the reemergence of graph theory, a branch of mathematics that analyses the relationship between different ideas by measuring them as a series of points connected by lines, forming a data graph.

The theory has been around for almost as long as computer science, but was traditionally only used to find insights in categories that were pre-determined by experts who would map the domain in advance. Machine learning techniques mean it's no longer necessary to define these graphs in advance.

"Once you've gone from data to information, the next step up is to go to knowledge, and to use that in your future interactions with data. And a learning graph is one of the ways to do that," says Lowe.

Streetbees is developing new techniques so quickly that Lowe believes the company will soon begin to conduct its own research.

"The market research industry has innovated much less than others in the past, so it's crying out for somebody to bring this kind of mobile technology to it, and to provide an alternative to the way things are working," he says.

"But actually the techniques we're using we think will apply in many other industries as well - pretty much anything where you need to gather large amounts of information from people around the world and make sure that it's statistically valid and that it's balanced."