Empowering predictions for human behavior

By: Eric Bender

Big data is a very big deal, but it would be even bigger if it were easier to analyze. Organizations are awash in data from mobile phones and other sources, while machine learning and related advanced analytic technologies are advanced rapidly. But properly deploying these software tools to make predictions requires experts who are rare and expensive, and these projects are rarely quick and never simple.

Endor Software, a STEX25 startup, offers a powerful alternative technology for predictive analytics. “Our product can allow everyone in the organization to ask any kind of a business predictive question and get fast and accurate results across multiple types of business initiatives,” says CEO and co-founder Yaniv Altshuler.

Our product can allow everyone in the organization to ask any kind of a business predictive question and get fast and accurate results across multiple types of business initiatives.

The company's platform is based on “social physics” technology spearheaded by Endor co-founder Alex Pentland, MIT Professor of Media Arts and Sciences, and his collaborators.

“Social physics is our discovery that any data that represent human activity, regardless of the nature of the data, must adhere to a certain set of mathematical equations,” says Altshuler, a researcher in Pentland's group.

Social physics algorithms follow the flow of ideas through social networks in which individuals participate, and make connections to resulting social behaviors, Altshuler explains. The technique can boost the power of predictions for many questions on how individuals in social groups will act, as Pentland, Altshuler and numerous colleagues have extensively demonstrated in papers in top scientific journals.

Additionally, social physics-based platforms can make it far easier than conventional analytic tools to pose multiple questions on data without changing the underlying algorithms, Altshuler says.

Moreover, since Endor's software focuses on a broad range of network interactions rather than scooping up specific details of individual interactions, it can handle encrypted data, as no understanding of the context is required. And the software can work across multiple databases that previously were almost impossible to integrate—and do so without breaking privacy rules.

“Instead of spending months and months cleaning and understanding data, which is what traditional data scientists have to do today, social physics enables us to automate the process,” Altshuler says. “We can connect almost seamlessly to data, be it raw data or cleaned data or even encrypted data, and extract a large number of behavioral measures. Those in turn let us predict who will engage in various behaviors.”

“Social physics allows you to translate digital footprints into different buckets of behaviors,” as Tomer Srulevich, Endor's Chief Business Officer, puts it. “By finding commonalities among individuals in these buckets, we can predict the behavior of individuals we seek to engage. We can look at what actually took place in the last week, and how are we all emotionally impacted by many things that we see—news, weather, Covid-19—right now when the past is so different than the future.”

Social physics gets practical

Founded in 2014, Endor has 17 employees in Tel Aviv and New York. The company's cloud-based Endor Enterprise software serves a range of large corporations in retail and financial services, plus several defense and intelligence organizations.

Among Endor's corporate customers, for instance, MasterCard uses Endor Enterprise to generate targeted recommendations for finding, assessing and cross-selling to their customers. Global defense agencies employ the software on a wide range of challenging operational problems from civil unrest and hostile campaigns to terrorism.

“We are not experts in loan origination or financial trading, or terrorism,” Srulevich says. “We are experts in analyzing the data and the correlation between the data points. What's beautiful is that the algorithm is exactly the same for everyone. We don't tweak it, we don't use a screwdriver to make it sharper for one application or another, but we can use it for various different sources. And it gives the organization the ability to ask an unlimited number of predictive questions, and to experiment.”

We are experts in analyzing the data and the correlation between the data points. What's beautiful is that the algorithm is exactly the same for everyone.

Endor recently launched an on-premises version of its software platform for government agencies with access to very sensitive information. Importantly, this version also will let large corporations derive information across data sets that can never leave the premises at all.

For example, Endor is working with one global conglomerate with many divisions that are forbidden to share their data. “The data also is encoded in a completely different way in each division, so you don't know that John Smith here is John Smith there, Altshuler says. “But this company still wants insights that are an amalgamation of everything they can extract out of their data that they spent so much time and energy curating.”

Endor's on-premises software will become the first offering anywhere that “can enable companies to benefit from all their data without disclosing anything about the data,” Altshuler says. “Nobody will disclose anything that they shouldn't be disclosing, but they will all benefit significantly from collaborating.”


Linking to blockchain

Endor is expanding its reach by exploiting another disruptive technology: blockchain databases, in which blocks of records are chained together with encryption methods enforced by global networks of servers to prevent unwanted access.

The company's blockchain technology is designed to deliver a new way for small companies to safely monetize their data, or to tap relatively inexpensively into data from other organizations. The first product, Endor Protocol, gathers information about specific ERC20 tokens, which are employed in the Ethereum blockchain cryptocurrency system. Endor also is working to extend such capabilities broadly to other forms of business data beyond blockchain.

With this approach, as with the on-premises version of Endor Enterprise, any organization will be able to participate but the data will never leave its computers. Instead, the Endor platform will pick up useful patterns from the data by analyzing social physics patterns.

This blockchain platform will give small businesses a chance to join in growing networks whose benefits become greater as they grow, Srulevich emphasizes. “When we combine blockchain with social physics, we use decentralization to enable small businesses with the opportunity to participate in a network effect that has been available only for the Googles and Facebooks of the world,” he says.

Overall, says Srulevich, “Endor wants to become the global standard of predictive analytics for organizations, by enabling everyone to have access to data and to understand things that will give them an edge.”

Key will be the speed with which its software can analyze any kind of big data. “Today, everyone has so much data, but they don't have the manpower,” says Altshuler. “What you need is an automatic machine that will enable a different way of analyzing this data. Organizations that will not transform the way that they automate this process will not be able to compete with those who do.”