Founders: Zohar Bronfman and Noam Brezis
Investors: S Capital VC, Dell Technologies Capital, GGV Capital
Year founded: 2018
Employees: 130
Capital raised: $116 million
Many Israeli start-ups are engaged in developing tools meant to help organizations apply machine learning (ML) to years of collected data, and generate forecasts for improving sales, retaining customers, and increasing efficiency. The vast majority of these startups offer a one-stop solution for one stage in the process, such as organizing data prior to input, training predictive models, or monitoring models over time. But Pecan AI provides an end-to-end solution, and moreover, promises to enable business intelligence, operations, and revenue teams to handle the entire process themselves, without needing to hire expensive data scientists.
Pecan AI's software, which falls under the Auto ML category, simplifies the use of machine learning through a library of pre-configured models. The models make it possible to predict customer behavior, such as a customer who is about to leave and with whom a preventive interaction should be initiated, or alternatively to whom it is worth making a special offer because the chances of making a sale are good. All that a Pecan user needs to do is to choose the query that interests him from the Pecan AI library and follow the system's instructions.
Pecan AI was founded in 2018 by CEO Zohar Bronfman and CTO Noam Brezis, who both hold PhDs in computational cognitive neuroscience from Tel Aviv University, and who left academia in favor of entrepreneurship. Since its establishment, Pecan AI has raised $116 million, and it currently employs a staff of 130. Among its clients are pharmaceutical corporation Johnson & Johnson, Canadian insurance company CAA, and Israel’s Phoenix Assurance Co. "In the last two years, our revenue grew threefold each year, and this year it doubled, bringing us to a significant revenue level, in the millions of dollars," Bronfman says, but does not elaborate.
Business analysts, the target market that Pecan AI addresses, already use business intelligence software such as Qlik Sense and Tableau, with the aid of which, for example, a segmentation of lost customers by gender or residential area can be obtained. The use of machine learning enables Pecan to take this to a higher level, and not only give a breakdown of customers who have left, but also to predict who is about to leave. "Business intelligence is a process for examining hypotheses put forth by the analyst in relation to the past. The answers that business intelligence gives are always at the group level, for example, whether more or fewer white men register for a service," Bronfman notes. "Machine learning provides predictions at the level of the individual customer, but the logic guiding these forecasts is not transparent and is hard to explain."
Over the years, no few criticisms have been levelled against the Auto ML approach, ranging from claims that the quality of the results is not high enough, to inflexibility and the inability to tinker with the models. An professional data scientist would probably not be caught using Pecan AI or any off-the-shelf Auto ML solution - which is just fine, as far as Bronfman and Brezis are concerned. They appeal to companies that cannot afford to hire data scientists, or to teams within the company that will never be assigned such scientists, but still want the business benefits of machine learning.
"One company wanted to check to which of its former customers it was worth showing advertisements. Our model divided past customers into those who would never go back to the company and those who were going to come back anyway; two groups at which there was no point in targeting ads. The third group was those who might return and on whom its was worth spending more money. In the end, the model reduced the company's marketing expenses by $3-4 million, and still raised the number of returning customers by 7%," says Bronfman.
There are plenty of Auto ML solutions on the market, such as that of US-based Data Robot, which has raised about a billion dollars to date. Bronfman and Brezis say however that most of the solutions on the market appeal to data scientists who know programming, while their solution does not require such knowledge. A more direct competitor to Pecan AI is Israeli startup Noogata, which has raised $28 million, and also offers software that does not require writing code. Further competition comes from business intelligence providers like Qlik Sense, which are adding machine learning functions to their products.
According to Pecan AI's founders though, their biggest challenge is not the competition, but educating the market. "Business teams understand that machine learning is important, but don't know that they can use machine learning themselves."
The Most Promising Startups rankings are part of the annual Enterprise Technology Summit held by "Globes" and JP Morgan. .
Published by Globes, Israel business news - en.globes.co.il - on December 14, 2022.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2022.