AI Startup Inferess Launches Supply Chain Analytics

By Sanford Bragg February 9, 2021

London-based Inferess is releasing AI-driven supply chain analytics this month, challenging more established providers of the data.  The firm leverages natural language processing and machine learning with the goal of creating more timely data than current sources dependent on manual processes.

Inferess’s new supply chain analytics cover 10,000 US public companies at launch.  As with current suppliers of supply chain data, notably FactSet’s Revere Data and Bloomberg, the starting point for US supply chain data is company 10-K filings which, under the SEC’s  Statement  of  Financial  Accounting Standards No. 14 (SFAS 14), require disclosure of customers which represent 10% or more of a company’s overall sales, and the revenues associated with that customer. 

However, 10% is a high bar and in a typical year only about 1,000 such relationships are disclosed through SEC filings.  The trick is to supplement the data with additional relationships disclosed in earnings calls, press releases, company websites and capital market presentations.  FactSet says it uncovers about 25,000 relationships a year from supplemental sources.  Inferess is seeking to streamline that process using NLP.  The firm also uses probabilistic methods to derive maximum-likelihood estimates of the revenues associated with relationships disclosed from supplemental sources.

For relationships gleaned from unstructured or text data, Inferess provides a confidence score for the degree of accuracy for the relationship.  The confidence score is comprised of factors such as the number of co-occurrences between two entities, the number of words separating the mention of two entities and the order in which the entities appear, among others.  The company says it uses high thresholds for determining relationships which “yields low-recall but high-precision relationship classifications”.

Inferess’s rollout strategy has been to initially approach quantitative hedge funds and other quant-oriented managers which use supply chain data as inputs for trading strategies.  It now begins a second phase this month with the release of analytics designed to appeal to a broader set of fundamental asset managers.  Inferess issued a white paper estimating the alpha associated with various trading strategies using supply chain data.  As one example, it found a Sharpe ratio of .926 related to a momentum long/short strategy using second-degree customers with a six-month lag.    

Inferess was founded four years ago by Vishal Puri, who previously worked as a derivatives analyst at UBS.  The company first focused on sentiment scores derived from news sources, a product that it has used to fund development of its supply chain offering.  Financing has been bootstrapped from revenues from customers of its news analytics service. 

The company currently employs six full-time staff, not counting part-time sales staff.  It is currently recruiting for two open positions, both being software engineers.

Our Take

Supply chain data is not new – FactSet’s Revere Data has been around for two decades – but Inferess is seeking a fresh approach leveraging artificial intelligence.  Its US coverage is nearly the same as that of FactSet, although FactSet covers an additional 20,000+ non-US firms.  Inferess says it will be able to add non-US coverage quickly based on its technology.  “We’re not encumbered by legacy systems or by data entry staff based in the Philippines,” said Vishal Puri in an interview. “We can scale up to different languages quickly.”

More broadly, Inferess is an example of increasingly creative applications of NLP.  As sentiment analysis has become more commoditized, firms like Inferess are diversifying into less crowded areas.  Rival Accrete is also leveraging AI to create supply chain data, although theirs is based on import/export data to monitor trade flows.  AI not only facilitates more timely data collection but also the mashup of multiple data sources.       

Related Articles

  1. Study: Web Scrapers Represent a Quarter of Internet Traffic (12)
  2. COVID-19 Push to Pass US Federal Privacy Law (12)
  3. Appeals Court Rules that Web Scraping Bots May Be Illegal (12)
  4. UBS Evidence Lab Hiring Slows in 2020 (12)
  5. Alternative Data Sourcing Firm Neudata Launches Compliance Service (12)
  6. Survey: Hedge Fund Demand for Alternative Data Remains Solid (12)
  7. White Paper: Alternative Data Intermediaries Have Mis-Aligned Incentives (12)