Heuristic for Business Intelligence

Published Date: 2019/10/10

Machine Learning solves known problems. Heuristic solves unknown problems.

Solving Problems with Machines

Statisticians model a known problem space with historical data to solve similar problem by matching it against the model.

Experts derive rules from experience in a problem domain to solve problems.

Computer Scientists enable a machine to learn with heuristic to solve problems on its own by discovery and accretion.

Limitation of Machine Learning

Quantitative analytics use statistical algorithms and principles of stochastic processes to build models, and then derive insights towards a problem by matching given variables against historical model. When variables are unknown and problems are new, Machine Learning cannot be used.

Deriving Business Intelligence from Natural Language Documents

Business Intelligence is insight derived from business practices. As much as business results are mostly quantified in a balance sheet, the result does not articulate the supporting facts behind the numbers. Often times beating earning expectation said little about the weakness and strength of the business operation. In general, executive summary and investor concerns are discussed during the earnings conference calls. MDA section of 10-K and 10-Q filings provides an official written discussion of the quarterly and annual results. Qualitative analysis of these documents is necessary to obtain the details on strategies and operations.


An analyzer with heuristics is fundamental to derive insights from earnings call transcripts and MDA embedded in 10-K and 10-Q documents. To accomplish this, SiteFocus has successfully implemented a generic solution for machine to understand the context and semantics of natural language documents. Using this technology*, together with a fully automated end-to-end streaming network, SiteFocus has created a WEB based platform for business strategists to gain insight into these earnings calls transcripts and MDA section of 10-K and 10-Q filings instantly, namely Earnings Intelligence (EI).

Applying Heuristic on Earnings Intelligence

Earnings Intelligence as a service is a search engine that analyzes earnings call transcripts and MDA section of 10-K and 10-Q filings of publicly traded companies in North America. Qualitative data is automatically analyzed for context and relevance. Result of the analysis is accessible by sector, industry, keyword, context, and semantics.

Example of EI heuristic:

To demonstrate Earnings Intelligence’s DNA on heuristic, here is a Tweet sent by EI automatically after it autonomously analyzed ERYP’s earnings call on Sept. 18, 2019:

According to EI’s heuristic findings, one of the tuples stated:

(metabolism,pancreatic), (manufacturing,pancreatic), (europe,pancreatic)

The tuples are composed by EI after analyzing ERYTECH Pharma S.A.’s Sept. 18th earnings conference call transcript. The above is the top three business focuses concluded by its heuristic analysis. The following is the excerpt that corresponds to the first tuple:

“You can see and not surprisingly the same profile as at our last update call, we are still the red cells company focused on cancer metabolism with … late-stage trials in pancreatic cancer Phase 3, triple negative breast cancer Phase 2, and IST in acute lymphoblastic leukemia in Phase 2. We have industrialized and scalable production with now two own manufacturing sites, one in Europe that we already had, but have now extended and then one in Princeton for the supply in U.S. that came on stream recently”

“So, we launched this study in September 2018, which is the pivotal Phase 3 study in Europe and U.S., about 500 patients to enroll, comparing standard of care chemotherapy to standard of care chemotherapy with eryaspase added, our product, overall survival end point. So, we already at the last call I think we had all 11 countries accepting our clinical trial”

The above excerpts are quotes that represent the focuses per management of ERYP during the conference call. It was considered as top challenges by EI’s heuristic. The selection of words that made up these tuples are results derivedby EI without any influence from pre-set preference, parameter, hint, guidance, dictionary, or sentiment. In fact, none of the words selected in these tuples are sentiment related words. It is important to point out that EI does not rely on sentiment analytics when it performs its NLU heuristic. By doing away with the sentiment aspect in analyzing context and semantics, it enables EI to provide qualitative assessment without any bias.

More on Earnings Intelligence ...

For articles of Earnings Intelligence analysis, please visit www.bizpage.com. It offers timely impromptu analysis and excerpts on strategies and growth related matters disclosed by executive teams of public companies during earnings calls.

*ELAINE is a Symbolic Artificial Intelligence System for Natural Language, for details, visit: https://www.sitefocus.com/sitefocus-elaine.

This article draws references from Earnings Intelligence, a service provided by SiteFocus