B23 is predicting that revenue for a corporation traded on the New York Stock Exchange (“NYSE”) will experience a 34.94% decrease for the business quarter ending March 31, 2017, as compared to previous Q4 business quarter of FY2016.  Comparing the previous year’s Q1 from FY2016 to our prediction in revenue for Q1 FY2017, we calculate a decrease of 20.94%. The quarterly revenue results are scheduled to be released to the public several weeks from now in May 2017.

B23 is not a financial services company and our primary business is not to predict stocks. B23 is a software company that has built a data automation and analytics platform to help businesses securely and quickly use data for competitive advantage. Our analysis is even more intriguing as at the time of writing [April 20, 2017], the general consensus in the financial media and financial analyst community is that of a bullish outlook on the performance of this stock.  Many factors influence the price of a stock, and it is important to note that our analysis is solely focused on sales revenues and its extremely high correlation to specific consumer behavior of the business we analyzed.  Since this company we analyzed has a distinctive seasonality to their business, we looked at several dimensions of their historical performance.   A graphical view of the difference between consecutive quarterly revenues is shown here:

Graph of Revenue Decrease

 

The significant decrease occurs in Q1 FY2017, and we show graphically the net decrease as compared to previous Q1 revenues in FY2016 and FY2015.  Comparing Q1 FY2016 with Q1 FY2015 shows relatively consistent revenue decreases.  Due to seasonality, we assume that this business’ Q1 revenues typically decrease relative to the prior quarter—but our predictive analysis shows a much greater decrease in FY2017 than what the company experienced in prior years.

A summary of analysis includes:

  • Comparing Q4 FY2016 to Q1 FY2017, we predict revenues will decrease 34.94%. This compares to a decrease of 17.78% from the same time last year, or from Q4 FY2015 to Q1 FY2016.
  • Comparing Q1 of FY2016 to Q1 of FY2017, we predict a decrease of 20.94% in revenues. This compares to a modest growth of 2.63% when comparing the same time the year before, or from Q1 FY2015 to Q1 FY2016.

Using a variety of data sources, B23’s Data Science team has performed a rigorous analysis and several machine learning algorithms to reach this conclusion. As part of our due diligence, we have correlated years of historical data on consumer behavioral patterns to known and documented “ground truths’ associated with the analyzed business. This resulted in a correlation coefficient of 0.93 between consumer activity associated with the business entity and that business’ previously announced earnings statements.

Historical Context Using Alternative Data

The term “alternative data” typically refers to non-financial data that can be used by financial services firms to gain unique insights.  At the moment, using so-called alternative data to predict the financial performance of businesses is gaining increasing attention.  There are several popular and well-documented success stories of how alternative data can be used to predict business performance.  These stories include the use of satellite imagery to count cars in parking lots, and Foursquare’s public relations coup validating Chipotle’s revenue drop after their E. coli experience.

Approach and Methodology

Our approach and methodology involves four discrete components including data, algorithms, software tooling, and people. B23’s combination of these components encapsulated within the B23 Data Platform is the unique capability we offer our customers.

Without quality data that can be synthesized efficiently, no machine learning or artificial intelligence (“AI”) algorithm would be capable of providing predictive insight.  Using a variety of meaningful data sets was a key factor in our ability to accurately predict the performance of this stock.

B23 has developed expert level knowledge and insight related to location-based data sets, places-of-interest (“POI”) data sets, customer sentiment-centric data sets, as well as other esoteric data sets.  Not all data is equal!  An intimate understanding of this data weighed against the strengths and weaknesses of the data providers played a leading role in the accuracy of this prediction. Knowing what data to use to solve specific problems is critical.

The democratization of algorithms over the past several years has allowed our team of data scientists to access sophisticated machine learning algorithms.  In our analysis, a series of sequential algorithms was applied.  Each stage of analysis required an appropriate algorithm to perform a specific set of analytical tasks.  There is no one algorithm to rule them all! An often-overlooked aspect of data science is the geospatial analysis domain. Analyzing billions of data points is now only practical using distributed processing systems.  A key aspect of understanding consumer behavior involves the broader understanding the aggregate movement of people.  Understanding where and when people are moving inside logical geographical boundaries continues to be a significant analytical challenge.

We leveraged Apache Spark to perform many aspects of our analysis, especially in our geospatial analysis pipeline.  In minutes, the B23 Data Platform can automate the secure provisioning and enablement of data pipelines directly to Apache Spark.  Spark is not a panacea, but it is an effective tool when combined with quality data inputs, and curious and intelligent data scientists.

The role of the B23 Data Scientist was a crucial factor in our ability to effectively formulate our prediction. Our data scientists are a unique blend of software developers and statisticians.  Too often, we interview job candidates with strong mathematical and statistical backgrounds, but who lack software development experience.  Developing software using higher order programming languages like Scala, Java, and Python is a critical factor to building scalable applications.  Notebooks and command line interfaces are effective for experimentation purposes, but not a substitute for creating performant software applications.

What’s Next?

We have deliberately not named the company we analyzed in this blog, or described more detailed characteristics of the company under analysis.  As we near the scheduled earnings announcement in May, our intent is to disclose the business and stock ticker immediately prior to the earnings release.

Stay tuned to our Twitter feed at @b23llc as we near this exciting confirmation of our prediction!

B23 will also be speaking about our data platform and its use in the financial services industry at the Fintech Exchange 2017 on April 27, 2017.

About B23

B23 enables businesses to gain competitive intelligence from data.  We have an axiom that all business are data businesses whether they know it or not. As part of our journey to enable our customers to fully leverage data for gaining unique insight, we acquire a variety of data sources and provide these within our B23 Data Platform. These data sources, when combined in creative, insightful, and sometimes unpredictable ways provide unique value.  Our Cloud-based B23 Data Platform allows our customers to securely access and analyze many different types of data. Based on our innovative software automation, it typically takes our customers minutes versus days, weeks, or months to start analyzing data.