Common Data Science Questions During a Stax Due Diligence for our Private Equity Clients

Many businesses have been affected by Covid in unexpected ways. For some of these businesses, the recent downturn has made Covid feel different from a “normal” recession. We are often asked to examine a business and understand what happens if the economy goes into a “normal” recession.

The best way to accomplish this analysis is with historical data. Even in 2021, we have been able to collect data going back to 2006 to understand how some potential acquisitions reacted to the 2007-2009 economic downturn. This often looks different from how businesses have been affected since the beginning of the 2020 downturn. With long-term historical data, we can make better projections for what will happen through the rest of the Covid downturn and beyond.

Historical data back to 2006 may not always be available. In some cases, businesses were founded more recently. In these cases, exposure to the business cycle can still be measured. The most common workaround we can utilize is to look at more recent local data. Some measures of economic performance can be measured locally, such as unemployment, household savings rates, GDP growth, etc. Using multivariate analysis, we’ve had success measuring the effects of macroeconomic indicators at a local level, which can be a good guide to how the entire business would react to changing national conditions.

Some clients are surprised that we can conduct data analysis to answer competitive questions during a diligence. The target rarely has data on their competitors’ performance. Instead, consulting firms need familiarity with common data sources to gather competitive data.

This data is usually limited but can answer a specific question, like how the target company is perceived compared to competitors. A quick and dirty answer to this question can be found through data that can be purchased about online ratings. For example, average ratings on Yelp, or even just the raw number of Yelp reviews can be a proxy for product quality or market share respectively. Although online ratings have many limitations, they can be worth a look especially in the absence of more traditional options (annual statements, industry reporting, etc.). Online ratings may also include reviews, which are an opportunity for text mining against competitors to identify areas where the target company may be differentiated from its competitors.

It’s common in a diligence project to estimate the Total Addressable Market (TAM) for a business. This question is sometimes answered without deep data science techniques, but other times, more rigor is required. For retail locations or other chains with a geographical footprint, it can be difficult to say how many potential favorable locations are out there.



For these businesses, we have completed deep TAM analysis even during a shortened diligence timeframe. To do this, we have partnered with data vendors to have quick access to granular geographic data, including demographics, foot traffic, purchase behavior, potential competitor datasets, etc. By combining this data, with details about where the target company is located, we can estimate where the company can go next. In some cases, we have utilized the target company’s customer data to estimate each location’s catchment area (the range where its customers come from), then apply that catchment estimate to other potential locations using geospatial software. Team members with various analytical skills support this kind of analysis: some are experts in geospatial software, others have many years of experience building statistical models that predict a site’s performance.

The most common questions we see in a diligence timeframe are around getting to know customers. These questions are so frequent that there are common acronyms for some of the metrics utilized throughout the industry (e.g., AOV for Average Order Value, etc.). These metrics fall into three main categories: spend, retention, and segmentation.

• How much do customers spend in an average order (Average Order Value)
• How much do they spend over their lifetime (Lifetime Value)
• How does this break down by product/service category
• How has this changed over time
• What sorts of bundles are products purchased together
• What do new purchases look like vs. returning customers' purchases

These metrics are helpful for comparing a company against its peers or to understand the company’s success or failure of expansion into new products.

• What percentage of customers are one-and-done (we never see them again)
• What's the average duration that returning customers continue to buy
• What is their frequency of purchase
• What percentage do we lose each month

These metrics help show how customers perceive the quality of a company’s products. If few customers return, that suggests a poor customer experience.

• Which of the customers spend more/less
• Which customers are likely to be lost/retained, etc.
• Which customers at the target company are most complementary with my current portfolio companies' customer base (merger opportunities)
• Are there unexplored segments of customers that we can identify through unsupervised learning (e.g., cluster analysis) to expand to new markets

We use these segmentations to understand if we can change customer behavior, or if we have some hypothesis about a potential merger and how customers compare between two entities.

These four questions (macroeconomic, competitive, market space, and customers) are the most common questions we encounter in private equity due diligence projects. It is rare that we would need to answer all these questions in a single diligence. It’s usually clear early on which are the most important questions to answer in diligence, and the analysis that needs to get done follows from those questions. We have long partnered with our private equity clients to help translate the strategic questions about the business into an analysis plan for a due diligence timeframe. It helps to have this kind of head start.

Joe Anderson is Senior Manager of Stax Data Analytics and has 15+ years of experience leading data science teams and developing analytics tools across multiple industries. As part of the Data Analytics team, Joe designs, extracts, and creates effective data science solutions, including products and resources to improve operational efficiency.