A Framework for Better Understanding Equity Valuations

 

Introduction

There is a lot of sentiment online and when I am talking to friends about a certain stock's share price that goes along the lines of, “No way I’m buying Zoom with a P/E of 1,000!”, or “I can’t believe Cloudflare is trading at 50x sales, that's crazy!”. Are these trading multiples outliers, general market euphoria over technology stocks, or are there explainable reasons for these outsized trading ratios? This post intends to provide a few hypotheses and better understand the logic (or lack thereof) for these trading ratios. 

This analysis started out by pulling standardized financial data from SimFin (no affiliation), removing companies that have less than $100 million in annual revenue to reduce noise in the data, removing companies with incomplete data in SimFin’s database, and only focusing on companies that have their primary stock exchange listing in the United States (Nasdaq, etc.). For the sake of this discussion, I’m focusing on the Application Software sub-sector. These parameters resulted in 147 companies to be included in the initial data set.

Initial Thesis

My initial hypothesis is that the financial metrics of a software company is correlated to its trading multiples, so I ran a regression of common financial metrics against Enterprise Value / TTM Revenue and PE ratio.

Here are the factors that I included in the regression and the reason for each one (skip this section if you want to just read the analysis):

1.      Gross Profit Margin – Many technology companies are not profitable nor do they intend to be profitable until their growth slows and their TAM matures. This results in net profit margin, operating margin, and EBITDA margin producing inconclusive data hovering near zero that creates noise in the data. The intent of including gross profit margin in this analysis is to use this as a proxy for the unit economics of each company – i.e., Salesforce has ~10% lower gross profit margins than Adobe despite directly competing in many customer segments – do investors reward this additional gross margin or not?

2.      EBITDA margin – Arguably the cleanest view of corporate profitability because it removes noise of one-time events, but the downside is this is a non-GAAP metric that is subject to interpretation that will vary by company

3.      EBITDA ($Millions) – I actually initially included this by accident, but decided to leave it in the analysis to test if the raw data of $millions of EBITDA for each company is a better proxy as a combined size and profitability metric for understanding if the market rewards company size and profitability on a standalone basis.

4.      Sales & Marketing (% of Revenue) – Measures the amount of S&M that the company is investing into the business relative to the total size of the business. An outlier on the high-side may indicate that the company is significantly investing in future S&M resulting in outsized future growth and an outlier on the low side may indicate that the company has very strong unit economics (high LTV:CAC) and a very efficient customer acquisition model.

5.      R&D (% of Revenue) – Measures whether the company is investing heavily in new technology, products, and features. An outlier on the high-side is potentially a Product-led Growth business (PLG) that instead of focusing on S&M to grow future sales, they instead focus on building an amazing product and drive their growth through a bottoms-up sales model (Slack, Snowflake).

6.      Revenue Growth Trailing Twelve Months (TTM) – The primary reason for running this analysis was the thesis that high growth companies receive outsized trading multiples - the market technology companies for their growth with profitability being viewed as a secondary metric that will be achieved after the company matures (i.e., Amazon).

7.      Market Capitalization – This measure was included to understand if company size can partially explain the trading multiples – does the market pay a premium for larger businesses? This could partially be because large investors do not see enough liquidity in small-cap names, fund parameters often restrict investments in small and microcaps, and market leaders such as Microsoft are often larger than their peers.

Data Analysis

Here is the regression analysis summary that I ran in Microsoft Excel (tutorial here) of the above financial metrics measured against EV / Revenue:

As a quick primary on regressions, P-values below 0.05 or 5% are considered to be statistically significant measures of correlation. I was not surprised to see that revenue growth had the lowest P-value, or statistical significance, but I was surprised to see that Market Cap and EBITDA ($M) had statistically significant P-values as well. My initial theory was that Gross Profit Margin would have the second strongest significance after Revenue Growth.

Here’s the regression summary of the same financial metrics measured against P/E: 

These P-values indicate that none of the financial metrics explain a strong correlation with predicting P/E ratio for software companies. I would expect other, more mature, profit-driven industries such as industrial manufacturing would have a higher correlation to the P/E ratio. There may be other financial metrics that are correlated to P/E that I did not analyze, but my data did not provide any conclusive results. Given the disappointing results of the P/E regression, I will focus the rest of this analysis on EV / Revenue analyses given there was some level of significance in the data.

Deeper Dive on EV / Revenue

Let’s take a look at re-running the EV / Revenue regression but only analyze the 3 financial metrics that had statistical significance in the original analysis: Revenue Growth Rate, USD millions EBITDA and market cap.

Although the P-value is indicating that all three values are statistically significant (<0.05), the coefficients in the regression are indicating that we may want to further reduce the number of factors in the regression. The column “Coefficients” provides the factor model that builds the expected value for each company, which is its “y = mx + b” formula. The version of the formula for this regression is Predicted EV / Revenue = [company market cap] x [market cap coefficient] + [company EBITDA ($M)] x [EBITDA ($M) coefficient] + [Company Revenue Growth] x [Revenue Growth coefficient] + Intercept). EBITDA ($M) and Market Cap ($M) are both showing coefficients around zero, which means that they are insignificant in testing our hypothesis that these metrics are partially explaining EV / Revenue. I was hoping that refining based on P-values would increase the Coefficients for the remaining metrics, but that did happen to be the case.

Analyzing TTM Revenue Growth (%)

Setting aside the two metrics with insignificant Coefficients allows us to focus on the one metric that had a statistically significant P-value and a meaningful Coefficient, which is Revenue Growth (TTM). So let's re-run the regression with this one metric:

Now that the analysis has been refined to a strong P-value and Coefficient, it's time to dig into the data. A scatterplot of the underlying individual company data provides a better understanding of all of these numbers (see larger image of the chart here):

If this chart isn't making sense to you, here's a fun example. Zoom’s predicted EV/Revenue of 72.5x is close but below the actual trading multiple of 84.7x, or a difference in the residual of ~12x. On the contrary, HealthEquity has a predicted value of 50.2x relative to the actual trading multiple of 7.9x, which indicates that it may be undervalued. 

Stack ranking each of the companies based on their residual values provides more value in telling us the relative under and overvaluations for the industry. I've sorted each of the companies in descending order based on the residual values with the higher residual representing overvaluation and a lower residual representing undervaluation:

So is Zoom overvalued? According to this preliminary analysis, the share price is overvalued, but there are other companies that have more alarming valuations such as Okta or Zscaler.  HealthEquity, Square and Zynga are considered to be the three most undervalued companies.

Conclusion and Food for Thought

One of the challenges of analyzing a single metric is that the R Squared of EV/Revenue plotted against Revenue growth only explains 39.6% of the correlation (as indicated in the top right-hand corner of the scatterplot). The majority of the correlation is unexplained and further work is required to test against other financial metrics, or perhaps refine the initial parameters such as excluding gaming companies or companies that have been significantly impacted due to COVID.

Share price performance nor forward-looking data was available as apart of this dataset. I would have much preferred to analyze Wall Street Consensus estimates and/or company provided guidance (when available) for each of the financial metrics. 

The raw data set that I used is not perfect. You may notice that a few of these companies have been acquired, and there's likely some incorrect financial data that was provided by the data provider. Even if the raw data was perfect, it's nearly impossible to entirely scrub the data unless you manually make all of the necessary adjustments. For example. Square's 77% revenue growth is primarily driven by the explosion of Bitcoin trading on their Cash App product, which contributes to their undervaluation in the regression. Square is required by the SEC to report gross Bitcoin trading volume as GAAP revenue, although industry analysts measure Bitcoin gross profit as an adjusted net revenue measure for Square. Therefore, a good next step is to perform a deeper dive on each opportunity such as my last post on Square.        

Disclosures: This is not intended to be investment guidance, please perform your own research before making any investment decisions, this analysis is strictly intended to be an educational framework for how to better understand equity valuations, and these opinions are personal in nature and not representative of my employer.