Data Trends for Investment Professionals

GO TO QUANDL.COM ⟶

How Email Receipts Predicted GoPro’s Q3 Earnings

The street got a big surprise last week when GoPro reported earnings. Shares dropped sharply in Friday morning trading after the company reported a loss of 60 cents per share on $240.56 million in revenue. Analyst expectations were much milder at a loss of 36 cents per share on $314.06 million in revenue. GoPro cited production issues as the cause for both the third quarter miss and the guidance for Q4.

Prior to the earnings announcement, most pundits were bullish – or at least neutral –  on GoPro: see MarketWatch coverage, AmigoBulls, and this note from Raymond James.

This is unusual. Modern markets are astonishingly efficient, and large surprises are very rare. Analysts have hundreds of data points, comprehensive business models, and frequent “guidance” from the companies they cover.  As a result, their forecasts are remarkably accurate.

This leads to the question: what could investors have done to foresee this kind of miss?

The answer lies in electronic receipts extracted from over 3 million inboxes. As it turns out, Quandl’s Email Receipt database told a different story from the consensus.

The analyst consensus was that GoPro would report a +42% Q/Q change in revenue.  In our own analysis using the Quandl Email Receipts data we started out by looking at receipts mailed out by GoPro, representing sales made directly on the GoPro website.  This set of receipts showed Q/Q growth of 25%; not quite as good as the consensus, but not terribly bad either.

But then we expanded our analysis to include third-party retailers of GoPro products, and a different picture emerged.

Sales via Best Buy were solid, but sales on Amazon, Walmart, Target and other retailers were low.  GoPro sales volume over a fixed cohort of Amazon users actually declined over the quarter, by 5%.  Amazon is by far GoPro’s biggest channel, accounting for ~50% of GoPro’s total sales, so this decline was significant.

Blending together sales across all channels, and weighting for relative volume, we estimated that GoPro’s revenue grew by +15% Q/Q in the third quarter of 2016.

The drop in our forecast was driven largely by declining Amazon sales:

GoPro revenue from Amazon saw a surprising drop in Q3.

Historical regression yields an R-squared of 85% for our blended estimate, large enough to be reasonably confident of our prediction:

Even more encouraging was the fact that our historical estimates seemed accurate across various sales channels:

The production issues GoPro identified on their earnings call had a direct impact on their ability to fulfil Amazon orders, which led to that channel’s underperformance. This was clearly reflected in the Email Receipts data.

We know that a single data point is not enough to drive an entire strategy. But when an alternative dataset indicates a departure from the consensus, the key outcome – whether it’s to an analyst or an investor – is to take a closer look. This is a new dataset that should be considered an arrow in the quiver for anyone making trading decisions with e-commerce-focused public companies.

The guidance for Q4 is continued underperformance, but with a return to profitability in 2017. What will the data say?

Abraham Thomas contributed to this post.

Contact Us For a Free Trial of This Data

VIEW COMMENTS

Leave a Reply

Your email address will not be published. Required fields are marked *


6 Comments
  • Jon Benp says:

    If it’s so accurate, how come this article only came out after earnings were released and not before? It’s always easy to find data to justify the view after the fact.

    • Raquel Sapnu says:

      You’re absolutely correct; it’s easy to cherry-pick favourable results post facto. It’s much more impressive to make predictions before the event.

      That being said, in this particular case, we’re constrained by the fact that we have customers who pay high prices for the underlying dataset, so we can’t really publish the key results for free on our blog.

  • Rob F says:

    How do we get the Quandl Email Receipt data?

  • MVSELL says:

    What do the axes on the graphs represent? This article could benefit from a little more clarification.

    • Raquel Sapnu says:

      Thanks for your question, MVSELL. The scatter plots all show Q/Q change in GoPro sales as predicted using email data, versus the same number as officially reported.

Fix This
Created with Sketch.