Staying Ahead Of The Market: Estimating iPhone Sales

| About: Apple Inc. (AAPL)
This article is now exclusive for PRO subscribers.


Search volume on Google for the word "iPhone" is highly correlated with actual iPhone sales.

Given that financial reports are always released with a delay, search volume could be used to get a sense of iPhone sales trend and potentially stay ahead of the market.

In this article, I propose an investment strategy based on search volume on Google and find that it can deliver superior trading profits.


The internet has transformed how we obtain and process information to make everyday life decisions. Tools like Google's (NASDAQ:GOOG) search engine have greatly reduced transaction and search costs, impacting consumption patterns.

For some consumer related industries, it makes sense that people use Google's search engine to familiarize themselves with the product before making a decision. I recently had to trade my smartphone due to an unfortunate incident. Before going into my local carrier's store, I took my time to read about the pros and cons of the phone I was hoping to buy. After some research, I decided it was a good idea and ended up with a new smartphone. I'm probably not the only one who buys certain items like this.

Apple (NASDAQ:AAPL) depends on iPhone sales for more than half of its revenue. As you would expect, iPhone unit sales are on the spotlight on every quarterly report. However, we have to wait approximately a month to find out in the 10-Qs and 10-Ks exactly how many iPhones were sold. If you could find a leading indicator and get a sense of the trend of iPhone sales before the actual report, you could stay one step ahead of the market. In this article I propose to use search volume on Google as a leading indicator.

Google Trends provides weekly and monthly data on search volume of a great variety of industries with a great advantage: it has a maximum delay of one week.


Google Trends show how often a particular item is searched relative to the total search-volume across various regions of the world. It offers weekly and monthly data, depending on the popularity of the item. For this article, I aggregate weekly data to match fiscal quarters.

Google Trends does not provide absolute search volume. Instead, data is normalized and presented on a scale from 0 to 100. It assigns the highest search volume in the period a value of 100 and presents every other search volume as a fraction of this value.

For example, compiling weekly data of Google search volume for the term "iPhone", the week with the most volume since 2004 up to this date was the week ended on September 14th, 2013. The data on the graph below shows that search volume is highly cyclical with peaks around September. It also suggests that its growth rate has flattened and could probably be headed south in the near future.

*Source: Google Trends

For readers more familiar with econometrics, more detailed information can be found on a paper by Choi and Varian. These authors show how Google Trends is correlated with automotive sales, among other variables.


My argument depends heavily on my guess about the behavior of the average internet era consumer. My intuition is that before purchasing an item with certain conditions, you'll probably want to "google it" first. Hence, before testing for a correlation between Google Trends and iPhone sales, I first address an important disclosure.

I don't expect all companies' quarterly results to be equally susceptible to be estimated using Google Trends. I expect only some consumer related companies to be a viable candidate. This estimate should be more useful to the extent that the company in question:

  1. Depends on a few flagship products for financial success. Back when iPhone sales accounted for only a fraction of the revenue of AAPL, guessing its trend correctly was probably less useful from a trader's perspective than it is today.
  2. Is exposed to cyclical consumer markets. Since we are interested in using Google search volume as a proxy for actual sales, we have to focus on industries where consumer decisions are more responsive to the economic cycle (i.e. more elastic in economics lingo). It would probably be pointless to try to estimate gasoline consumption using Google Trends, since it may depend more on necessity.
  3. Offers heavily branded and differentiated products. If you don't care about the brand of the product, for example, your choice of gum, you'll probably not waste your time searching for the right one on Google.
  4. Depends on relatively expensive products. The more money you have to put down to buy something, chances are you'll take more time to make sure it is the right decision.
  5. Depends on customers who have access to the internet. This one explains itself.

Estimating iPhone Sales

The smartphone industry reasonably adjusts to most of the conditions above. It is highly cyclical and consumer oriented. Its products are expensive enough and branded enough to justify gathering at least some information before making a decision.

Given that AAPL is heavily dependent on its iPhone sales for most of its revenue (57% as of 1Q14), it is a suitable candidate to propose later on an investment strategy based on Google Trends.

Using Google Trends, I track the search volume for the term "iPhone" under the "Product Line" category. For iPhone unit sales I use the company's 10-Qs and 10-Ks available at Edgar.

Since the raw data is highly seasonable, I take the quarterly growth on a year over year basis of both iPhone unit sales and iPhone search volume on Google Trends.

The first iPhone units were sold on 2Q2007. Its growth rate has been spectacular ever since, including a couple of quarters with +500% growth (given its small initial base). Below I graph together the quarterly year-over-year growth of iPhone's search volume and unit sales. As expected, for most quarters its trends are highly correlated. A notable exception is 2Q2009, when the search volume dipped and iPhone sales skyrocketed.

*Source: Google Trends for iPhone search volume; Apple Inc.'s 10-Qs and 10-Ks for iPhone quarterly sales, available at Edgar.

If I omit the first quarters of the existence of the iPhone, the correlation becomes more evident.

*Source: Google Trends for iPhone search volume; Apple Inc.'s 10-Qs and 10-Ks for iPhone quarterly sales, available at Edgar.

For those more familiar with statistics I provide a simple pairwise correlation test for the period between 3Q2009 and 1Q2014:

Pairwise Correlation 0.7039
p-value 0.0002

*Source: My own calculations using STATA, based on data from Google Trends and Apple Inc's filings at Edgar.

Recall that correlation is a measure of linear dependence between two variables. It gives a value between -1 and +1, where +1 is perfect linear correlation and -1 perfect linear inverse correlation. The p-value indicates that the correlation coefficient is significantly different from zero.

What I'm Not Claiming

As an important disclosure, let me address two of the many limitations of my analysis.

  1. Small amount of data. Even taking into account every quarter with iPhone's unit sales data there are only 28 unique observations, too small a sample to propose more accurate analyses.
  2. Correlation doesn't imply causality. It's one thing for two variables to move together and a completely different thing for one variable to cause another. It is plausible that iPhone's search volume causes its unit sales, given that a consumer would want to check its characteristics before choosing to buy. However, it is equally plausible to assume that the more iPhone sales, the greater its search volume.

My only claim is that search volume is significantly correlated with iPhone unit sales.

A Trading Strategy using Google Trends

If you could get ahead of the market and bet on a increase or decrease in unit sales of AAPL's flagship product before each quarter's earnings release, it would be possible to generate superior trading profits.

I propose the following trading strategy:

  • Trading Signals. Exactly the trading day before the earnings release, if the sequential y/y growth rate in search volume for the word "iPhone" dropped go short. If it increased, go long.
  • Holding Period. Once on a long or short strategy, keep the position and sell exactly on the day before the next earnings call, in which you'll make a new trade based on the new trading signal.
  • Period Selection. My strategy makes sense only for the period in which iPhone sales account for most of AAPL's net revenue. Hence, I select the period from 4Q2011 to 1Q2014. During 4Q2011, iPhone sales crossed the 50% mark as a percentage of total net revenue for the first time in the history of Apple (vs. 34% for 3Q2011). On average, iPhone sales accounted for 52.5% of total revenue for the period.

In the chart below I present the results of this strategy on a quarterly basis.

Earnings Release Date (t=0) Quarter Price t-1 Price Change (%) Search Volume y/y Search Volume Change (%) Trading Signal Return (%) Accumulated Return
25/01/2012 4Q2011 404.08 NA 766 40.0% Long NA 100.00
25/04/2012 1Q2012 538.52 33.3% 624 15.8% Short 33.3% 133.27
25/07/2012 2Q2012 577.58 7.3% 566 10.8% Short -7.3% 123.60
31/10/2012 3Q2012 583.03 0.9% 688 34.6% Long -0.9% 122.44
24/01/2013 4Q2012 498.43 -14.5% 677 -11.6% Short -14.5% 104.67
24/04/2013 1Q2013 396.12 -20.5% 629 0.8% Long 20.5% 126.16
24/07/2013 2Q2013 411.36 3.8% 577 1.9% Long 3.8% 131.01
30/10/2014 3Q2013 510.62 24.1% 735 6.8% Long 24.1% 162.62
28/01/2014 4Q2013 547.22 7.2% 676 -0.1% Short 7.2% 174.28
24/04/2014 1Q2014 524.75 -4.1% 614 -2.4% Short 4.1% 181.44

*Source: Edgar for Earnings Release Dates; Yahoo Finance for Apple's prices on the trading day prior to the Earnings Release; Google Trends for the search volume and my own calculations.

To illustrate this strategy, suppose that you invest $100 on 24/01/2012. You observe the following signal: for 3Q2011, year over year (y/y) growth in search volume was -14.5% (not included in the chart). Since the y/y growth in search volume on 4Q2011 stands at 40.0% (larger than -14.5%), you chose to go long. You sell your position on 24/04/2012, netting a profit of 33.3%. On the same day (right before the next earnings release) you notice that y/y growth in search volume for 1Q2012 is 15.8% (smaller than 40.0%) and you chose to invest your entire position short, and so on.

After a few ups and downs, the holding period return of this strategy stands at 81.4% vs. just 30% of a buy and hold strategy during the same period.

In case you are wondering, search volume growth stood at -2.4% for 1Q2014, so this trading strategy would indicate a short signal for the period between 23/04/2014 to 24/07/2014.

This is only one of many possible trading strategies using Google Trends. Perhaps there are others even more profitable.

Disclaimer: There is no guarantee that Google Trends data is accurate. I proposed this investment strategy as an academic exercise only. I strongly recommend investors to do their own due diligence before making decisions.

Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.