Estimating 2Q 2012 US Smartphone Sales

Last week, Strategy Analytics published its estimates for 2Q12 US smartphone shipments; however, they can't been seen as reliable due to a number of problems. Whenever I read research or an article, or hear somebody quote market share figures from the large industry research houses, such as IDC, Gartner, Strategy Analytics, or Canalys, I cringe. Too often, I find wide disparities between these industry estimates and the numbers I know to be true- figures reported by manufacturers, carriers, resellers, etc. in press releases, SEC filings, and on conference calls. Another problem is that these research firms give little public explanation behind their methodology. A better framework is to first calculate sell-through, or sales to end-users which the top 4 carriers provide smartphone unit figures. These aren't 3rd-party estimates; they are actual sales figures from the horse's mouth. Since these numbers can be attained with a high degree of confidence, we can then hash out shipment estimates based on analyzing market and seasonal trends, as well as incorporating any guidance available from vendors and carriers/resellers. 

Strategy Analytics 2Q12 US Smartphone Shipment Estimates:

source: Strategy Analytics

The first problem with SA's Q2 US shipment estimates is that its figures for Apple aren't shipments, they are sell-through. The 7.9M unit shipment estimate for Apple is the sum of activations reported by AT&T, Verizon, and Sprint. To be consistent with measuring shipments, as we assume SA is doing for the other OS platforms, sales to iPhone carriers and resellers (i.e. Best Buy, WalMart, Radio Shack, etc) need to be measured. 

Second, the activation figures used only include 3 carriers, however iPhone is available through 16 carriers in the US. Even though these carriers are very small and many weren't online during the entire quarter resulting in relatively low activations, there would be considerable iPhone shipments into these channels due to the need to build a stock of initial inventory.    

SA is assuming that iPhone shipments equals sell-through. Apple reported that channel inventory declined by 300K globally, hence 26.3M units sold-through versus the 26M units shipped. Apple specifically stated that China shipments were impacted from a reduction in channel inventory which was high at the start of the June quarter due to the iPhone 4S launch at China Unicom and the addition of China Telecom in the March quarter. Therefore, sell-through in China was significantly higher than sell-in (shipments) which resulted in the reduction of channel inventory. Apple stated the China channel inventory impact was more than $1B of iPhone revenue, which would equate to more than 1.5M units. Therefore, channel inventory had to increase by 1.2M units outside of China to equal the net 300K decrease in global channel inventory. Given that Apple added 12 carriers in the US, coupled with only a 12% sequential decline in activations at the top 3 carriers, we can attribute much of the 1.2M channel inventory build to the US.    

Third, we know the 7.9M figure for 2Q12 and the 5.9M figure from 2Q11 aren't the correct numbers to use. Tim Cook stated on the conference call that iPhone sales were up 47% year over year in the US. Using the activations figures as SA did, the Y/Y increase is only 34%. Therefore, we know that shipments don't equal the activation, or sell-through numbers that SA incorporates. 

So for SA's estimates to be reliable, or even taken seriously, they must be consistent in what they are measuring. If they are measuring shipments, then SA should use shipment figures for Apple, not sell-through. Obviously, shipment numbers are a lot harder to ascertain than sell-through figures since US carriers report smartphone sales or provide metrics to calculate it. 

Using the erroneous activation number, SA calculates that Apple captured 33% of shipments. SA also says Android shipments were 13.4M units and RIM shipped 1.6M units. Where or how did SA come up with these figures? We know the RIMM figure doesn't make sense. Taking the revenue numbers RIMM reported for the US for the May quarter, and adjusting for June month being much weaker than February, ASP's are highest in US, and significant portion of service revenue comes from high ARPU customers (such as enterprises) in the US, Blackberry shipments were most likely around 1M units, not 1.6M units SA estimates. So what about the Android number? That's a tough one to reliably estimate. However, we can make some meaningful assumptions for estimated shipments from sell-through estimates which are much easier to calculate given that carriers provide smartphone sales. Using the underlying trend of sales to end-users, we can then make assumptions of whether or not carriers & resellers would be increasing/decreasing or maintaining inventory levels.

Financial Alchemist 2Q12 US Smartphone Sell-Through Estimates:

AT&T & Verizon reported smartphone sales of 5.1M & 5.9M, respectively. Sprint stated that 81% of handset sales were smartphones, and given the upgrade rate and gross additions, Sprint sold around 3.3M smartphones. T-Mobile hasn't reported quarter results, but sold 2.5M in Q1, so estimating 2.4M units would likely be appropriate. Then comes the wild card: the other carriers.  They consist of very small regional post-paid players and national prepaid carriers which probably account for about for 30M subscribers. It's a hard number to quantify. It's also hard to quantify smartphone sales for those carriers as well. However, they only make up 10-15% of the market, and the percentage of smartphone sales is lower due to the high cost of hardware which isn't subsidized much for prepaid plans. A conservative estimate would be that the "other carriers" roughly resembles T-Mobile. 

In total, I estimate that 19.2M smartphone units were sold to end-users in 2Q12. SA estimates that 23.8M units were shipped, or sold into the channel. So is it reasonable that shipments exceeded sell-through by 4.6M? If sell-through were expected to increase by 4.6M units in Q3 it would, but it's very unlikely that it will by that much, if at all, for a couple of reasons. 1) Carriers are tightening their upgrade policies, charging higher fees and lengthening the time period until one is eligible for a fully discounted upgrade. 2) Smartphone penetration has reached over 50% at the top carriers  (62% at AT&T, 71% at Sprint), thus the pool of potential first-time smartphone buyers is shrinking. 3) Seasonality- typically demand softens in the summer, and gradually accelerates into the back-to-school period and holidays. Q3 generally sees the widest divergence between sell-in and sell-through as carriers build inventories. In sum, unit shipments likely exceeded sell-through slightly (20M units), or at best, only modestly (21M). The vast majority of any excess sell-in would consist of Android units, however. 

At the top 3 carriers, Apple captured 55% of smartphone sales to end-users. For the entire US Market, I estimate Apple's market share was 42%. In terms of share of sell-in, iPhone would likely only be lower by 2-3%. It's highly unlikely that it would be as low as 33% Strategy Analytics estimates. Estimating the sell-through for the other platforms is more difficult, since only Verizon provided Android figures. But taking data reported by RIMM and Nokia, as well as other company sources, unit sales and market share can be estimated with a modest degree of confidence.  

source: FA estimates, company reports

My method for estimating US smartphone sales mirrors the method used by Ben Evans from Ender's Analysis. In being conservative, I model a higher volume of unit sales to the other carriers (non-Big Four). 

In short, industry research firms such as the ones mentioned in the opening paragraph often don't produce reasonably accurate estimates. This instance is just one example of many where I have encountered considerable factual evidence to the contrary.  The best approach is to always take the figures reported by the industry players as we know that those are accurate. From there, we can make more reliable estimates with regards to the unknowns.