Are Alibaba's numbers fake?
In reference to: Job interview questions: The size and scope of Alibaba
Let me start off by saying that I have an enormous amount of respect for the Sydney-based Mr. Hempton. His work on Valeant and a number of other situations has been absolutely superb in all respects from his level of conviction, the detail of his work as well as his style and approach which I find very entertaining.
Now to the matter at hand.
In his blog post, Mr. Hempton hypothesizes that Alibaba’s numbers might be fake. The key talking points are:
Alibaba’s reported volume appears to be too high
Measured in terms of gross merchandise value (GMV) (this is from an earlier blog post)
Measured in terms of total deliveries
Alipay’s reported transaction velocity also appears to be too high
Measured in terms of peak transactions per minute
After going through this in some detail, I came to the following conclusions:
I actually agree that gross merchandise volume is overstated, meaning that volumes do not necessarily match actual transactions. But “overstated” does not mean “fake” and I will explain below why this ultimately does not matter as it relates to Alibaba’s financials.
Total deliveries is most likely not misstated [1]. I believe the comparison between Alibaba and Amazon/UPS/Fedex/USPS is flawed in a few meaningful ways.
Alipay’s reported peak transaction velocity is also most likely not fake. This is more related to lack of context as to what the data point he cites actually applies to.
(1) Gross merchandise volume
Earlier, Mr. Hempton had written a blog post (Alibaba - yeah right Jack) where he expresses his incredulity around Alibaba'‘s reported gross merchandise volumes:
The GDP of China in 2014 is $10.3 trillion.
The population of China is (estimated) at roughly 1.4 billion people.
This makes for GDP per capita of roughly USD 7,350.
China is also a country which has a very low private consumption rate. The World Bank estimates a domestic consumption rate of about 36 percent of GDP. This is the flip side of China’s huge saving rate.
Calculate it: household consumption per capita is roughly USD 2,650.
So the Alibaba amount of spending per user (USD 1,056) is almost 40 percent of that number.
Yeah right Jack.
His point is that he does not believe how Alibaba can account for even close to 40% of (his back-of-the-envelope calculation on) Chinese household consumption and also expresses surprise that China’s online spend as a proportion of consumption exceeds developed countries like the United States (looking at Amazon as the key comparable).
First, there are a couple reasons why the blunt comparison between Alibaba’s reported GMV and China’s household consumption needs to be fine-tuned:
B2B. Many small businesses use Taobao or even Tmall to purchase goods that do not qualify as retail consumption. I do not know what percentage this is but I suspect that it is not insignificant. Therefore, some portion of the overall GMV should be classified as B2B spending.
International shipments. While the vast majority of Alibaba’s volume is domestic, a small but not insignificant portion (around 10%) goes to international customers (e.g. Hong Kong, Taiwan).
However, even after adjusting for these two factors, I agree with Mr. Hempton and think that the GMV figure still overstates actual goods and services delivered.
This is where understanding one of the key differences between the Alibaba and Amazon business model matters.
Alibaba’s business model is more a pure platform instead of an end-to-end provider like Amazon. Unlike Amazon, whose revenue is primarily based on the sale of merchandise, Alibaba takes fees from merchants for “pay-for-performance marketing services, display marketing and placement services” [2] as well as commissions on transactions paid through Alipay. The Taobao merchants are primarily responsible for handling logistics after the initial transaction is struck on the platform.
It is well-known that many Taobao sellers generate fake transactions (a.k.a. “brushing”). But the underlying reason is less nefarious than one might initially think – it is simply done to increase their rankings in the Taobao marketplace. This is because transaction volume is one of the key factors in getting your products better placement. This also means that GMV in this case is not necessarily an accurate measure of a real underlying transaction volume. Frankly, I don’t think anyone (including Alibaba) knows what this number is with precision. I have seen estimates of fake transactions on Taobao range anywhere from 10 to 30% of GMV.
The only number that Alibaba can track is the transaction as it crosses its Taobao platform and that is the number it reports. I would be incredibly surprised if they are faking that number, although I do think it is a bit “sales-y” to tout the number as if it fully reflected reality. But what do you expect? Jack Ma is one of the most talented salesmen that I’ve ever seen.
And here’s the rub: It doesn’t even matter. Overstating GMV will not impact Alibaba’s financials and profitability. Effectively, the fees that sellers end up paying on these fake transactions are merely another form of marketing spend. This is very analogous to a good chunk of SEO/SEM money spent on trying to optimize placement within Google's search results. If Alibaba were to crack down on this practice, fees on fake transactions would just end up going into another revenue bucket for them (it's not like sellers have much of an alternative here).
(2) Delivery volume
Mr. Hempton expressed further incredulity that Alibaba had the necessary infrastructure available to deliver so many packages:
But at this point I wanted to explore the 278 million deliveries number. So the question is “how many internet users in China”. Most people guessed a number that was reasonable - so I pointed them to official numbers. Roughly 650 million would be a good guess. The official number at the end of 2013 was 618 million - and the growth rate has already slowed.
So there was 278 million deliveries for a total prospective 650 million users. In one day! The question is how much bigger can this get and what drives it to be bigger?
I then pointed them to official numbers that said that Amazon (worldwide) had 244 million users. That was 244 million separate accounts made at least one order in the past twelve months. The Amazon numbers are here.
Alibaba delivered more parcels in a single day than Amazon had users in a whole year.
He goes on to look at the number of employees required to handle the number of deliveries Alibaba claimed both for “Singles Day” as well as over the course of the whole year. Later, he compares the maturity of the logistics infrastructure used in America compared to China, noting that Amazon’s appears to be far more automated.
I looked into the source links and did not find any claim that Alibaba actually delivered all of those 278 million parcels. First, they don’t actually directly handle the delivery, unlike Amazon (for its direct sold and fulfilled merchandise). Second, it is also well-known that orders placed on Singles Day could take up to 7-10 days to actually get delivered as merchants clear through the mountain of orders received. The number was referring to orders placed.
But more broadly, my sense from reading the article is that much of Mr. Hempton’s skepticism in this area is based on comparing how things work in the U.S. vs. China. However, I think one needs to note some major differences between the two in areas that would directly affect logistics:
Population density. China is simply a lot more densely populated than the United States. China’s population density is >4x and this ignores the fact that most Chinese people are crammed along the coastal regions and moreover that these regions are where people who have the disposable income to purchase things online tend to live. The population density of our 4th-largest metropolitan area (Dallas-Fort Worth) is roughly the same as an entire Chinese province (Hebei) which is not even close to being the densest one. San Antonio is more sparse than certain areas in China that are classified as rural.
From a logistics perspective, this means that delivery distances are much shorter and your logistics network can be handled more with “scooters and electric bikes” than “trucks and delivery vans”, especially since scooters and bikes can far more easily navigate China’s traffic-riddled urban streets. While here in downtown Manhattan (i.e. the absolute densest place in the U.S.) it is very common to see Fedex and UPS trucks parked on my street delivering packages, it is far more common in Chinese urban areas to see this:
Lack of incumbents and implications on the development path of China’s retail/logistics infrastructure. One must also keep in mind how little existing infrastructure the e-commerce giants had to compete with. Before Alibaba came along with Taobao, there was very little entrenched retail infrastructure to compete with. There was no Fedex or UPS equivalent – Alibaba had to build it from scratch [3] and only started this in earnest in 2013. Because of different starting conditions, China’s logistics industry is evolving very differently from how the U.S. industry evolved.
First, it is going to be a lot more labor-intensive and spread out. It is no secret that China’s labor costs are still extremely low by developed world standards. But on top of this, many of its industries (especially e-commerce) are growing so quickly that it is more important to go for a “land grab” strategy than invest time and money in automating processes with slick robots and software [4]. That can come later.
Second, quality standards are generally lower and more varied. The industry is extremely fragmented. There are literally thousands of logistics providers, some not much more advanced than a single guy with his scooter in a small village. That is why instead of getting a nicely packaged box with an smile on it, you often end up with this:
China may not have been very efficient using massive amounts of concrete to build a bunch of empty buildings but it sure is pretty efficient at re-using cardboard boxes!
This will improve over time as the industry consolidates and as Chinese consumers expect higher quality.
Logistics remains a very labor-intensive process no matter how much technology you have. Mr. Hempton is skeptical how Alibaba could possibly coordinate to delivery of significantly greater volumes than the combined efforts of the world’s most valuable e-commerce company [5] and the world’s largest logistics companies (emphasis mine):
We should also note the scale of the 8.6 billion packages the network claims to have delivered. UPS delivered 4.6 billion packages per year and had 435 thousand employees, 539 aircraft (including charters), about 100,000 delivery vehicles and almost 2000 operating facilities.
To truly deliver at a larger intensity than Amazon, Alibaba and its outsource network would need more staff or capital (or both) than Amazon and UPS combined.
Despite vast amounts invested in complicated software, shiny delivery fleets and infrastructure, the fact remains that labor accounts for the bulk of the costs in logistics, even in the world’s richest large country. Here is a snapshot from UPS's most recent annual report:
Here you can see how “compensation and benefits” accounts for 58% of all operating expenses. The next-largest line item is “purchased transportation” which accounts for less than a third of compensation cost. The point is that logistics – whether it is picking, packaging or delivering – still requires tons of labor and if anything that is something China has in abundance.
Jack Ma thinks that another 7 million or so service sector jobs can be created over the next decade just from demand for new delivery personnel. That’s certainly one way for the tertiary sector to absorb job losses in the primary and secondary sectors resulting from China’s economic transition!
The bottom line is that I highly doubt that Alibaba is faking the actual delivery (or more accurately, orders placed) numbers. At least based on the figures and logic from the blog article.
(3) Alipay transaction volume
Mr. Hempton is skeptical about the reported size and scope of Alipay based on a comment in Jack Ma’s letter that at its peak, Alipay handled “2.85 million transactions in a peak minute”. He goes on to extrapolate this to a “full day” based on a number of his own assumptions and then compares it to a world-class company like Visa to assess its technical feasibility:
A pretty good guess is that the peak day should be between 4 and 9 hours of the peak minute (in other words the peak day is somewhere between 2.85*60*4 million transactions and 2.85*60*9 million transactions). This seems sensible. I have discussed with several people in the payment industry the ratio of peak minute to peak day volumes and got numbers in this range - but I do not have definitive numbers. [If someone from PayPal or Visa wants to come through with something definitive I would like it.]
7 hours of peak minute (which my informal research says is reasonable) is about 1.2 billion transactions in the day. That is two per internet user in China. That is to say a large number.
1.2 billion transactions in one day per does seem extremely far-fetched. But that is not what the quote said. The problem with this analysis is simply extrapolating off what is very likely to be an incorrect assumption. In the blog’s comments section, “Anonymous” added a very thoughtful note which best illustrates how changing the assumptions drastically alters the conclusion (emphasis mine):
I think your 7 hours of peak transactions significantly overstates the # of transactions. There is huge volume at the very beginning of the day as customers try to take advantage of limited quantity deals or time-expiring deals. (Think midnight crowds outside of Best Buy on Black Friday) The peak # of transactions almost certainly occurs in the first few minutes.
On Singles Day, Alibaba and/or its employees post updates of the cumulative GMV for the day. Alibaba reportedly generated 57.1b RMB of GMV on Singles Day. The #s I’ve seen:
- 1b RMB GMB in the first 2 minutes
- 10b RMB GMV in the first 38 minutes
- 20b RMB GMV by around 7am
- 30b RMB by around 11am
- 40b RMB by around 4pm
- 50b RMB around 9pm
- remaining 7.1b RMB of GMV was generated from 9pm to midnight
This implies GMV per minute peaks in the first two minutes and then drops off significantly (by more than 90%). On my calculations, GMV per minute was:
- first two minutes: 500m RMB per minute
- 12:02 to 12:38: 250m RMB / minute
- 12:38 to 7am: 26m RMB / min
- 7am to 11am: 42m RMB / min
- 11am to 4pm: 33m RMB / min
- 4pm to 9pm: 33m RMB / min
- 9pm to midnight: 39m RMB / min
A few further calculations:
- if we assume the “peak” 2.85m Alipay transactions per minute applies to the first 38 minutes and then scale the number of transactions/minute in line with Alibaba’s GMV per minute, then on my calculations, Alipay processed 618m transactions
- if I assume the “peak” 2.85m Alipay transactions per minute only applies to the first 2 minutes and then scale the # of transactions/minute in line with Alibaba’s GMV per minute, then on my calculations, Alipay processed 325m transactions
Unfortunately the quote is unclear on what “peak of the event” applies to, but the range of Alipay transactions could be 325-618m if “peak” refers to 1) the first 2 minutes; or 2) the first 38 minutes
So without knowing what time period the quote was referring to, it is hard to assume based on the quote that Alibaba is just making this number up.
Notes
[1] Update (3/16/2016): Now whether all of those deliveries are “real” transactions once again is a separate issue. The Wall Street Journal recently reported on fake transactions and corresponding fake deliveries where deliveries take place but are of empty boxes. Once again, similar to my points above about GMV, this is merely another form of marketing spend.
[2] Source: Alibaba IPO Prospectus (F-1) pages 19-20
[3] More accurately, Alibaba invested in and worked with a number of logistical partners to build out its logistics network. Unlike Amazon, Alibaba does not directly own its logistics infrastructure although it has certainly played a key role in developing it.
[4] In fact, Amazon tried to apply its approach in China a decade ago and failed. Amazon is incredible at building amazing software and technology and optimizing everything to the last significant digit. But these skills do not translate into one’s ability to hire thousands of young people to go door to door convincing merchants (many whom were just not tech savvy especially in the early years) to sell their wares online. Or to sit in bars and cafes and show and convince other young people how to buy stuff online. These were some strategies Alibaba used to win and they have nothing to do with technology.
[5] As measured by market cap as of February 2016.
This was originally published on Quora in February 2016.