Big Data is worshipped in virtually every organization today, but its journey to the mainstream is a perilous one that needs to be carefully managed. All of us, from the casual hobbyist to the data aficionado, have to know where the pitfalls are and employ greater caution on its usage, particularly when factoring or predicting humans. Warning: Contrarian arguments ahead.

Ah, the promised land of Big Data.

The holy grail, we are being told by every media outlet, that empowers us to make successful decisions in today’s organizations. Big Data helps us deliver more targeted marketing to customers, identifies inefficiencies to streamline costs, and provides real-time insight into areas of crisis and opportunity. CEOs pay close attention — fail to master Big Data, and your organization will probably become obsolete come the next decade.

A gumptious assertion, and one I’m about to appraise, but before that, what is Big Data anyway?

Strictly speaking, Big Data refers exclusively to the data, but technology and analysis help bring meaning to the data. Together, the trio form Big Data solutions, which is a market worth billions today.

Let’s break it down further.

First, the data itself. Universally, there are two main forms of data. Structured data, which follows a pre-defined model, like customer transactions, and unstructured data, which has no regular format, like images and news documents. Big Data, as most people imply from its name, refers to structured and unstructured data sets so voluminous it would take years to process via traditional methods. But volume isn’t the only characteristic that makes data “big”. Classification of data as “Big Data” is also contingent on its velocity (data is streamed in high speeds), variety (data must be extracted from many different sources and manifest in many different forms), as well as veracity (data is frequently incomplete, uncertain and unreliable).

Second, the technology. When we talk about Big Data today, we aren’t just referring to the data, but also to the technologies and techniques (things like NoSQL, Apache Hive, Spark, MapReduce etc) which make it possible to even extract, aggregate, process and understand the data. The technology is important because having data without technology would be akin to having gallons of fuel without an engine to burn it.

Third, the analysis. This is typically governed by the human operator. Other than deciding what data to analyze, the operator also needs to decide what the success measure, or in other words, what the goal of the analysis, is. That in turn determines what kind of data analysis is required. This ranges from constructing models for predicting future outcomes, to measuring success rates of current initiatives. Sort of like regular analytics, but on a much greater and faster scale.

In theory, this sounds like pure gold, and from the viewpoint of a data evangelist, I am excited about the wide-ranging applications of Big Data. Big Data can help barbecue chains sell more ribs, but it can just as well help save lives by predicting natural disasters. However, we ought to be cognizant of its pitfalls and consciously avoid them where possible. In practice, if managed carelessly, Big Data can lead to serious ramifications, several of which may not be immediately obvious to the best of us.

1. Big Data perpetuates and normalizes the surveillance society.

Big Data or Big Brother? The loudest critique of Big Data perhaps lies in the nature of the data being collected. As Big Data technologies become more powerful, the more extensive the data we can collect, and the higher the value we can extract out of the data. In hoarding the data necessary for prime motives like detecting terrorist threats and calculating credit scores, governments and conglomerates are justifying the installation of all kinds of sensors and trackers to monitor our every move, both online and offline. But is it okay for companies to use the same sorts of surveillance to do targeted marketing and boost sales? Are we collecting more data than what we actually need?

Today, privacy is a bygone right — not only do most of us not know that we are being tracked, we are beginning to accept surveillance as the new normal. Caveat — While Facebook, Google and the like are quick to point out that the data they collect is faceless (doesn’t identify the individual), there is still a possibility that systems can still be hacked, or exploited in secret by malfeasant leaders, in violation of civil rights.

2. Big Data is only as objective as its creators.

Big Data does an excellent job of objectively modelling the factors which are fed into its system. But that objectivity is also a pernicious weakness, serving only to reinforce prejudice if the choice of data is biased in the first place. Garbage in, garbage out. For instance, if we were using Big Data sourced from smartphones and other mobile gadgets to approximate the patterns of pedestrian movement in a city, the data might be inherently biased because it wouldn’t be inclusive of people who do not carry around smartphones. Our models might perceive less movement in neighbourhoods which have more lower income or older residents who are less likely to be smartphone carriers. As a result, Big Data hasn’t negated bias, but simply amplified the bias of our data choices.

3. Big Data black boxes may (unintentionally) flirt with moral boundaries.

Subjective data choices become a bigger problem when they toe ethical and moral boundaries. Worse still, these ethical violations can happen unintentionally even when the creators set out to use Big Data with good intentions. In her book Weapons of Math Destruction, Cathy O’Neil, a former quant and data scientist, points out in an example that Big Data is used by prisons to ascertain recidivism risk of prison leavers (the risk of prison leavers returning to prison) and for police to conduct predictive policing. On the surface, attributes such as neighborhoods are simply another facially neutral statistic, but when analyzed under the Big Data microscope, they become highly accurate proxies for racial discrimination. Because Big Data is assumed (incorrectly) by many to be fair, companies are now more easily able to, under the guise of Big Data “black-boxes”, justify making the kinds of unfair decisions that would be heavily protested if made by a human.

4. Big Data evangelizes overall outcomes and ignores individual failures.

“Logic clearly dictates that the needs of the many outweigh the needs of the few”, said Spock famously in The Wrath of Khan. The goal of Big Data solutions is always to maximize success as a whole, even at the expense of a few individual failures. That’s alright when we are talking about Netflix’s or Amazon’s prediction system occasionally recommending me a couple of movies/merchandise which are ill suited to my tastes. But it’s much less forgivable if those small margins of error had more serious, individualized consequences. For instance, if Big Data was being used by, say, courts to determine convict sentencing and wrongly judged someone at trial. Or if a company conducting a retrenchment exercise misidentified under-performing staff thereby causing them their jobs. Or if a resume screener which leverages Big Data ignores a candidate with an incredible personal story because his resume didn’t contain the right blend of keywords. In all of those situations, a human assessor might have helped drive a very different outcome.

5. Big Data, in redefining new systems of scoring or evaluation, could gradually become destructive to the sanctity of society’s systems.

Furthermore, Big Data driven evaluations can encourage people to learn from patterns in the system’s behaviour and adapt to the systems in unnatural, destructive ways. Take the example of using Big Data to drive teacher evaluations- assuming the Education Department sets in motion a drive to replace human, in-classroom observers and peer interviews with evaluations purely based on Big Data. It may start to utilize easily documented data, like the students’ test scores, their level of engagement with co-curriculars, or even their parents’ feedback. What is more difficult to objectively measure, however, is how teachers help students to build character and socio-emotional competencies. When teachers begin to realize that test scores play a more poignant role in their own grading, they may become pressured to focus on those statistics instead of the holistic goal of education.

Another example — when we start to analyze customers via their online behaviour, demographics, and spending patterns, instead of relying on actual, physical interactions and surveys, we begin to redefine traditional notions of customer satisfaction, creating an “empathy gap” between ourselves and our customers. Big Data is extremely savvy at modelling what the customer wants, but it is incredibly difficult to model what the customer feels. And that may not be a positive change.

6. Big Data can engender vicious feedback loops.

In addition to redefining how we are judged on a systemic level, Big Data has also begun to influence our behaviour. Big Data algorithms are being employed by social networks to constrain us into clusters they believe we are interested in. As we interact more with our friends from these clusters, we reinforce the belief that we are engaged by content from the clusters, slowly but surely excluding us from the rest of the world. For millenials who obtain their world view largely from social media, this can be really petrifying. Big Data, boosted by Internet trends, can allow companies like Facebook or Twitter to wield petrifying amounts of political and socio-economic influence.

7. Big Data, if pursued vacuously, can cause a drain on valuable resources.

Last but not least, organizations are increasingly being pressurized by their management to jump on the Big Data bandwagon without truly understanding what their objective functions are. This is further perpetuated by business managers who lack the technical acumen to drive Big Data operations, or even understanding what Big Data can or cannot do. In many situations, this means millions of dollars spent on infrastructure, software, technologists and consultants to do Big Data exploration, which in many situations generate little value.

All things said, there is nothing wrong about Big Data or Big Data technologies themselves and we should not be starting an anarchist protest against the institutions who use it extensively. The issues I mention above stem from misguided use from us human operators, and several of these issues are really issues with data analysis in general, “big” or not, just magnified on a much larger scale with Big Data. But, as Big Data proliferates into the mainstream, all of us will eventually become, on different scales, operators of Big Data. When that happens, we must individually have the awareness to complement Big Data with our intuition to make smarter, ethical decisions. At an organisational or societal level, we must ensure that we have the governance and legal frameworks to identify and purge Big Data misuse. This is the only way for Big Data to truly become that progressive force we’ve been promised.

The Internet in America is awash with chatter on the FCC’s plan to “dismantle net neutrality”, and I need to add to that chaos no further. But it’s interesting to ask a few questions of the whole debate, and drive the context closer to home.

Is Net Neutrality the biggest problem in the room?

Net Neutrality gets a bunch of press because it’s a trendy topic often linked with strong themes like innovation & fair competition for small businesses. But that’s just the tip of the iceberg when you consider the real elephant in the room Ajit Pai has continued to endorse — the oligarchy of cable providers.

In America, broadband is dominated by a few providers (Comcast, Charter, TWC etc), which in my own experience, over-charge, under-deliver, provide terrible customer service, and have poor coverage of rural regions. Internet speeds are decent if you’re in the correct zip code, but still dismal compared to the cream of the developed world (think South Korea). In one of my favorite publications (BackChannel), Susan Crawford, a Harvard Law professor, writes a riveting article on how a smaller cable company, which runs a local monopoly, is upsetting Big Cable because it’s raising its data prices too fast — and is therefore increasing the risk of regulation & oversight for the whole industry. In other words, forget the political high ground of the internet, we have deeper problems to address. Before we debate about Open Internet, the real fight should be in keeping the Internet provider oligopolies in check so high-speed Internet access can remain affordable & available.

What about Singapore?

In Singapore, Net Neutrality is a bit of an outside term — in fact, I don’t think most people have heard of it at all. Most consumer complaints today center around network interruption (is the wifi down again?), network speeds (where’s that 1 gbps connection I was promised?), or unfair pricing (more a pay-TV thing though). There is a flavour of consumer indignation against the Big 3, though the landscape is accommodating new players. By and large, pricing is competitive/fair, and ISPs do a relatively good job of what they are paid by consumers to do (for now). I dare say that many would find it fascinating that an innocuous-looking zero-rated plan, like Singtel’s Spotify Plan, is violating the purist principle of Net Neutrality. What’s generally accepted without furore here is the subject of widespread protestation in America.

But just because it isn’t a problem yet, doesn’t mean it will always stay that way. What is Singapore’s stance on net neutrality?

Back in 2010, when Open Internet was established, the IMDA (then IDA) issued a consultation paper on Net Neutrality, seeking input from industry players. All of the Big 3 focus on attacking IDA’s proposal for them to publish “average” speeds instead of “maximum” speeds for their plans, citing a list of reasons which make average speeds “misleading”, “difficult to determine”, and “counter-productive”. IDA relents, but goes on to say that they will publish a “Guide to Residential Broadband in Singapore” to help guide consumers (found here by the way). I’ve only just seen this guide for the first time, so I’m doubtful it’s of any use to consumers. And if all the ISPs abide to regulation & display those average numbers, the playing field will be levelled again — so how is this “counter-productive”?

There’s more. Here’s an excerpt from Singtel’s response.

“SingTel is supportive of the spirit of net neutrality…However, SingTel submits that there are no issues arising from differentiated pricing, QoS and terms and conditions for usage and service providers should be allowed to continue to do so as long as they are transparent about the terms and conditions with their end-users.”

Basically, what the telco(s) were saying here, in a subtle way, was that it’s okay for them to have partnerships with OTT providers to provide zero-rated or differentiated plans and throttle legitimate content, where necessary, in order to meet a level of service quality for all consumers. While IDA largely took the middle ground — agreeing with allowing ISPs to manage bandwidth, they did take a hardline stance on not allowing ISPs to charge consumers or providers extra for OTT services. (Singtel’s CEO Chua Sock Koong did briefly run into a PR fiasco in 2015 when she publicly commented that network providers should be able to charge content providers for their OTT services.)

So consumers get their “subsidised” Netflix & Spotify subscriptions, and providers don’t have to pay extra to deliver their services — win-win for all right? Not so fast. Right now, ISPs may not need to throttle bandwidth as often because there is “sufficient bandwidth for all”. But as consumers get more online-savvy, and activities such as video streaming become more prevalent, the challenge would be to be for the telcos to keep up the same level of service quality — particularly prescient in the congested mobile networks — without raising prices significantly. Network congestion is a real issue today especially on 3G/4G mobile networks — telcos have already scrapped their 12GB plans once in a major shakeup— and they now have to make sure zero rated plans don’t cause more strain on the network and force them to penalise consumers.

As zero-rated plans become more mainstream, there is also more pressure put on other OTT providers to get on board. Studies have already shown that having zero rated plans gives content providers a serious advantage in terms of user-ship. While not as dangerous as discriminatory throttling or banning of certain web traffic, it is difficult to say if zero rating plans put smaller players at a significant disadvantage in their quest for market share. And if ISPs begin to double dip by providing zero rated plans for as many OTT providers, then OTT providers are almost paying a tax just to be competitive. As more OTT providers emerge, the regulators may want to consider safeguarding their interests as well, on top of those of the ISPs and the consumers.

Meanwhile, I still kind of like not being charged for mobile data on my Spotify app.

“Hello, would you like to make a donation?”

It’s lunchtime in the Central Business District. A meek, bespectacled girl stands near the cross junction between two office skyscrapers. In her hands, a shiny, circular tin with several thumb-sized stickers lining the edge. She urges the tin valiantly towards the ravenous office workers, who stream past her in all directions. The vehement roar of lunch hour traffic does her soft voice no favours. Most passers-by skirt past in a hurry. Some look away, using their smartphones to avert eye contact. Others acknowledge the girl’s existence, but express apologetically that they have no spare change on them today.

Finally, reacting to the girl’s exasperated pleas, a middle-aged gentleman stops in his tracks. As he reaches for his pocket to pull out some change, the girl relaxes, breaking into smile — her persistence finally paying dividends. The man probes politely where his donation would go to and how it would be used. Startled, the girl takes a quick glance towards her friend, who is also canvassing for donations a few steps away. Her friend doesn’t notice. The girl quickly glances at the paper glued to the tin can and stammers the name of the charity, but she cannot provide any further details. Either satisfied by the answer or feeling sympathetic, the man drops in a single golden coin and continues on his quest for sustenance.

It’s a situation many working adults empathise with. Several of us in Singapore remember this flavour of donation drives (non-profits mobilising swathes of volunteers with coin tins to collect donations from strangers on the street) as one of those mandatory activities we did as part of a co-curricular activity or the school’s Community Involvement Programme. During a Flag Day, as it was typically called, you had two main group of participants. The first group viewed it as a matter of pride — to do well meant you had to report back with the heaviest coin-laden tin. They canvassed furiously and purposefully. The second group didn’t care at all and spent the day chatting with their friends, eventually putting in some of their own loose change to avoid the wrath that came with submitting an empty tin. Whichever group you fell in, it was rarely explained to you in detail what the organisation cared or how the donations would benefit it. To be fair, few of us ever asked anyway.

Despite ‘coin tin’ Flag Days being so grossly inefficient on a dollar raised per volunteer hour basis, their impact on the philanthropy scene in Singapore, in terms of amounts raised, is still pretty substantial even today. A typical campaign — and there are countless through the year — easily mobilises a couple thousand volunteers a day, and over a week could potentially raise a six digit sum. Proponents also contend that there are intangible aspects of this method of fundraising, such as increasing awareness among both volunteers and members of the public, or having an educative impact. Yet, there are clearly more effective ways for charities to raise awareness, such as targeted online campaigns, or even more efficient means of running donation drives, such as this cashless example, but that’s a topic for another day.

Instead, let’s draw our focus to you, the individual donor. Recall the earlier anecdote of the middle-aged man? That’s most of us when we either feel like it’s time to take the moral high ground for that charitable act for the week. Or when we get ambushed and guilt-tripped to give up our lunch money by persuasive teenage volunteers. In many situations, we barely know the name of the charity we are donating to, less what our donations would be used for.

Does it matter? Giving is unequivocally good, right?

Yes, the intention to give is mostly indisputably good, but can we give badly? Donating your spare $5 to feed a low-income family for a meal? Easy choice. But, what if I told you that same $5 allows for the procurement and distribution of a mosquito net, which could protect two children in Malawi from malaria for 3–4 years, and potentially save their lives? The decision becomes a lot less straightforward.

Central to this rational form of giving is a movement gaining recent popularity, aptly called Effective Altruism. In a nutshell, Effective Altruism promotes a philosophy of giving where your dollar will be stretched the furthest. It takes the “emotion” out of giving, and ultimately encourages people to give to organisations which are run most efficiently and have the largest impact, say having the lowest “cost per life saved”. (Here’s a Ted talk by Peter Singer, a moral philosopher at Princeton and staunch advocate of Effective Altruism.)

An important tenet of this school of thought is that all lives are equal — the life of an underprivileged child in Africa must be as equally worth saving as that of an underprivileged child in your home country. If there is a charity out there in a world which can make a greater impact, then we must dedicate our gift to that charity.

According to GiveWell, an organization which collects and analyses data on charities, the best charities to donate to currently include the Against Malaria Foundation, Deworm the World Initiative, and Give Directly (a nonprofit which delivers cash to the extreme poor in Kenya and Uganda). Unsurprisingly, they are all based in Africa.

Of course, with all decisions in life, it’s never that straightforward in reality. Even if one was able to to donate without prejudice and sentiment towards a specific cause — which is difficult considering it’s the emotion which drives giving in the first place — there are still numerous stumbling blocks.

• People who run non-profits aren’t social scientists. Therefore, they are often poor at collecting or reporting statistics about themselves.
• Statistics are difficult to collect and may be inaccurate. It could be difficult to measure a metric like cost per life saved.
• Even if we were able to measure accurately, not every organization can be measured by a common statistic. For instance, if you were donating to a school for underprivileged children, how could you measure the impact of education? How do you benchmark a nonprofit which helps the blind against a foundation which educates orphans?
• If everyone abides to such a philosophy, only the high profile charities with complete and transparent statistics will receive aid. If everybody is donating to the Against Malaria Foundation, who will donate to the poor in our own local communities?
• A judicious perspective towards giving may not be for everyone. Many donors are guided by emotion rather than rational thought. Will encouraging them to be more rational instead backfire and cause them to give less?

While it would be impractical for everyone to fully adopt Effective Altruism principles, we should embrace some of its ideas and adopt the middle ground. Regardless of motivations, it is always a good idea to do some due diligence on organisations before making a weighted decision to donate or volunteer. As a product guy, I love driving decisions through metrics. Here are some metrics to consider before donating:

• Track Record: Are the beneficiaries legitimate? Are there well-documented examples of how the organization has benefitted its beneficiaries?
• Cost Per Life Saved (Or more generally, impact per dollar donated)
• Expense Ratios: Are the charities fiscally responsible? Are they overpaying their executives or spending too much money on irresponsible admin costs or marketing campaigns? In other words, how lean or efficient is the charity being run?

Other than these metrics, we should also compare a particular charity against its comparable peers, for instance, other charities which benefit the same cause. Often, organisations which tackle the cause, rather than the issue itself, will be able to serve a greater impact. If donors and volunteers alike work with the most efficient charities, the less efficient charities will eventually fold and we will be able to see a consolidation of individual charities into bigger but more effective operational units.

Every dollar counts. Donate with your heart, but think before you donate.

Everybody knows that they key to staying healthy is a balanced combination of nutrition and exercise. Yet, every year, people around the world spend billions on “expertly curated dieting plans” and “dieting supplements”. In America alone, complications arising from these dieting supplements send an estimated 23,000 people to emergency rooms every year. In short, it’s a complete, unadulterated mess.

But is there a better, healthier, and cheaper solution to this madness? In the hottest technology of today, virtual reality (or augmented reality), we may find a clever way to address the problem of overeating — by tricking the mind into eating less without diminishing levels of satisfaction.

Much of the talk surrounding VR has been entertainment-focused, from video games to sport, and movies to travel experiences. But none of these applications are as ubiquitous as eating, an activity we spend roughly a tenth of our day doing.

The psychology of food is nothing new. For decades, scientists have been exploring the connection between our sensory experiences and the way we eat. Common findings include the effects of the Delbeouf Illusion Bias, which states that smaller plates and high color contrast between plate and food cause us to put less food on our plates. In other words, we tend to pile on less alfredo pasta on our plates when we use a black plate instead of a white one.

Takuji Narumi, assistant professor at the Cyber Interface Lab at Tokyo University, contends that humans, like animals in the wild, are genetically programmed to consume as much food as possible to maximize their chances of survival. While this has worked well for us in ancient civilisations, it’s clearly a problem with the high availability of food (especially junk food) in today’s world. Our body’s hardware, that is our mouths and digestive organs, have not adapted quickly enough to the rapid shift in paradigm to highly available and highly processed foods. Narumi’s idea of food hacking is to use virtual reality to hack our body’s software (i.e. our brain) and manipulate our appetites, through tapping on well known food psychology principles like changing the size and color of the food we eat.

By attaching other add-ons like smell manipulators, we can also affect how our brain perceives the actual taste of the food. For instance, the smell of chocolate may lead us to believe that our food tastes sweeter, allowing us to reduce the sugar content in our food without affecting the pleasures we derive from eating.

Beyond dieting, virtual reality headsets could also have endless, wide-ranging applications in dining. Imagine altering the shade of a glass of wine and introducing olfactory components to make it taste like your favorite 1980 bottle of Bordeaux, or changing your tableware and surroundings so you can dine in a beachfront restaurant in the Caribbean, all from the comfort of your own home.

Watch out, world.

Good deal for Goldman, right? Except for one minor detail — that startup was Alibaba, and ten years later it was returning the favor to Goldman in a very unexpected way, employing the investment bank to underwrite its IPO on the NYSE. Which, by the way, valued Alibaba at $168 billion, more than double that of what Goldman’s worth today.

Success stories like Alibaba’s and mobile upstart Xiaomi’s ($45B and still private) have romanced investment in Asia’s startups to reach unprecedented levels over the last five years. Riding the Alibaba dragon, e-commerce startups have been particularly successful at attracting prodigious funding, even at a nascent stage. With Asia having 45.7% of the world’s Internet population and yet only 34.7% penetration rate (the rest of world is at 51.8%), it’s not hard to see why. Nearly every country has either been colonized by Groupon or Amazon, or has invented its own localized clones. But with so many e-commerce startups emerging with billion-dollar valuations with no profits, and talk of yet another Internet bubble forming, can we realistically expect any of them to replicate Alibaba’s success? To make sense of these heady valuations, let’s take a deeper look at four comparable Asian e-commerce startups which have caught the public eye a lot lately. I’ve chosen these subjects because they share the following traits and may make for interesting comparison:

• Founded in 2010 or later
• Achieved Unicorn status (> $1B valuation)
• Originated from a country ranked in Asia’s top 5 by number of Internet users

1. Coupang (South Korea), founded 2010 :

At a $5B valuation, Coupang is a behemoth in startup terms. Beginning operations as a Groupon clone, Coupang has pivoted and is now a full-fledged ecommerce platform maintaining its own inventory and logistics chain. South Korea’s answer to Amazon, Coupang’s pages are refreshing, colorful, and modern — filled with the latest fashion items and advertisements featuring KPop celebrities. Its founder, Kim Bom, is a Harvard Business School dropout who idolises Masayoshi-son of Softbank fame. And one can imagine it’s no accident that of the $1.4B it’s raised so far, Softbank alone contributed a cool $1B in one of the biggest single startup investment of 2015. While South Korea has a (relatively) small market of roughly 45m Internet users, it has a tech-savvy population and one of the highest Internet speeds in the world, conditions which have allowed e-commerce to flourish rapidly.

2. Snapdeal (India), 2010:

Gurgaon-based Snapdeal has doesn’t do its own logistics, and pretty much still operates a Groupon-like business model. Like Coupang, its two co-founders went to prestigious universities, in UPenn and IIT respectively. Very recently, it received a $500 million injection from Alibaba and Softbank, valuing it at an impressive $5 billion in a country with roughly 250m Internet users.

3. Meituan (China), 2010:

Meituan is the equivalent of Snapdeal in China — by far the biggest market with over 600 million Internet users. A group discount website, it specializes in localized consumer services and last year alone sold $7.4B of deals and vouchers. It’s also received a monster round in 2015 ($700M in January), and at $7 billion is currently the highest valued of the four.

4. Lazada (Singapore/Southeast Asia), 2012:

For good measure, I’m throwing in Lazada into the mix to give some perspective from one of the hottest regions in Asia right now — Southeast Asia. The youngest of the quartet, Lazada is an outlier from the above three in many ways — It’s only valued at $1.25 billion, it didn’t start out as a Groupon clone, and it wasn’t founded by a native (it’s of Rocket Internet* origin). But having hit the $1B mark in 2.75 years there’s no questioning Lazada’s ambition. At this level of valuation, it’s slightly ahead of the other three at the same stage — though with Rocket’s backing it’s almost expected to immediately outperform. Operating primarily as a B2C platform (but also some C2C) and spearheaded by a McKinsey alumnus, Lazada has been allowed to expand with ruthless, German-like efficiency in courting the over 190 million Internet users in Southeast Asia.

*For the unacquainted, Rocket Internet is a German Internet company which runs a global empire of businesses primarily founded by replicating the strategies of other successful companies in new markets.

Some Numbers:

Numbers don’t lie. Well, assuming we can trust the authenticity of the numbers in the first place.

Let’s begin by looking at the cumulative investment in the four companies over the last five years:

Snapdeal has raised the most with $1.56 billion, followed closely by Coupang with $1.4B and Meituan with $1.07B. Most of the investments have come in 2014 and 2015 (Snapdeal was particularly prolific in 2014 and Coupang in a similar sense in 2015 with that $1B Softbank infusion). Meanwhile, Lazada has yet to publicly disclose any funding rounds in 2015, and lags behind the pack at $680m. Yet, Lazada is only three years old, and is well ahead of the other three companies at their respective 3-year marks.

Cumulative investment, however, only tells us how much cash the companies have to burn, but indicates to us little about their value. While it’s difficult to ascertain the valuations at each funding stage, we can at least confirm that at the moment, Meituan is valued the most. For reference, Meituan, Coupang & Snapdeal are all — for now at least — valued more than the very company they set about emulating —yes, the father of daily deals, Groupon ($2.8B).

1. Meituan: $7B
2. Coupang/Snapdeal: $5B
3. Lazada: $1.25B

How do these astronomical valuations compare with how the companies are doing? Here’s where the picture gets slightly murky. Having no access to properly documented income statement disclosures, I’ve had to rely on whatever’s been said in press releases or the companies’ own reports. Profitability is shrouded in secrecy but it’s probably safe to assume none of these companies are profitable. Still, I’ve gathered sufficient information about their sales numbers to get a good-looking graph.

Here’s where Meituan’s performance — in line with its valuation — outshines the rest. Meituan claimed $7.5B GMV in 2014 and is forecasting a $16B GMV run-rate for 2015. In comparison, Snapdeal, also a daily-deals pure play, only broke $1B in 2014 but is growing at a similar rate as Meituan, forecasting $10B by end 2015. While we’d expect the GMVs of an end-to-end e-commerce model to take off slower than a third-party, daily deals model — Lazada & Coupang both illustrate that, Coupang is impressively predicting its GMV will break $3.9B in 2015. Lazada grew its GMV from 300% through 2013 to 2014, and should break the vaunted $1B mark this year. If they manage to hit these numbers (remember this is a prediction), here’s how their valuation/GMV multiples will look like:

• Meituan: 0.44
• Snapdeal: 0.50
• Lazada: 1.25
• Coupang: 1.26

How do the industry veterans compare? Here are the Price/Sales multiples for Amazon, Groupon, and Alibaba as of 18 Sep 2015:

• Groupon: 0.910
• Amazon: 2.635
• Alibaba: 12.85

Revisiting the earlier valuations, they don’t seem so crazy after all. Meituan’s and Snapdeal’s multiples are half that of Groupon’s, while Coupang’s and Lazada’s are half that of Amazon’s and four times lower than Alibaba’s. The mechanics of this direct comparison are flawed (especially when we consider factors like Groupon charging merchants 40-50% of their listed deal price while Meituan only charges something like 5%), but at least we get a clearer picture of why they are valued that way.

The immediate phase ahead : Domestic supremacy

At a high level at least, we’ve justifiably established that Asian investors haven’t gone mad and there is still some sensibility in the market today. But what lies ahead for these power-hungry unicorns?

In theory, the next phase is straightforward: Establish oneself as the undisputed domestic leader by channelling as much money as possible towards customer acquisition. Get customers, and deals will flow right? With so much money on the table, it seems impossible to fail. E-commerce is continuously growing and Asians love to shop, so what’s there not to like about this strategy? Competition, that’s what. Asia is synonymous with competitiveness and it takes a lot of investment — and strategy — to become peerless on the domestic scene. It’s a large factor that explains why Amazon and Groupon, despite their scale, have been largely unable to penetrate into the region. The other, of course, is that Asian markets are less homogeneous than Western markets.

China: A thousand Chinese Groupons

There’s no better place than China to begin a conversation about competition. It’s no secret that the Chinese are ridiculously good at “copycat innovation”. As a result thousands of e-commerce sites exist and there is no model more copied in China than Groupon’s. The stories of the rise and fall of these Chinese Groupons contain so much drama it would take days to tell them all. But here’s the gist: When Groupon tried to launch in China five years ago, it found — to its horror — that someone in China had beat them to the Groupon domain and had already started a business called Groupon.cn. As a result original Groupon had to launch under a different name (Gaopeng) when it expanded to China. Groupon then had to acquire Groupon.cn to boost its business, but that’s another story. In 2011, possibly the peak of the Groupon craze where there were at least 1500 clones, the market leader was Lashou, and things were looking up for the fledgling startup. A few years, a failed IPO and some market stabilization later, Lashou has fallen way off the pace, dozens of other startups have traded market share leadership countless times, and Gaopeng still hasn’t even been able to break into the top ten.

Meanwhile, a pattern of what might become the most interesting three-headed fight in China over the next couple of years has emerged with increasing clarity. A direct showdown among the big three in China — Baidu, Alibaba, and Tencent is happening right now, not just in e-commerce but in all sectors of the Internet economy. B.A.T, as they are otherwise known, made their name through disparate channels — search engines, e-commerce, and social media/messaging respectively, but have since expanded so astronomically that they that it almost takes the backing of one of them to enjoy prolonged success in the Chinese Internet market.

Meituan’s arsenal? A direct and big-money backing from the Alibaba Group (which also owns Juhuasuan, another successful Daily Deals platform that’s even enjoyed expansion to Hong Kong and Taiwan). With 51% of the B2C market through TMall and 95% of C2C through Taobao, Alibaba’s support will be crucial to Meituan’s prolonging its longevity. Alibaba led the $50M Series B round for Meituan and also helped finance the $300M Series C round. Meituan has grown rapidly to become the industry leader with 60% market share, which is promising, especially with the powerful asset of Alibaba’s user reach and e-Commerce engine.

But don’t expect the competition to back down without a fight. Meituan’s closest competitor now is $4B valued and Sequoia-backed DianPing, in which Tencent bought a 20% stake in 2014. The advantages of having Tencent in your boardroom? An immediate incorporation into WeChat, China’s top messaging app with 600 million registered users, and an important component of Tencent’s online-to-offline strategy. Not to be outdone, Baidu announced in June this year it was pumping in $3.2B into Nuomi, a group buying site Baidu has fully acquired. The impact? Imagine if Google bought Groupon today and integrated Groupon into Google Search and Google Maps. The current results are favouring Meituan, and I’m hard-pressed to bet against it, especially if it gets Alibaba’s full backing, but as history has taught us, it’s not how one starts, but how one finishes.

India: East versus West heats up

India doesn’t have the same level of economic xenophobia that China has, so American tech companies have traditionally enjoyed a great degree of success in India, which is also a recruitment mecca for precocious, talented engineers. Unlike in China, Google, Whatsapp and Facebook are all market leaders in India. But in 2007 a couple of ex-Amazon employees set out to prove that an Indian-born tech eCommerce startup could emulate the successes of its Western counterparts. They founded eCommerce portal Flipkart to become India’s Amazon, and with overwhelming success — Flipkart now reportedly commands 44% of the market share by sales. Realizing India’s potential, Amazon invented Junglee in 2012 to compete with Flipkart(and also to cleverly skirt the regulations around a foreign company running a full fledged service in India) and commands around 12–14% of the market now. (There’s an interesting thread here outlining the differences between Flipkart and Amazon.) There’s still a reverence for Silicon Valley ideas and companies even amongst traditional Indian tech conglomerates, but there’s an increasing feeling that homegrown tech startups now have enough machinery to compete with the Valley’s finest.

Where does Snapdeal stand in all of this? Snapdeal’s currently at 22% market share, but in the purest sense of competition, Flipkart and Amazon aren’t really group-buying sites so Snapdeal doesn’t directly compete against them. Groupon has a presence in India, but has achieved more mixed success than Snapdeal, having exited once and been rebranded twice. It has recently given up control to Sequoia, been rebranded as Nearbuy, and is planning to nip at the heels of Snapdeal again. However, Snapdeal seems to have held up strong. And I think Flipkart and Snapdeal can both co-exist to some degree given their different focus. Snapdeal’s biggest competition to profitability is probably itself —we earlier mentioned that Asian versions of Groupon have much lower margins per deal sold and this has probably led to Snapdeal’s rapid acquisition of customers. Will Snapdeal be able to generate enough volume to make its low margins profitable or will it be able to pivot to a model which can allow it to charge higher margins?

Korea: A titan emerges

In Korea, the picture seems to be a lot clearer. I’ve narrowed the key competition down to three. There’s the two with enormous financial muscle — Softbank-backed Coupang and KKR-backed (formerly Groupon-backed) Ticket Monster [TMON]. But then there’s also the odd case of WeMakePrice, which is still 100% owned by its founder Huh Min (although there are plans to change that to a small degree). There’s also a significant player in eBay Korea, which recently acquired GMarket for $1.2B, but C2C is really eBay’s strong suit while the other three are more similar in adopting Amazon’s model — selling both direct goods & third party brands.

For a relatively smaller market, their achievements have been phenomenal. Mobile commerce in Korea is dominant (e.g. Coupang and Ticket Monster’s mobile sales account for 75% and 70% of revenue respectively). Coupang’s mobile app has been downloaded 25 million times from the App store. To put things in perspective, Korea’s population is 50 million. That’s one download for every two people in Korea. While the number of customers may not have that much room to grow further, the transaction volumes do. eCommerce in Korea, like in several other advanced Asian economies, still take up a small percentage of overall commerce, and investors are clearly banking on mobile commerce sites to disrupt commerce further.

TMON was first in 2010 to introduce social commerce in Korea, but was quickly followed in the same year by the other two. Since then the trio have traded positions in a closely balanced market, and in 2014 all three were still regarded as even players. Fast forward a year, and Softbank’s faith in Coupang seems to have transformed a close dogfight into a David vs Goliath contest. Coupang’s value is now more of five times TMON and [x times] WeMakePrice (I say x because WeMakePrice is still 100% owned by its founder and difficult to value for sure). With this level of financial muscle and Softbank’s impressive track record, my money’s on Coupang to make the transition from market player to market champion.

Southeast Asia: No clear winner

There’s a solid piece on Techcrunch which kind of details the eCommerce scene in Southeast Asia, so I’d recommend that read for a detailed analysis. In short, the authors argue that despite Lazada’s best efforts to establish itself as the market leader (it’s present in six Southeast Asian countries and has a reported 20% market share), the competition is still very much open — much more so than China and India at least. They highlight the $100M investment partly financed by Softbank in Tokopedia as a potential threat to Lazada’s foothold. Which, given the kind of genius investments Softbank has made over the last decade, should definitely give us reason to believe that Tokopedia can become a force to be reckoned with.

In summary, achieving domestic supremacy is hard — even with a ton of cash in the bank. There’s only one thing for sure — competition drives prices lower, and consumers are going to love every bit of that.

What lies beyond — Long Term Survival

What does a tech startup do with hundreds of millions of dollars? Aside from the obvious need for to acquire customers through publicity campaigns and discount programs, a substantial sum is often reserved for acquiring similar companies which already have a sizeable user base. This allows the startup to grow at a far more burgeoning pace than if they tried to grow organically, and at the same time clear the market of any pesky competitors.

Assuming our eCommerce startup gets past the domestic hurdle, somehow overthrowing or acquiring its competitors, establishing a superior position in the domestic market and getting a ton of loyal customers, what can we expect from it in order to continue growing in the longer term?

Let’s consider some possible approaches —

1. Expanding internationally into Asia/rest of world with the same business model

All four startups certainly have the pedigree to go global once they have taken care of their own domestic challenges. (To some extent, Lazada’s already done that since they are now present in six different countries, although succeeding in Southeast Asia is still very different from succeeding across Asia.) However the main difficulty is the adaptability of their businesses outside of the domestic niche in ultra-heterogeneous Asia. Amazon and Groupon have failed in China and haven’t really taken off in the rest of Asia. But Asian companies will do better in Asia, you say. Not true. Take the example of Rakuten, the shining star of Japanese eCommerce, who failed spectacularly in China and, having enjoyed limited success in other parts of Asia, is now looking at focusing on its U.S. expansion plans instead.

But as they say, if you can’t beat them, join them. And as investment bankers and venture capitalists love to say — if you can’t beat them, acquire them. Thus, what’s likely to happen, if we ever see Meituan trying to expand to India, or Snapdeal trying to enter Korea, et cetera, is that it will be achieved via a merger or acquisition, rather than a head-to-head battle. Consider this fascinating investor matrix, which outlines the investors who have invested in more than one of the four startups:

Surprise, surprise. There is an incredible synergy among the main investors of Coupang, Meituan, and Snapdeal. I’ll call this the S-A-S connection, short for the Softbank-Alibaba-Sequoia alliance. Softbank is the main investor in Coupang and Snapdeal, Alibaba is highly involved in Snapdeal and Meituan, and Sequoia is considered a main investor in Meituan and a co-investor in Coupang. To reinforce this alliance, Softbank (majority) and Sequoia (minority) both own stakes in Alibaba, so they are really vicariously invested in anything that Alibaba decides to lay its hands on. It’s a clear and powerful message to the entire continent that SAS is after domination in every major market in Asia, at least where eCommerce is concerned. But why is Lazada conspicuously absent from the SAS portfolio — it’s links to the other three merely through Temasek Holding’s minority stake in Snapdeal? Well, that’s because — as hinted earlier — Softbank and Sequoia are anti-Lazada. They’ve both invested in Indonesian upstart Tokopedia and are betting that Tokopedia can upstage Lazada to become the kingpin in Southeast Asia. When I began writing this piece I picked the four startups because of their similarities in value and age. But now looking at things this way, it becomes clear that the biggest factor Meituan, Coupang, and Snapdeal have in common is who’s filling their seats in the boardroom.

2. Evolving business model to stave off competition

While the Groupon model is quick to implement, enjoys excellent initial traction and good profit margins, it’s always been a concern — considering Groupon’s recent stagnation — whether the model is at all sustainable.

The biggest critiques of the Groupon model is that Groupon

• Is too easy to replicate by competitors, driving margins lower
• Attracts too many one-time buyers looking for specific discounts
• Has its reputation tarnished by merchants having poorer quality services/products
• Leads customers to believe the merchants have to be inferior in some way to have to aggressively discount their prices

Daily deals sites like Groupon often perform poorly compared to full fledged eCommerce sites like Amazon and Rakuten because of they attract customers through the cheapness of the service/product, rather than product quality and customer experience. For Meituan and Snapdeal, this could mean that their impressive growth numbers now give a poor indication of continued growth over the next few years. Coupang may have realized the opportunity when they pivoted to build end-to-end eCommerce infrastructure, but trying to do the same would be a gargantuan challenge in China and India, as it would involve upstaging Alibaba and Flipkart at their game. Besides, one might argue that it’s easier for Alibaba and Flipkart to enter the daily deals market than it is for Meituan and Snapdeal to try to do end-to-end eCommerce.

An alternative way for eCommerce companies to succeed could be to specialize. Companies who have differentiated themselves well from the rest of their peers include vintage marketplace Etsy, designer-good-centric Fab and members-only Gilt Groupe. These niches are great, but unfortunately being too specific doesn’t present the best option for our four unicorns because of the small market size and lack of scalability.

What about diversification? Let’s not forget that both Amazon & Alibaba, while still largely e-commerce at their core, have already diversified into an array of other products and services, most notably on-demand media and cloud computing infrastructure. Over its two-decade lifespan, Amazon has acquired anything from shoe e-tailer Zappos to live streaming site Twitch.tv. Alibaba, apart from its investments in other eCommerce outfits, has also shelled out billions into things like online education (Tutor Group), video streaming (Youku), and ride-sharing (Lyft). End-to-end eCommerce often means the need for things like great data processing capabilities and an extensive merchant/customer network, which can be naturally extended to other online-to-offline commerce platforms like education or taxi services.

Our subjects are beginning to do exactly that. They are acquiring either companies who have a useful piece of technology, or companies which help to plug the gaps in the overall value chain. For instance, Meituan’s invested $35M just this month in Zhai.me, a Shanghai-based campus-oriented delivery app. Zhai.me will help Meituan overcome its last-mile problem by providing food deliverymen access to campus dorms. Snapdeal’s acquired a dozen startups itself, the most notable being the $400M acquisition of Indian mobile recharge platform Freecharge, which is popular among the 18–25 year old segment of the population. Coupang’s even begun acquiring abroad, turning to Silicon Valley startup CalmSea, which specializes in large-scaled database systems and distribution optimizations, irrevocably useful for Coupang’s massive logistics operation.

Will we ever see the next Alibaba in any of these four startups? Probably not, but only time will tell.