Category Archives: Strategy

Distribution v Innovation

The battle between every startup and incumbent comes down to whether the startup gets distribution before the incumbent gets innovation.

The TiVo Problem

In 1999, ReplayTV and TiVo invented the Digital Video Recorder (DVR). It was an incredible innovation — allowing you to “pause” live television.

But TiVo had no value without “content” to pause. That content, by and large, was distributed via cable and satellite TV networks.

And because TiVo was separate from your cable box, using it was far from simple. If you wanted your TiVo to “know” what shows were on (and consequently record them), you’d have to have it connect (via modem/phone line — remember, TVs were not placed near phone jacks) to a TiVo server to download them.

Clearly TiVo had an enormous channel opportunity. What if Comcast, Adelphia, Cox, and other large cable companies simply distributed TiVo to their customers? Wouldn’t that be a home run for TiVo and the cable companies — a new service that would delight customers with a massive new revenue stream to boot? And, integrated with the cable box, the TiVo product itself would get better, too.

But — and this is what I call the TiVo problem — that doesn’t normally work out well, and if you look at your cable box today (with a generic DVR function, I would bet, built-in), you know how this story ends.

If you’re TiVo trying to cut a deal with a Comcast, one of the below normally happens:

  1. You partner with Comcast, but Comcast dominates the economics of the deal, in some cases restricting your cooperation with its competitors. (rare to partner)
  2. You sell your whole company to Comcast, but you’re not selling a company, you’re selling an awesome product…and somebody else might have an awesomer product (or a worse one that is deemed better by the technology team at Comcast). Moreover, if you already have commercial deals with Comcast, Adelphia, Cox, et al…, Comcast won’t value your ex-Comcast revenue (because it will disappear upon acquisition by Comcast!), dramatically reducing your independent valuation. (rare to sell)
  3. You get screwed by Comcast. Comcast builds a crappy version of your product, but because they have the distribution, they can and will beat you. (common)

TiVo did not fail, but it became a patent troll of sorts. It has a market cap of less than $1B today, despite having collected more than $1.6B in patent settlement funds from the “Comcasts” of the world.

The Winning Strategy: Go Boring

Given that the common outcome to the “TiVo problem” is getting clobbered by Comcast, how do you deal with this situation?

The answer is often to “go boring and be patient. This was a big mistake I made at TrialPay, which put relevant offers around the payment flow. We built a great product/service/business on top of payments, but it wasn’t core — merchants didn’t start off looking for or needing our product. They started off looking for what I thought was boring, cheap, commoditized payment processing. Going back to the analogy: Consumers want/need Comcast more than they want/need TiVo. Or at the very least, the chronology starts off with Comcast.

In 2006, I thought “Why build a ‘boring’ commodity payment business like Stripe or Square (that did not yet exist), when we could build the lucrative feature missing from all the commoditized payment processors?” We had insanely better unit economics than they did.

But these payment processors had the customer relationships, and they had the starting product that the customer wanted. Eventually we sold TrialPay to Visa, and I think a lot of value will be created for Visa from that deal, but not nearly as much for TrialPay shareholders had we owned the channel.

This is the flaw with looking at Square and Stripe and calling them commodity players. They have the distribution. They have the engineering talent. They can build their own TiVo. It doesn’t mean they will, but their success hinges on their own product and engineering prowess, not on an improbable deal with an oligopoly or utility.

Being Judged on the Present

There are two ways somebody can interpret this video.

-“I’m much better than that kid at golf! [says a 33 year old]. I have a 12 handicap and can outdrive that joker by 200 yards!”

-“Wow, that’s remarkable for somebody of that age…if he continues like that, he could someday win 14 Majors.”

Both assessments are logically correct. But as a young company selling into enterprises, you will often get the first reaction.

In my experience, a lot of larger corporations (and people who work at them) can generally only see the present — the present capabilities, the present revenue (or trajectory), the present limitations. In startups, you need to see the future. Not as a fortuneteller would (impossible) but to judge teams and ideas on their future potential / adjacencies.

There is a natural lesson here for an entrepreneur — which is to beware showing “leanness” of product when interacting with a large company. Saying “we can/will add that later” unfortunately lacks credibility, because large companies are often incapable of building anything quickly, and hence their employees tend to doubt this statement. “Blockers” in the large organization will try to scrap any deal with a “deficient” startup.

The right way to build a typical startup is “lean.” Overbuilding before product-market fit can be catastrophic; building sophisticated management and operational processes before you need them is normally a vast misallocation of resources and actually prevents learning of what the market wants.

But present your lean startup to a large company and you’ll hear “where’s the beef?” When selling a product to a large company, or even selling your OWN company to a large company, you’ll be thoroughly evaluated on the present — which sometimes is good in, say, M&A when you are on an unsustainably high growth rate. The hard part of a company is generally making the whole thing work; it’s not the sophistication of a set of algorithms, but having the whole product perform at scale with an organization that can support it.

I saw this firsthand at TrialPay, which was, at its simplest level, an advertising technology company. An early potential deal with a large company did not materialize because said company was displeased with our optimization systems, even though we could add a better optimization algorithm in a few days (GitHub shows 329 collaborative filtering projects). But we were judged on the present, and if we had the opportunity to do it over, I would have actually invested those few days to look “fatter” — even if it had no impact on our business.

The Danger and Opportunity of the Intermediate Metric

Are social media companies overvalued? The question is not just a matter of revenue multiples (low or high), but rather whether that revenue is actually generating new sales for advertisers. Google convinced the world to believe in the click, Facebook has done the same with the Like, Twitter with the follower, and Pinterest is planning on unveiling the same with the Pin. Creating these “intermediate” metrics between impression and ultimate purchase is a great move to boost revenue, but must stand up to scrutiny as software eats the marketing funnel. For startups seeking to build a valuable advertising business, creating an intermediate metric is crucial, but so is ensuring that that metric holds up to scrutiny.

Let’s rewind a bit, though. Without commerce, without transactions, there would be no advertising. The point of an advertisement is to generate sales. Full stop. Brand building, goodwill, mindshare, buzz, and a lot of other niceties might come about, but even those are meant to eventually lift sales. Without a transaction at the end of the line, advertising has no raison d’être.

The challenge, though, is that it’s often difficult to draw a straight line between “person sees an advertisement” and “person buys a product.” Impression and transaction are the two endpoints of the advertising-commerce lifecycle.

And, the chronological delta between impression and purchase can be wide. A 15-year-old might be bombarded with BMW advertisements for 10 years before she finally pulls the trigger on a fancy, brand-driven car purchase. Deciding to buy Advil vs. Tylenol might take years of external inputs and supermarket trips.

Enter the intermediate metric, which is anything that falls on the continuum from impression to purchase: clicks (the Internet’s first intermediate metric), likes, bookmarks, views, shares, app downloads, recall, followers, retweets, mentions, pins, etc. Intermediate metrics help publishers (e.g., Google, Facebook, Twitter, Yelp, Pinterest, etc.) attempt to show their impact when sales are not readily measurable — either because of chronological disconnect or because the transaction data is not readily accessible. Or, cynically, and in some instances, because there are no downstream sales — making the intermediate metric the best way to obfuscate while purportedly showing performance.

intermediate metric

Intermediate metrics help advertisers show internal and external stakeholders that they’re doing a great job. It’s hard for Clorox’s marketing team to be given an instruction of “sell 20 percent more bleach this quarter and you get a big quarterly bonus!” A national “must wear white to participate” tomato fight might increase sales of Bleach without Clorox lifting a finger. So many advertisers will compensate and reward their teams for the achievement of intermediate metrics.

  • “Your goal for the quarter is to get 10 million Facebook Likes, and to get a 15 percent increase engagement on Twitter.” (This must increase sales, right?)
  • “Twenty percent of your bonus this quarter will be based on getting 100,000 mobile app downloads.” (Mobile is hot and people are using mobile phones everywhere, so it must drive revenue!)

The greatest intermediate metrics allow for the broadest attribution tracking possible (accounting for marginal intent generation), while being somewhat unique to the medium. At scale, Quora might charge for a promoted corporate answer; Gmail might charge for a bolded email; Waze might charge for a “route added.” These would all be intermediate metrics, knowing that none of these actions yield an immediate purchase but hopefully contribute to one. Without an intermediate metric, there would be a publisher-advertiser marketplace failure, since immediate “transaction” tracking would undercount efficacy and cause metrics-driven advertisers to abandon the platform.

The greatest intermediate metrics allow for the broadest attribution tracking possible while being somewhat unique to the medium.

The smartest thing that Google did was charge for the click, not the sale, because it isn’t Google’s fault if your site converts poorly (or if a sale/action is not relevant, as it is for, say, auto research).

The smartest thing that Facebook did was define the like not just as an intermediate metric, but as a quantum of self-worth. Watching Samsung hit 20 million Likes must have made HTC mighty jealous and want to respond accordingly. When I asked a large restaurant chain where they spend most of their money online, the president said “Facebook. We get a lot of likes, and that must be better than not a lot of Likes.” A click — Google’s classic intermediate metric — isn’t too relevant for a restaurant that doesn’t deliver or allow online transactions. Facebook has a potentially broader audience, yet less transactional intent — so ultimately those likes will need to turn into revenue.

As Twitter goes public, it probably needs a stronger intermediate metric that can resonate with the long tail of advertisers. It might not make sense for regular people to “follow” an advertiser like Oreo in the same way they might follow their favorite moviestar, thus making followers a poor metric; in fact, The Bronx Zoo’s Cobra (an actual snake) has more followers than Oreo. The famous Oreo Superbowl tweet was retweeted only 16,000 times. The most retweeted brand advertisement on Twitter (from Nokia) has yet to top 50,000 retweets. Yet perhaps Oreo was seen by millions of people on Twitter, yielding a spike in supermarket sales, and thus followers and retweets — the intermediate metrics with which pundits seem to be measuring Twitter, are the wrong intermediate metrics.

The danger of intermediate metrics is that they feel quantitative — these are numbers, people! — but they might actually be meaningless. Ironically, both parties, advertisers and publishers, often have a vested interest in separating them from sales — for the short term — lest the music stop. Separation allows for “quantifiable metrics” when sales are just too hard to perfectly measure, so advertisers can keep spending and publishers can keep charging.

If a company’s revenue is based on selling questionable intermediate metrics, be cautious — no matter how quickly that revenue is growing. Sometimes metrics are purely about internal vanity and do not last. As an example, “number of app downloads” feels like a key performance indicator, whereas for many companies (Supercuts has an app?!), “apps” make little sense as a paradigm. Depending on how this intermediate metric (app downloads) stands up against actual incremental sales, the whole app download market could suffer. The same goes for many other intermediate metrics. When advertisers start thinking of the intermediate metric as the final action (the R in ROI = achievement of intermediate metric), the market is inflated.

For any company — whether buying traffic or selling it — intermediate metrics are often a crucial strategy in building a broad revenue model and in having a metrics-driven approach to customer acquisition and retention. But it is unwise to divorce the intermediate metric from the final, and crucial, metric of the transaction — to ignore it, or to exaggerate it, is penny wise and pound foolish. Plenty of startups and established industries (television advertising!) will be obliterated when data finally lights the path from impression to transaction and, in some cases, reveals it to be seldom traveled.

Payment Data Is More Valuable Than Payment Fees

We are in the midst of a great revolution in the payments space: anyone with a phone can now accept credit cards; online-to-offline commerce is allowing online payment for offline purchase and significant friction is being removed from the consumer purchase experience thanks to mobile. All of this innovation (read: competition), combined with government intervention, means that payment fees are falling, threatening revenue streams for incumbents and startups alike in the payments space. But a broader opportunity exists: using the data of payments to build a more valuable, more defensible business model, one not dependent on fees. The result will revolutionize offline commerce and online advertising.

Today: It’s All About Fees, and They’re Heading Towards Zero

Payment companies make money by charging fees to “process” a payment from buyer to seller. Square charges 2.75% (or $275/month for volume up to $250K/year). PayPal Here charges 2.7%, as does Intuit GoPayment. Groupon and Amazon are both supposedly working on their own dongles, and prices will continue to fall, especially as these new devices create “one-sided” networks without significant defensibility outside of switching cost and inertia. “Pay with Square” is a potential game changer, as the millions of Square user accounts can ONLY be used with Square. But basic “acceptance of credit cards” is becoming a commodity where prices will keep going down.

Competition between payment companies is only one leg of inevitable downward pricing pressure. Government intervention is the other. Not too long ago, the Australian government decided that payment fees were too high, so now most Australian merchants pay less than .5% for credit card swipes, a fraction of the cost here in the US. The European Union is likely to enact similar legislation. The Durbin Amendment of Dodd-Frank and the $6B+ (pending) Brooklyn Settlement are US-based government and civil attacks on the business of payment fees. Many of these fee-cutting regulations help intermediaries like PayPal and Square short term, by reducing their cost (owed to the Visa/MasterCard infrastructure), but eventually it limits what they can charge, too.

Wherever fees end up, most merchants will still dislike paying them. They are a “cost of doing business” that every merchant has an incentive to bring down. Payment companies generally aren’t delivering new customers; they’re taxing the flow of existing ones. Google effectively charges 20-30% to deliver a customer (if you back out the cost-per-click to percentage of realized sale) to an ecommerce merchant, yet merchants are competing to hand Google more money because each dollar “in” produces more than a dollar “out.” Payment companies charge a fraction of Google, but are often despised (witness the lawsuits and legislation) or treated with promiscuous disrespect.

It comes down to something rather simple: Connecting the bank accounts of buyers and sellers will never be as valuable nor defensible as connecting buyers and sellers. Google delivers customers at the top of the funnel, and payment companies serve the prosaic, but necessary, task of shuffling funds at the end.

Tomorrow: Payment Data Will Revolutionize Commerce & Advertising

As society goes increasingly cashless, payment companies will have a larger business, and a more valuable one, in closing the loop for offline transactions and helping deliver customers. The data they possess is without equal; did somebody buy something? How much did he spend? What did she buy? Paper money cannot be tracked in this manner. In order for Online-to-Offline commerce to take flight, every merchant needs an ability to track online/mobile action to offline purchase, and PayPal Here, Square, GoPayment and others could provide just this for a whole new class of small merchants.

Imagine that Wendy’s, or even a local handyman, wants to advertise on the Internet. What’s the point? What does a click, or an impression, really mean? It’s clear what it means online, since every click can be measured to “action” (e.g., purchase) for an ecommerce company. Who can tell Wendy’s, or the local handyman, if that online advertisement worked?

In an increasingly cashless society, the answer is pretty clear: the payment infrastructure. Tracking that purchase back to the originating source (Google? Yelp? Patch? etc) is known as “closing the loop” and will revolutionize offline commerce and advertising alike.

The million-plus merchants walking around with Square, PayPal Here, and GoPayment dongles want more customers, and these dongles provide a means to “close the loop” and let those merchants acquire more customers, remarket to those customers, understand those customers, and do everything that ecommerce companies have taken for granted for over a decade. Legacy POS systems were poorly integrated and insufficiently verticalized, often requiring a merchant to have separate relationships with every player in the payment chain (hardware vendor, merchant bank, CRM system, etc); moreover, they were priced out of reach of the sole proprietor.

Beyond closing the loop, payment companies can utilize data from existing transactions to generate more transactions. Companies who maintain a direct relationship with the consumer — such as American Express, PayPal, Square, Discover, etc — are in the perfect position to serve as an Amazon recommendation system for “everything.” You bought a tennis racket at Sports Authority? How about tennis lessons with Saul the tennis pro, at a discount thanks to your purchase of a tennis racket, only redeemable with the same payment instrument? You weren’t searching for Saul, and you wouldn’t want an unsolicited email from Saul, but seeing an advertisement for Saul shortly after buying a tennis racket (say, on your purchase receipt) would likely produce a response. It’s a way topreeempt search for a large class of “secondary” purchases (e.g., charcoal after buying a grill; tennis balls after buying a tennis racket, etc), in a “pull” based way.

None of this is to say that the fees charged today are wholly unreasonable and unconscionable; they’re just not long-term defensible as more parties offer the same conduits to existing credit card infrastructure. I have $40 cash and five credit cards in my wallet right now, so any merchant wanting to charge $100 for some widget can either get 97.25% of $100 (if using Square), or $0. That’s an easy decision and shows why things like Square and PayPal Here are hugely beneficial to merchants and consumers alike. But longer term, as those fees continue to compress to the benefit of merchants, the larger business will be in applying the data of payments to the benefit of merchants, consumers, and payment providers alike.

Service as a SKU

(Originally guest-written for TechCrunch)

The biggest ecommerce opportunity today involves taking offline services and offering them for sale online (O2O commerce). The first generation of O2O commerce was driven by discounting,push-based engagements, and artificial scarcity. The still-unfulfilled opportunity in O2O today is tantamount to tacking barcodes onto un-warehousable services by standardizing and normalizing the units being sold, something I call “Service as a SKU.” Just as Amazon figured out how to build the best warehouses and technology in the world for delivering boxes, somebody will do this for “unboxed” services, with customers driven not by discounts or scarcity, but rather by the Internet’s hallmarks of customer experience and convenience. And unlike how “ship stuff in a box” ecommerce seems to be gravitating towards a few winners, Service as a SKU is still a wide open playing field.

The idea is to turn every service, or unit of commerce, into what retailers typically call a SKU (Stock Keeping Unit). Imagine the following as “items” you can buy, and have “delivered,” with a simple click or tap:

“1 Unit of Plumber-Fixes-Your-Leaking-Toilet”
“1 Unit of Dentist Fixes Your Crown”
“1 Unit of 12-Inch Hole-in-Roof-Is-Fixed”
“1 Unit of Piano Tuner Tunes Your Piano”
“1 Unit of Set Up a Home WiFi Network”

Groupon and LivingSocial, early leaders in O2O commerce, started a wave I wrote about a few years ago, but have historically focused on discounting and creating demand by artificial time or quantity scarcity. There are two main problems here:
-Adverse selection: Groupon et al tend to attract customers looking for deals. This is not what Amazon does, and not how most consumers shop for necessities (e.g., fix my toilet!).
Push v Pull: Groupon et al tend to rely on “push” (e.g., email) to drive a tremendous amount of sales. Unlike Google, eBay, Yelp, or Amazon, people don’t tend to go to Groupon “unprompted.”

To successfully create a SKU for every service, you need to normalize both the service provider (price/quality) and the service being rendered. It’s more like buying produce than buying something mass-produced in a factory. Or, perhaps more accurately, it’s more like booking a hotel reservation, where the rooms are anything but identical, there exist varying degrees of quality, but there are also quite a few commonalities.

The company that pulls this off will need to have the following:

-A seamless scheduling system, deployed at various service providers, to allow real-time inventory management. OpenTable does this for restaurants, and hence can provide a marketplace for “tables” at You can’t sell boxes without knowing how many items are in your warehouse; you can’t SKU-ify a Service without knowing how many hours are available.

-A trusted ratings system to allow for normalization of services and parsing of consumer feedback. How do I compare a $100 “fix my toilet” plumber to a $175 “fix my toilet” plumber? Ideally this will work like hotels: every service provider has a “star rating” and an associated cost. Hotel rooms are reasonably similar; consumers can choose between a 5 star hotel or a 2 star hotel, and even different star levels have significant variance. Yelp and Angie’s List have tremendous assets in their community-based feedback, although payment companies like PayPal and Square have perhaps an even better potential asset on their hands (chargeback rates are a good proxy for merchant quality, every completed transaction can solicit quality feedback and not just from aggrieved/fanatical customers, etc).

-A no-discounts, no-push site. OpenTable gets people looking for restaurants, and needs neither emails nor discounts to make that happen. Yelp, Google, eBay, Angie’s List, and Amazon are all contenders as they all have consumers “coming back” unprompted. If the product and site are sufficiently convenient, this often happens organically; having a well-designed and convenient search, shopping, payments, and redemption experience avoids the need for push marketing.

-Relationships with offline service providers. Despite the flash nature of Groupon and LivingSocial, their merchant relationships are significant. Yelp has virtually every business profiled but perhaps not every business engaged in an economic relationship.

It’s important to note that Service-as-a-SKU is not lead generation for offline services, nor is it just a glorified scheduling platform. “Leadgen” has been around since the beginning of the internet, but there is no standardization or normalization, not to mention the convenience of “one-click” purchase. There are leadgen services for housing relocation, laser eye surgery, insurance, etc, but none let you actually make a purchase online. The hard part is in “normalizing” to create a single “service item” that can be scheduled, paid for, and “delivered” with a mouse click or smartphone tap. As an example, Uber has done this for black cars, and EXEC is fixing hourly prices and limiting SKUs to low-wage labor services.

At 8:01 AM on June 26, 1974, a shopper named Clyde Dawson bought the first item — a 10-pack of Juicy Fruit gum — to ever be scanned with a UPC (universal product code). Today, barcodes are a part of every mass-market product bought and sold throughout the world. You won’t see plumbers, dentists, limo drivers, or gardeners walking around with UPCs on their backs, but we are poised for another shopping revolution of equal magnitude.

Push v Pull

What makes email, Facebook, and Google so valuable? Answer: Visiting them is largely unprompted, notwithstanding the synapses that fire in your brain that make you check your email, your Facebook feed, or decide to research something on Google. In other words, people pull content themselves, rather than having that content be pushed — or foisted — upon them.

The best way of looking at consumer web applications is as a complex stack of “pulls” and “pushes.” Lest these terms be confused with an earlier generation of push: a “pull” is an unsolicited action by a consumer, whereas a “push” is a solicitation by a seller/producer.  The consumer ultimately “pulls” from a mobile phone or computer. Everything else is “pushed” to the consumer, through ads, e-mails or other marketing efforts from companies eager to get business and traffic.

The greatest trick that Facebook ever “pulled” was transforming itself from a push platform (dependent on email to woo users back) into a de facto pull platform.  Facebook touts that 50%+ of its users log-in every day, and my guess is that the vast majority do so with no prompting. Push is still valuable but simply complements the massive pull that Facebook has developed.

Why is Pull so essential for a web company? The intersecting forces of human psychology and economics.

First, psychology: consider how most people hate being “sold” to. “Being sold to” is a form of push. Consumers get hundreds of unsolicited offers and emails pushed to them every week. They learn to tune these solicitations out, especially if they are not in a buying mindset. Relevance is a function of offer-consumer fit paramaterized by time.

Second, economics: A pull platform doesn’t need to spend any money to reach or acquire customers; a push platform does. Facebook’s marketing spend per user has to be the lowest of any company known to man. Granted, Facebook is intrinsically viral and laden with network effects, but the unprompted pull phenomenon has been crucial to Facebook’s dominance.

The value of pull is not just for consumer companies. Any Business-to-Business company knows the value of “demand generation”: catalyzing a “pull” by customers. The quickest and cheapest sales cycles start with a pull by the prospective customer.

For any web company, fostering Pull is essential to creating value and engagement.  There is no shortage of great applications and amazing technologies which stagnate due to a lack of pull.  But the greatest economic achievement of being a “pull” platform is in becoming the mechanism by which “push” companies must engage with audiences, paying handsomely to do so. This expectation is why a company like Twitter can be valued in the billions with minimal revenue.

Here are some ways of thinking about fostering pull:

Plan Around Events

Groupon Now is Groupon’s attempt to add Pull to its traditionally Push service. I want to eat, where do I go? Groupon. Every human desire has a natural pull tendency. Being the “first responder” to a human desire is incredibly valuable.

Find Offline Analogies

Most forms of pull fit a predefined social pattern, per the comment on “human desire” above. Before Google, people used phone books (unprompted) to find services. Before email, people would check their postal mailbox, generally at a given time (after the mail was delivered).

Answer Recurring Questions

There are certain types of content that consumers will invariably pull (or want pushed to them). These types of content generally answer recurring questions of a consumer. How much did I spend Receipts, bank websites)?  Where am I going (Google Maps)?  How do I get there (Kayak)? What’s wrong with me (webMD)?

Build Brand and Familiarity

Once one of the above is satisfied, brand and credential storage foster pull. A frictionless and “known” experience catalyze pull for transactional activities. While Amazon, as the largest spender on Google, does a fair amount of push, they also benefit from a tremendous amount of pull when consumers decide to shop. This is a combination of the brand but also their accumulation of user/payment credentials.

There is no substitute for pull in establishing success for a web company; the key is producing something sufficiently valuable in repeat interactions. Reid Hoffman has notedthat “social networks do best when they tap into one of the seven deadly sins.” It’s no coincidence that people have, unprompted, “pulled” those sins since the dawn of humanity.

Say Goodbye to the Long Tail of Product Resellers (online)

The 1980s and 1990s witnessed the slow death of the “mom and pop” general store, replaced by superstores like Walmart that sold everything from butter to guns.  Regardless of one’s position on this trend, it makes classic economic sense: by buying in bulk, Walmart commands better prices with suppliers, and then passes on lower prices to consumers. (Walmart has even been accused of “predatory” pricing to drive mom and pop stores out of business, raising prices after their disappearance.) By aggregating every product under the sun, Walmart can lure consumers in to buy staples (sometimes sold at/below cost), and cross-sell them other impulse items.

There’s one primary reason why Walmart hasn’t completely taken over the world: geography. is a drop in the bucket compared to Walmart’s offline retail presence (remember that people spend far more money offline than online). Some communities keep Walmart out, New York City being one such example. And some people just live far away from Walmart.

But nobody can keep UPS or Federal Express trucks away, and the Walmart effect is going to be even more extreme online. This time Amazon is the big gorilla.

Consumers traditionally shop at retailer A versus B based on the intersecting calculus of five variables:

Price (actual price to consumer + “friction” in ordering process)
Geography (proximity to consumer)
Selection (do they have X in my size, or sell rare item Y?)
Service/Brand (do I trust/like them?)
Experience (is it easy/designed to shop for X?)

Internet commerce has witnessed incredible price transparency, where the Walmart effect can play out without any pesky geographical barrier for most items that UPS will ship; this explains why there are 41,000 shoe stores offline in the US but maybe only 5 of scale online.  That leaves Selection, Service, and Experience.  Selection explains why a small site like is likely thriving, and Service shows how Zappos got to $1B in sales.

The danger is that when a niche becomes big, it will simply be invaded by Amazon, the Internet’s Walmart. I’m pretty certain that if Squash becomes the number one sport in America, Amazon will “go big” and put out of business by squeezing better prices out of suppliers and providing lower prices to consumers, combined with a world-class logistics engine.

If you’re an entrepreneur itching to get into e-commerce, remember that you can’t compete on geography (unless you’re cloning an existing retailer in a region where there is no Amazon), and you can’t compete purely on price.  But here’s what you can do:

Cultivate a better shopping experience: BlueNile is simply a better place to shop for engagement rings. Zappos is a better place to shop for shoes. In some cases, what makes great (every shopping experience is the same) is also its greatest weakness.  Some things are designed to be bought differently.

De-Commoditize: If you’re just another reseller of a generic commodity, you better have a pretty clear advantage outside of price…but these are often tough to come by. is one of very few companies that has out-Amazoned Amazon. If there’s something unique you can add to the order (e.g., proprietary software that consumers can use with the commodity good) it makes it easier to differentiate and provide value to the consumer in excess of a nominally higher price. For example, a vitamin reseller might be wise to develop a smartphone app to remind consumers of pill times…and bundle it with every order.

Build a marketplace for buyers and sellers, don’t be a reseller. Etsy, eBay, IronPlanet, Copart, Elance and others have built great value by focusing on the defensible art of the network effect.  This area is far from played out, and there are many marketplaces waiting to be created for verticals from babysitting to piano lessons. The best marketplaces tend to be for frequently purchased items with a diverse quantity of sellers and few repeated interactions.  For example, you want to eat at different restaurants, but typically go to the same piano teacher for years, so it’s easy to see why OpenTable might be bigger than a piano lesson marketplace.

Distributed commerce: Who can beat Amazon on price? The companies whose products are sold on Amazon!  Outside of the Kindle, Amazon is merely a reseller — marking up the price of others’ products, so those “others” could theoretically beat Amazon in selling direct to consumer.  But most manufacturing companies do not do a very good job selling products direct to consumer, and hate to risk channel conflict.  And consumers prefer to shop at supermarkets, not “silo” markets.  Imagine a world of decentralized commerce — where you can shop at any number of manufacturers within the context of one meta-shopping cart or wallet.  It might be a pipe-dream, but it’s a huge opportunity that could beat Amazon on price and selection if the experience and service components could be filled in.

Preempting Search

Google: 65.8%
Yahoo: 17.1%
Microsoft: 11%
Ask: 3.8%
AOL: 2.3%
(Search Engine Market Share, source: Comscore, August 2010)

Outside of a tectonic shift in search results/quality – think how offering 100x more email storage encouraged people to switch webmail companies back in 2004 — people are not going to ditch Google as their primary search engine. And Google isn’t taking any chances – by paying Dell $1B for their search toolbar to be pre-installed on new Dell PCs, or pushing Android (who’s the default search engine?), they are doing their part to make current habits continue and lock down their whole “supply chain.”

For Google’s enemies, the best way of hurting the search goliath is not to build a better search engine, but rather to give people a reason to stop searching for a wide class of goods and services by preempting search on Google. Given Google’s dependence on harvesting “transactional intent” for its revenue, the key is to move transaction initiation off of Google. The ComScore search marketshare numbers at the top are somewhat meaningless; Google could lose massive revenue while their overall search share, for non-transactional search, stays strong or even grows.

What can preempt Google search — or at least the money-making parts of it? There are two things for Google to worry about: Vertical Search and Intent Generation. Vertical Search will nip away at vulnerable parts of Google in the same way that Etsy, Copart or IronPlanet has nipped eBay – think OpenTable for restaurants, Kayak for travel, Amazon (yes, Amazon) for traditional e-commerce, etc. And Intent Generation catches people further up the funnel, before they search, and delivers them what they want, and gets them to purchase, before they start searching. Intent generation can also spawn impulse purchases and overcome inertia to get people to buy more quickly.

Intent Generation is perhaps the more dangerous, because it is stealing purchases from Google’s clutches – bypassing any kind of search.

Vertical Search

Amazon: if everyone in the world signs up for Amazon Prime (unlimited, free 2-day shipping) and becomes a loyal Amazon customer, who would search for anything shopping-related on Google? We’re a long way from this happening, but imagine Amazon as the “e-commerce search engine” and Google as the “random stuff I’m looking for when not buying” search engine. I believe the long-tail of ecommerce resellers will deteriorate due to economies of scale and lack of geographical differentiation (e.g., 40,000 offline shoe stores, but only 5 of scale online), thereby making Google less relevant for a whole category of searches, and benefiting Amazon as the largest, broadest ecommerce company.

ZocDoc, OpenTable, and Yelp: Since becoming an OpenTable convert and Yelp user, I have not searched once for a restaurant on Google, and I bet these two companies are quickly taking away searches from Google for the dining category. I’m a big believer in ZocDoc, and if that can become the Expedia of medicine (long way to go for that to happen), Google could lose another category.

Kayak and Expedia: Expedia is a great example of what Google needs to avoid. If you’re looking for a hotel in Phoenix, you probably head straight to Expedia, Kayak, or another online travel agency (OTA). Google doesn’t have much to lose here because it’s never had a foothold in travel search, but its purchase of ITA is very strategic as a way of reversing that.

Intent Generation and Catalysis

Groupon: for “impulse” purchases, things like Groupon are pushing offers to consumers rather than relying on consumers to pull (search). The half-million or so Gap Groupons sold on 8/19/2010 represent half a million customers who won’t be searching for Gap, much less any other clothing retailer, on 8/20/2010. Groupon snatched these customers (and their discretionary clothing spend) before they got a chance to search. Some of this is accretive and not preemptive, but consumers only have finite income and a million Groupons every day will have a substantial impact on Google.

Facebook: With more traffic than Google, Facebook only has an estimated 5% of the revenue of its rival. Social recommendations, a catalyst for Groupon’s success, can help preempt search, but these tend to further curate intent rather than harvest it, as Google does. The holy grail is the ability to show the perfect advertisement at the perfect time (precognition, like in Minority Report), something Facebook has a better chance of doing than anyone. The popularity of gaming on Facebook is another angle we have seen be effective – encouraging people to buy something (e.g., a new sweater at Gap) in order to get credits in a game. This is both an example of intent generation and intent catalysis; perhaps you knew you were going to buy a sweater eventually, but you decide to buy it today, and buy it from Gap and not Macy’s, in order to deck out your virtual restaurant on Restaurant City.

Payment Companies: By knowing how much you spend and where, payment companies have tremendous opportunity to change future behavior, generating and catalyzing intent. American Express recently sent me a very nice coupon/gift certificate for Barneys. A month later, when I thought about going shopping, I went straight to Barneys, and didn’t search elsewhere. It preempted my search and changed my behavior. Unlike Groupon, which offers great deals to everyone, payment companies have nonpareil data to use in targeting offers to consumers, and furthermore allowing merchants to target specific consumers. PayPal, American Express, or a resurgent Google Checkout could fundamentally change the nature of ecommerce through intent generation in the same way that Catalina Marketing has altered the CPG and supermarket industries.

With Bing, Microsoft has made a laudable attempt to out-Google Google, but Google has thousands of engineers who can quickly out-Bing Bing. The battle for search is over for now — Google won — but the battle for the underlying revenue is just heating up.