Four Principles for SaaS Pricing, Part II

The Endowment Effect, a behavioral economics term coined by the University of Chicago economist Richard Thaler, contends that people attribute greater value to a good that they already own or have experienced. In 1990, economists Kahneman, Knetsch and Thaler "found that randomly assigned owners of a mug required significantly more money to part with their possession (around $7) than randomly assigned buyers were willing to pay to acquire it (around $3)." This notion guides the pricing strategy of many start-ups: founders believe that once customers experience or own the product, they will ascribe a greater value than if they know about the product but have not yet experienced it.

While the endowment effect is real and observable, it can lead to a lot of mispricings in the market if used insouciantly. The A16Z podcast on Pricing Free elucidates the dangers of pricing using a "bottoms-up" strategy where companies use a low or "freemium" pricing model, thinking that it would obviate the need for a sales team. "Companies that take that approach can become fairly dominant on the bottom-end of the market, but you may have another competitor who comes in with a sales team, and maybe even an inferior product but a couple additional layers of functionality" which will lead to a price war in an already low-price environment, according to Mark Cranney in this episode. So, the moral of the story is, don't price yourself at the bottom of the market just to get customers to try your product because you might get stuck there.

This leads me to the crux of my post, which is delving into the next two of the Four Principles of SaaS pricing (the first two are covered here).

To review, the Four Principles are to price so that you:  

1. Cover Your Costs
2. Increase Your Valuation 
3. Grow Customers 
4. Signal Appropriately

3: Grow Customers: I want to immediately point out that this tenet could be in conflict with #1, cover your costs, and I talk about covering Customer Acquisition Cost in my previous post. Expanding your business, however, requires expanding your customer base, and it is important to acquire those customers at the right price. And while it isn't the primary impetus for my post, I do briefly want to touch on pricing strategies as they relate to pricing "correctly".

Using the Endowment Effect theory, you'd want your potential customers to experience your product so that they can get attached to it, perhaps even to the point of feeling Loss Aversion at the prospect of not subscribing to it. There are many pricing strategies that achieve this by allowing customers to try out the product in some way, and I won't rehash them here. Instead, I'll direct you to this seminal white paper from Bessemer Venture Partners called "Software as a Service Pricing Strategies." Some of the strategies recommended here are:

• Capacity-based freemium models such as those offered by Dropbox 
• Feature-based freemium model such as that offered by Skype 
• Time-based freemium models, meaning that the company gives potential customers a free trial
• Use-case freemium models such as that offered by Adobe to just the readers 

I like time-based freemium models (with the caveat that it really does depend on your business model) because it allows you to achieve the Endowment Effect with the customers and even lets them feel a little Loss Aversion in case they don't sign up. Once they do sign up, however, you need to make sure that you are pricing appropriately to the rest of the market, something I'll discuss more in the Signal Appropriately section.

For more on pricing strategies, be sure to read Tunguz's posts on How to Price Your Product and The Three Part Tariff.

Using Opportunity Costs in CAC:

One twist to calculating CAC that I recommend is that while traditionally you would use just the paid sales and marketing costs to calculate customer acquisition costs, I would consider forgone income, or opportunity costs as customer acquisition costs as well. So if you would normally charge $100 a month for a product and you give away 2 months of it for a free trial, include those two months of foregone revenue ($200) in your fully-loaded CAC calculation. Obviously, your revenue payback period will be longer when you include the opportunity costs into the calculation. Maybe it'll take you 12-15 months to recover the CAC instead of the benchmark 11 months noted by Tunguz. But I think that this is a good metric to measure so that you don't become cavalier about handing things out for free. It will help you quantify the costs of letting your customers try out a portion of the software, and it will be important to take a pulse of where that metric stands over time to make sure you're headed in the right direction.

4: Signal Appropriately:

In the late 1960's George Akerlof wrote about asymmetric information, an idea that later won him the Nobel prize, in a paper called "The Market for Lemons". In a used car market, for example, buyers may be unable to tell the difference between a good car, worth $1,000, or a "lemon", worth $500.  To account for the possibility of a lemon, buyers cut their prices. Astute buyers may offer $500 to start and sellers, knowing the quality of the car, may or may not accept the offer. And so, Signaling Theory was born, asserting that the quality of an item is endogenous, and that the price of a product signals the quality of it. Hence, it is imperative that the product be priced appropriately. Here, I will elaborate on three types of signals: Market Segmentation Signal, Competitive Landscape Signal, and Customer's Return-on-Investment Signal.

Market Segmentation Signal: 

First, determine to which segment of the market your product will be sold. Who will your customers be? If you want to sell to large enterprises, then don't price your product for small businesses. That's giving the wrong signal, and may actually turn off some enterprises who think your price is indicative of the quality. (There are exceptions to this, of course. If you've figured out a way to bring costs down in a way that others haven't and are passing on those savings to your customers, then by all means, price lower. But you'll have to be very clear in your messaging as to why your quality is better AND your price is lower).

Competitive Landscape Signal: 

Second, price relative to your competitors. Here's where a quantitative model called value-based pricing can prove to be useful. Value based pricing works like this:

1. Determine the cost of your closest competitor


2. Outline the functions that your competitors have that your product does not have. What are those components worth. Try and be objective here, and use market based data where possible. Are there two similar products in the market, one with some functionality missing? What is the delta between the two price points? 


3. List the additional functionality that your product provides, and ascertain the value of that functionality. Again, try to be as objective as possible, and use market data where available. 


4. To arrive at the price of your product, take the price of your closest competitor, subtract what you calculated in #2 and add what you established in #3. 

Especially when you have competitors, value-based pricing presents a great way to calculate the differentiated worth of your product and and set your price accordingly. For more on value-based pricing, check out this Harvard Business Review article. 

Customer's Return-On-Investment Signal: 

Lastly, the price of your product can be a signal how valuable your product will be to your customers. Each customer makes an investment when buying your product, and if you can show him that the return on investment (ROI) will be substantial, it will make his decision to purchase the product a lot simpler. That return can come in many shapes, according to Bessemer, such as increased revenue, time savings, efficiency gains, resource or cost savings, or reduced errors. 

I really like this article by Lincoln Murphy of Sixteen Ventures on pricing. In it, he reveals the 10x rule. "If I sell something for $100, I want to provide at least $1,000 in value to them... at least," he says. 

"You do this by understanding your customer’s Desired Outcome, goals, opportunities, problems, etc. You do this by offering Price Anchors that are not competitive services, but what it would take to replicate this in-house, with low-efficiency, high-cost human beings, what you (or the industry) has paid to create this solution, or the fear of not meeting some level of compliance and the costs associated with that."  

So this may be an exercise in quantifying both the tangible and intangible benefits to arrive at an ROI. When you can quantify the ROI for your customer and compare that against your price, which will be 1/10 or lower than the return, you are signaling your product's value in a very quantitative manner, making it very easy decision to purchase your product. 

The primary take-away from my two pricing posts should be that pricing for SaaS involves both art and science. There isn't one equation for it, and you can't just say, "let's price at where our marginal costs will be." At the same time however, there are plenty of quantitative tools that will help guide and set benchmarks for pricing considerations, and those should be utilized in conjunction with the "art" of knowing your customers, your competitors, and the market.

This post was edited by Andrew Amos.

Four Principles for SaaS Pricing, Part I

"I want you to be very careful, because no single decision will impact your company's valuation more than the one you're about to make on pricing." That's Ben Horowitz as quoted on the A16Z podcast on pricing. As the podcast explains, pricing isn't intuitive, and the partners at funds constantly push their entrepreneurs to raise prices. 

Unfortunately, higher prices don't automatically lead to higher valuations. There will always be a price/quantity struggle especially as you start having competitors, as I explained in my previous post on auctions. And if you don't get enough people to buy at the higher prices, your revenue, and hence, your valuation suffers. 

In that vein, I've designed a framework for deciding on how much to charge for your SaaS product. This framework could work for any product, but the first tenet deals specifically with subscription services. Another caveat: this framework applies well to companies operating in post-chasm industries. That is, once the early adopters have embraced a new technology and competitors have entered the market, this framework can be used to set pricing in relation to your own costs and prices of your competitors.

The Four Principles are to price so that you:

1. Cover Your Costs
2. Increase Your Valuation 
3. Grow Customers 
4. Signal Appropriately 

In Part I of this blog post, I'll cover principles 1 and 2, Covering Costs and Increasing Valuation.

1. Cover Your Costs:

Sure, this sounds backwards, but stay with me. Most start-ups worry about how much they can burn given how much cash they have. Covering your costs, however, proves to be an interesting thought exercise, especially when you look out into the future. And the exercise can yield an excellent benchmark. In two years, when you've acquired the number of customers you've predicted, what will your costs be? How should you price to cover those costs? Let's dive in a little deeper.

When the early majority begins adopting a certain technology and slew of companies begin offering solutions to a specific problem (even if the solutions are differentiated), the market begins to resemble monopolistic competition. I wrote about monopolistic competition in my post on newspapers. With low barriers to entry in starting a software company and product differentiation, each technology can set its own prices (as in, they aren't price-takers), but the industry provides a guideline. 

Economic theory in relation to monopolistic competition would tell us that firms need to price so that average revenue is greater than both their average total costs and marginal costs in the short-run to make an economic profit. In the long run, average revenue will equal average total costs. Both graphs are shown below. 

But when speaking of revenue covering costs, pricing subscription products gets tricky. For the most part, costs related to selling the product will exceed the first month's revenue if you're expecting the lifetime of the customer to be several months to several years. So here's how I would think about covering costs. 

COGS: The first thing I would say is that your price every month NEEDS to cover costs of goods sold (COGS). Those are direct  incremental costs that are incurred every time the technology is sold. Usually, the biggest component of COGS for a SaaS business is hosting costs, but can also include set-up fees or platform fees. So, never price so low that the respective monthly hosting costs aren't met by what the customer is paying you every month. I've seen too many companies fall flat because they have negative gross margins; it's hard to price yourself out of that kind of a hole. 

CAC: When considering how much to price in relation to your customer acquisition costs (CAC), which is more of an indication of your sales and marketing costs, a payback period is a good way to think about it. Tomasz Tunguz at Redpoint points out that a median start-up has a payback period of 15 months on a gross margin basis. That is, it takes 15 months of gross profit dollars from a particular customer to cover the sales and marketing costs of a new customer. In terms of revenue, he notes that CAC is typically 11 months of revenue. I think those are good benchmarks to work against, and as Tunguz points out, the faster your payback period, the more working capital you free up for other initiatives. For more on the right way to calculate CAC (using full cost of customer acquisition), check out Andreesen Horowitz's "16 Startup Metrics". 

Total Costs: Lastly, I don't hear much discussion about total cost of operations in relation to revenue. The question many managers ask is, "How much can I burn given the amount of cash I have raised and have on the balance sheet?". And while that mentality is important to insure against insolvency, that burn number usually doesn't factor into how much to price your product, and I think it should. 

When it comes to covering total costs, it is helpful to project yourself into the future, as I mentioned above, and think about what your costs will be when you reach a certain number of subscribers. For example, if you forecast having 100 subscribers over 2 years, how many sales people will you need to get there? 5? 10? 20? What will you have to pay each one? Will you need a Chief Revenue Officer? What will your marketing division look like? Who else will be on your payroll, and what will your rent be? Build a financial model for just the costs, which will help edify how much revenue you will need to cover those costs or make a profit. Once you have the revenue number, you can divide that by the number of subscriptions you anticipate having (in this case, 100), to get to your price. Of course, you don't have to price exactly there, and pricing may actually differ from customer to customer depending on negotiations and how early you are in the process (your earliest customers will likely get a great deal), but this will give you one benchmark to enlighten your process. 

A quick aside on measuring total costs using financial models: I've had some cognitive dissonance here. My accountant side likes Statement of Cash Flows, my investor side likes EBITDA, and my economist side prefers Net Income (to include non-cash costs). I've reconciled all of those to find that EBITDA minus Capex minus Interest Expense is the best proxy of how the core business is doing. It takes out the noise of non-cash items, such as depreciation, and exogenous cash items, such as capital inflows/outflows for debt or equity. If you want to measure just the expenses, take out the revenue line on the EBITDA, subtract capex and interest expense, and you'll be left with just the costs of running the business. Those represent the total costs you'll need to cover. 

2: Increase Your Valuation:

In the second tenet of my framework, I am deliberately eschewing saying "Increase Your Revenue", knowing full well that revenue serves as a large component of valuation. But as an example, Company A with revenue of $100MM with 80% gross margins is very different from Company B with revenue of $100MM with 50% gross margins. Company A deserves higher valuation. 

A step further, Company C with revenue of $100MM with 80% gross margins and 80% annual retention rate is very different from Company D with revenue of $100MM, 80% gross margins and 50% annual retention rate. Company C deserves higher valuation as it likely has lower sales and marketing costs needed to replenish its churn. 

Bill Gurley wrote an article called "All Revenue is Not Created Equal: The Keys to the 10X Club", and I think it is worth a read in its entirety. He discusses the distinguishing factors between high quality revenue companies and low quality revenue companies. Some of the characteristics that warrant a higher revenue multiple, in his opinion, are:

• Visibility of revenue
• Defensible nature or sustainable competitive advantage
• Customer lock-in or high switching costs for customers
• Gross margins
• Scalability or marginal profitability 
• Organic demand versus high marketing spend

So when it comes to pricing for valuation, you want to ensure that you're not squeezing gross margins in order to obtain higher revenue or market share. That is, don't price so low that you're not making any money. Some startups think that once they dominate the market, they will be able to raise prices. That works in some cases (Opentable for example), but you really have to understand the elasticity of demand of your customers, as well as your competitive landscape.  If, in the process of gaining market share you also build up solid network effects, then you are more likely to succeed in raising prices later.

On the other hand, don't price so high that your customers can't afford the product in the long run, and you're constantly spending more on sales and marketing to replenish your customer base. It's a balance.  

In Part II of this blog post, I'll explore the other two tenets of this framework, Grow Your Customers and Signal Appropriately. As you'll see, growing your customer base can come in conflict with covering your costs. Keeping in perspective (i) maximizing valuation and (ii) signaling appropriately to both your customers and competitors, however, should help define your customer acquisition strategy in a cohesive way.  

Marketplace Transactions: Auction Pricing

Platform's Rake: 

If you're in the process of starting a virtual two-sided marketplace, there's plenty of information out there on how much to charge and who should pay. For example, Bill Gurley's article called "A Rake Too Far" illustrates how much existing platforms skim off the top (termed "rake" to follow casinos' syntax) to formulate their revenue. Marketplace Academy's Juho Makkonen expounds on Gurley's findings with his own data to contend that the average rake among virtual marketplaces is 9.2%. Makkonen then gives guidance on figuring out how much rake you should charge for your own marketplace, using variables such as marginal costs, competition, network effects, and transaction size.  

Secondly, there's also plenty of advice on which side to subsidize. Eisenmann et al demonstrate in their paper "Strategies for Two-Sided Markets" that in a two-sided network, there is typically a "subsidy" side, which is a group of users, who, when attracted in volume, are highly valued by the "money side", the other group. Because the strength of the network hinges on the subsidy side, that side ends up paying little to nothing. Eisenmann et al suggested that factors used to determine which side ought to be subsidized should be figuring out which side is more sensitive to price and quality. The following table, from the paper mentioned herein, depicts different two-sided markets and highlights the subsidy-side of the market with a star. 

Marketplace Transactions:

That said, I'm interested in exploring pricing structures for individual transactions within a marketplace. I don't think there's as much thought given to structures of individual transactions, and for that reason, there may be room for optimization. Structures of individual transactions, as I've observed, fall into the following four categories:

1. The platform sets the transaction price. Ride-sharing apps such as Uber and Lyft, as well as Handy, practice this method of pricing their transactions. I view these platforms as psuedo-resellers: they buy a service/product from provider and sell it at a higher price to users. These platforms facilitate the transaction and the meeting of buyer and seller.

2. The seller of the product or service determines the price. Airbnb, Kayak, and Amazon Marketplace fall into this category. These marketplaces mimic a traditional physical market where sellers sell their goods at their prices, and buyers choose what to buy based on prices and other factors.

3. Auctions establish pricing in this case. Ebay is the prime example of this, but other marketplaces make use of auctions as well, such as Priceline for travel, Beepi for cars,and Ubid for electronics. 

4. No dollars are exchanged in this marketplace. Marketplaces such as Opentable and Yelp have no pricing at the transaction level, but there is still a "money side" who pays the platform fee.

Avoiding Auctions:

If we take a look at Gurley's list of how platforms charge, it becomes apparent that not many marketplaces use auction style pricing for their transactions. There could be a few reasons for that: 

1. Friction caused by timing. Auctions may take a while to conduct. If we have to wait to bid on a a car service before booking it, we may forgo taking that time to negotiate and go with another option that's quicker. Therefore, auction pricing may not be the best idea where time is of the essence. 

2. Platform's marginal costs. The platform wants to make sure that it covers its costs, and the easiest way to ensure that might be by setting the transaction prices and basing its commission off the transaction price. For example, if Uber knows that it's marginal cost for a particular ride is $1, it might charge $10 for the ride and take a 20% commission. That way, the company would make a $1 profit ($2 commission -$1 marginal cost). If there were an auction instead and the ride only went for $5, Uber's commission would only be $1 and profit would be $0. 

3. Consistency. The platform may not want to have dynamic pricing for the sake of offering consistent prices and quality. Handy, for example, may want to make sure that you get a consistent quality of cleaning done for the same price, hence setting the transaction price each time. 

4. Sellers' marginal costs. In this scenario, the platform may permit sellers to elect their own prices to protect the sellers' marginal costs. Etsy serves as a good example here since craftsmen on the site may have vastly different costs, requiring them to charge varying amounts. They may also have varying levels of quality of workmanship, allowing them to charge higher or lower prices than their competitors. The Etsy platform wants to cater to that kind of variability in quality and competition by sanctioning sellers to establish their own pricing. 

Types of Auctions:

Auctions have the benefit of allowing price to be close to where supply meets demand without the platform's intervention (more on that below). There are lots of different types of auctions, and I won't go through all of them, but they are English, Dutch, First Price Sealed, Second Price Sealed, and Reverse Auction. You can learn more about all of them here. The two most relevant ones for our purposes are the Second Price, Sealed Bid Auction (used by Ebay) and the Reverse Auction (used by Priceline).

A Second Price, Sealed Bid Auction is where bidders submit what they are willing to pay. Their bids remain undisclosed to other bidders. The highest bidder wins the auction, but pays the price offered by the second-highest bidder (in Ebay's case, a premium is added to the second-highest bid). Therefore, the buyer still ends up having a consumer surplus. It is important to leave some consumer surplus on the table to keep customers coming back to the platform. The seller can set a minimum starting point to cover his costs, so issue #4 from above is solved.

A Reverse Auction requires that the buyer name his own price and the seller either accepts or denies that price. In this case, the buyer starts with a price that is much lower than the maximum that he is willing to pay and then gradually moves higher. Priceline uses this for its travel, which makes the buyers to feel like they got a great deal, but also ensures that sellers are able to get some producer surplus.

Why Auctions Are Important:

Even though there's a myriad of reasons for eschewing auctions, they might be the best way for sellers to discover the right price, and therefore, for a platform to build momentum over its competitors. In my opinion, the auction strategy works best when the outcome of the transaction is not time-sensitive.

As I discussed in my previous post on Ride-Sharing Apps, it is difficult to know how much a customer would be willing to pay for a product or service. In economics terms, it is difficult to draw a demand curve (how much of something would be sold for each price level) in real life. Therefore, it is difficult to price where supply and demand would meet. What happens if you can't price where supply meets demand? Let's look at the following graph:

At P1, the price is too high and supply exceeds demand. For example, you are setting prices for renting out apartments for a night, and you set the price at $100 (P1). You get 10 owners willing to lease out their apartments for the night, but you only get 5 renters. This results in excess supply. The number of transactions is 5, revenue is $500.

In the second scenario, you see the excess supply and decide to lower prices to increase quantity demanded. The next night, you set the price at $50 (P2). The scenario flips, and you get only 5 owners willing to lease out their apartments, but 10 renters. The renters get the apartments on a first-come-first-served basis. Pricing at P2 results in excess demand. The number of transactions is 5, revenue is $250. For this reason (lower revenue), most platforms are more likely to overprice than underprice.

In the third scenario, EP, which stands for Equilibrium Price, is where supply would meet demand. In this case, the right price could be $75, and you might end up with 8 lessors and renters. It's tough to know where to price, however, and it could become a guessing game. Here, supply meets demand, the number of transactions is 8 and revenue is $600 (higher than if you had over-priced).


Let's look at a fourth scenario, where price isn't set by the platform. Each individual apartment owner puts up his apartment for a reverse auction, sort of like Priceline, where the prospective renter submits the daily rent she'd be willing to pay. She'll start off lower than the maximum that she's willing to pay, and he will either accept or reject. That is, she'll submit a bid closer to P2 and move higher if he rejects the bid. There's no guarantee that they'll end up at EP, but if they agree on a price, it will likely be lower than the maximum that she's willing to pay and higher than the lowest amount that he's willing to accept. In large quantities and with free-flowing information (how much did similar apartments rent for on the same day?), prices will move towards EP naturally, without the platform having to do the guesswork.

Auctions can be a very powerful way to grow the marketplace precisely because they allow prices to be set dynamically through supply/demand. Especially for commodity services and products, buyers may be able to pick up what they need for very cheap, giving them the incentive to keep coming back. For example, Handy's cleaners may be willing to work for less on weekdays during school hours, which, if Handy were an auction, would be reflected in the pricing. Or, someone who only needs a sub-par writer should be able to bid a much lower amount for my writing skills on Fiverr than for an excellent writer. Were some of these platforms' prices auction-based rather than pre-set, I'd surmise that prices would be lower in many instances, resulting in higher consumer surpluses and a higher number of transactions. 

From sellers' point of view, price discovery becomes easy since they can price discriminate. Apartments will demand a higher rent price on July 4th weekend, for example, but owners shouldn't have to guess how much higher. Moreover, apartments on July 4th weekend can then be allocated to those who are willing to pay for them instead of on a first-come-first-served basis, which is what would happen if rents were set arbitrarily than through an auction. 

While platforms may fear that dynamic pricing may result in variable or lower commission (rake), I believe that the higher growth (from greater transactions) and the network effects that come with it will make up for that. Furthermore, Gurley gives good reasons for keeping the rake low, which is what may happen if overall prices are low and the platform takes a low percent of transaction as its rake. He states, "If your objective is to build a winner-take-all marketplace over a very long term, you want to build a platform that has the least amount of friction (both product and pricing). High rakes are a form of friction precisely because your rake becomes part of the landed price for the consumer. If you charge an excessive rake, the pricing of items in your marketplace are now unnaturally high (relative to anything outside your marketplace). In order for your platform to be the “definitive” place to transact, you want industry leading pricing – which is impossible if your rake is the de facto cause of excessive pricing. High rakes also create a natural impetus for suppliers to look elsewhere, which endangers sustainability." 

And while having high consumer surplus and low prices is imperative for building the demand side, Mike Russell of Paintzen talks about building the supply side of the marketplace on the Traction podcast. A good marketplace will strive to know and solve for the sellers' pain points in a specific industry, but reducing friction in terms of pricing (rake), as well as providing an avenue for sellers to price discriminate and enlarge producer surplus will make the platform sticky for the sellers. Therefore, auctions can be an elegant way to solve for price discovery where time is not of the essence, and they can help expand a new marketplace by attracting buyers due to higher consumer surplus.  

Ride-Sharing Apps

An Economist's Dream

Steve Levitt captured my sentiment precisely in his recent Freakonomics episode "Why Uber is an Economist's Dream" when he commented about the demand curve, "I've been dreaming of the day I could answer this question, and it probably says a lot about me."

A little background: Levitt, along with Peter Cohen, Robert Hahn, Jonathan Hall and Robert Metcalfe, used consumer reaction to Uber's surge pricing data to estimate a demand curve and the resulting consumer surplus. Uber's data makes this calculation easy to conduct since a consumer's willingness to pay can be tested for almost the exact same product (getting a ride) at many different price levels. They estimated that for every dollar that someone spends, he would have happily spent another $1.60, that $1.60 being the consumer surplus. The economists contend that "back-of-the-envelope calculations suggest that overall consumer surplus generated by the UberX service in the United States in 2015 was $6.8 billion"! Wow, people are willing to pay a LOT to get a ride!

Another dream come true would be figuring out at what price the drivers would have driven at compared with what they actually got paid, or what is known in economic theory as the producer surplus. In order to know the producer surplus, the supply curve itself would have to be estimated by figuring out how many drivers would be willing to drive at every given price. The upward sloping supply curve transpires because the number of drivers willing to drive increases at each price level. 

Ride-sharing apps in reality do espouse many virtues of perfect economics experiments. They are cross-sided or two-sided marketplaces where each side requires different functionality from the platform: the driver requires a rider and a rider requires a driver with a vehicle. Ride-sharing also has merits that other two-sided marketplaces lack: near perfect competition. Although Uber clearly dominates the ride-sharing apps today, I think there is a necessity for multiple apps due to the unique dynamics of this two-sided marketplace and the propinquity of the industry to a perfectly competitive market.

Two-Sided Marketplaces: No study of two-sided marketplaces is complete without revisiting the war between VHS and Betamax. This famous incident has the makings of everything needed for a "winner take all" landscape: people weren't going to buy multiple players or videotapes in both formats before one became the clear standard, and then, manufacturers began choosing sides. In the end, JVC won with its VHS design. Will there be a single winner in the ride-sharing market the way there was in the videotape market? I don't think so, and here's why.

In the article called "Strategies for Two-Sided Markets", Eisenmann, Parker and Van Alstyne proffer that a networked market is likely to be served by a single platform when the following three conditions apply:

1. Multi-homing costs are high for at least one-user side. "Homing" costs comprise of all expenses network users incur-including adoption, operation, and the opportunity cost of time, in order to establish and maintain platforms affiliation.
2. Network effects are positive and strong, at least for the users on the side with high multi-homing costs.
3. Neither side's users have strong preference for special features.

Ride-sharing apps fail test #1. Multi-homing costs are the cost that consumers had of having two different players for videotapes, and that just doesn't exist with ride-sharing. It's easy for riders to have both Lyft and Uber on their phones and to check both for time-until-ride and price. There's no cost of having both apps on the phone and very little opportunity cost of time spent checking both apps. Multi-homing costs aren't even an issue for drivers, as many drivers drive for both companies.

Network effects are strong and positive for both sides and neither side really have a preference for special features, but given that it's really easy to download and compare trips from both apps or to drive for both companies, I don't think ride-sharing is inherently a winner-take-all industry.

Near Perfect Competition: Another reason that ride-sharing apps are such a great case study is that the service of ridesharing is almost (but not quite) a perfectly competitive market. In economic theory, a perfectly competitive market has the following characteristics:

• No barriers to entry: drivers can decide to enter and exit the market as he pleases without too  much trouble. Because of this, there are too many drivers in the market to measure.
• No single firm (in this case, driver) can influence the market price or condition. Each driver is a price taker, and price is determined based on supply and demand.
• Homogenous, almost identical output: one ride may be better than another, but main output is to get from point A to point B. (The slight difference in each experience actually does matter, as discussed below). 
• There is perfect knowledge with no information failure or time lags. The ride-sharing app platforms make the transmission of information much easier than in the non-virtual world.

Ride-sharing ought to be a perfectly competitive market if all information is available. That is, absent regulation, no one entity ought to be able to set a price, given that there are lots of drivers who provide a seemingly indistinguishable service. One would think prices would naturally be set dynamically where supply and demand meet.

In practice, figuring out the demand curve is difficult, as discussed above in reference to the paper that Levitt co-authored. You may accept a ride at $5.00 but not at $5.50, but it's difficult for Uber to know that without asking you about both of those prices. And this is the reason the market will support at least one more competitor to Uber, to ensure that there is a price checking mechanism for a commodity service. Without Lyft or other competitors, would Uber have created the consumer surplus that it did in 2015?

Another aspect that contributes to the likelihood of multiple ride-sharing apps existing is that, unlike in a theoretical perfect competition, each ride is a little bit different from the one before. The variance in each individual experience results in much more uncertainty about customer satisfaction about the output than if the output were a consistent product, like Campbell's soup. It only takes one bad ride for a rider to desire another option for an app. Uber tries to solve for this with the drivers' ratings system to ensure a consistently pleasant experience. That's not fool-proof (the driver can just have a bad day), however, and it doesn't solve for the negative experience being exogenous to the driver.   

The price and treatment checking will happen for drivers too. Anecdotal data suggests that more drivers prefer Lyft to Uber for various reasons, and the normal ebb-and-flow of business policies will cause either one of the apps to be the darling at any given time. Regardless, I think the people on both sides of the ride-sharing equation like having the choice, and that choice serves as a bit of a price-check on the market.  

What's interesting is that obviously Uber the platform itself is in nothing close to a perfectly competitive market: barriers to entry are enormous, which is why the company is so successful. But it is facilitating transactions that, if done correctly, would buoy the perfect competition of the ridesharing world. It could eschew the need for price-checking by competitors if it could figure out how to let the market set the price dynamically based on supply and demand; that is, if somehow, each ride could be priced by the rider instead of the company itself.  This still doesn't solve for riders just wanting the second choice in the case of disenchantment with one of the platforms. 

Ben Thompson Disagrees: Since I get so much inspiration from Stratechery, I'd be remiss if I didn't mention that Ben Thompson kind of disagrees with me. Thompson argues that since the number of riders is far greater than the number of drivers, and as of now, Lyft has fewer number of riders (less demand), drivers will be too busy serving Uber customers, which will lead to a winner-take-all dynamic.

He explains further, "It doesn’t matter that drivers may work for both Uber and Lyft. If the majority of the ride requests are coming from Uber, they are going to be taking a significantly greater percentage of driver time, and every minute a driver spends on a rider job is a minute that driver is unavailable to the other service. Moreover, this monopolization of driver time accelerates as one platform becomes ever more popular with riders. Unless there is a massive supply of drivers, it is very difficult for the 2nd-place car service to ever get its liquidity to the same level as the market leader (much less the 3rd or 4th entrants in a market)."

Thompson also claims that people build allegiances to a brand and persist with that brand, unless they are given a reason to change; it's simply not worth the time and effort to constantly compare services at the moment of purchase, and that Uber and Lyft would ensure that their prices are pretty similar anyway.

To his first point about drivers, I think people act irrationally, and that explains why drivers drive for Lyft and other apps when liquidity might be the best at Uber. Just check out the reasons people give for enjoying Lyft; there are people who will drive with the less-busy app because they like it that way, so I don't think the liquidity of drivers, at least in large cities, will be an issue.

To the second point about riders having allegiances, again, I think people are fickle. One bad experience, which is easy to have in ride-sharing, and they will open up the "other" app for the next ride. Because unlike a can of soup by Campbell, each ride by Uber actually isn't exactly the same.

Also, people do gravitate towards their niches, especially in dense areas, so ride-sharing apps with a twist could also do well. A pet-friendly car service? Paw-fect! A neighborhood service for hauling kids by a certified parent? It takes a village to drive a child, right? And I'm sure there's already lots of cannabis-friendly ride-sharing schemes being conjured up in a few of the states.  

A Note on Financing: According to Crunchbase, Lyft has raised $2B to date while Uber has raised $8.7B. The thesis of having two successful ride-sharing apps is predicated upon having two apps with pretty good brand awareness and customer acquisition ability. And to do those two, you need financing. Lyft, the #2 player has been able to pay for its customer acquisition costs thus far, but if the financing market dries up and the company has challenges raising its next round, it will have trouble keeping up with its customer acquisition goals. A stalled financing market is bad for both companies, of course, but especially bad for Lyft, since it's trying to play catch-up. A prolonged tightness in the fundraising market, therefore, would end up aiding Uber in retaining and expanding its market dominance.

 

Update: We're seeing financing woes play out for Ola, India's Uber rival, as reports indicate that the Indian ride-sharing app may settle for equity financing at a 40% lower valuation. Still, the company would be valued at $3 billion and is planning on raising approximately $600MM, which would still give it enough ammo to acquire customers. Until funding dries up entirely, I don't think we'll see any of Uber's competitors shutting down.



Bundling the Newspapers

Where's the Spotify or Netflix for Newspapers? 

As I either come up against paywalls for, or renew subscriptions at, some renowned bastions of journalism, I keep wondering: why isn't there a Spotify or a Netflix for Newspapers? You know, an all-you-can-eat subscription model where I pay a flat monthly fee and get access to the New York Times, Economist, Washington Post, Wall Street Journal, Financial Times and the New Yorker, just to name a few.

To clarify, we're not talking about news aggregators like Digg, Reddit or Google News, RSS feed readers, or curators that work above the paywall to show you free articles or spinets of otherwise for-fee articles. In that regard, no one beats Facebook; PewResearch Center shows that over 60% of users get their news about government and politics from Facebook. Rather, we're talking about what's referred to as bundling of paid content. 

A Little Background: The term "bundling" was previously used in reference to cable companies selling you a package deal of phone, Internet and cable service. More recently, bundling is referred to those companies selling you a subscription to a panoply of cable channels including ones you'll never watch. The idea is that high revenue generating channels subsidize low revenue generating ones and everyone is happy. But viewers got tired of how expensive those cable bills were and moved to online bundles like Neflix, Hulu, HBO Go and Amazon Prime. The irony is that the bundle was dis-aggregated on TV and re-aggregated online, and can cost up to the same amount.

I bring up this background because newspapers are going through the same thing. Everyone knows that print circulation is down, and the bundle of content that a newspaper used to be is becoming dis-aggregated. You have your "new" news, which is "what is happening right now. That kind of news has become a commodity, thanks to Twitter, Facebook, other social media, and anyone with a smartphone. Then, you have the "think pieces", investigative journalism, and local news, which could demand a monetary value from readers. According to an American Press Institute Study published in February 2016, 78% of US newspapers with circulations over 50,000 are using some kind of a digital subscription model. More specifically, 63% of newspapers are using a metered model, 12% are using a freemium model and 3% are using a hard subscription model. With a panoply of newspapers with the multitude of subscription choices, it's a shame that no one has figured out a simple way to access all the finest journalism for one monthly fee yet. 

To be sure, the idea has been tried and tested, but mainly for magazines. Next Issue Media, now Texture, launched in 2012, and has an all-you-can-eat monthly subscription for magazines, and its competitor, Magzter, is now available for us too.

The Startups: Moving on to newspapers, Blendle and Inkl are the two visionaries trying to solve this problems, but they're still in their early stages. Inkl allows you to read newspaper articles by paying either by the article (10 cents) or by a monthly subscription ($15). Blendle, which is in beta, is a pay-per-article model. The pay-per-article model makes model may make sense for journalism, because unlike music, t.v. shows or movies, it is unlikely that readers will consume an article repeatedly. 

The pay-per-article model doesn't allow for higher cost articles to subsidize lower cost articles, which wouldn't matter as much if, like songs in an album, all articles in a publication were written by one author. With the prevalence of this type of model, for better or worse, we'd eventually lose the articles for which not enough readers wanted to pay. Both models tout ad-free reading zones without click-bait, but if your revenue stream is per article, headlines will have to be flashy enough to ensnare readers. Regardless, neither of these bundlers have been able to get all the good papers, and it may be impossible to make that happen (#analysis).

 

The reason goes back to simple economics (you knew this was coming). The market for a newspaper most closely resembles that of a monopolistic competition, where firms have many competitors, barriers to entry are low, but each firm offers a slightly differentiated product. In this case, newspapers aren't perfect substitutes for one another and can set their own price, albeit taking the industry price as  a guideline. Granted, most newspapers aren't doing super well, but Pulitzer Prize winning papers can typically elicit a higher price than other papers. In the long-run, papers price where Average Revenue = Average Total Costs as shown below.

So as long as news organizations are recovering their average total costs and maybe a little more, they should sign up to be part of he bundling service, correct? Not so fast. In the short-run when differentiation between papers is sharp, some can charge above Average Total Costs Curve and make economic profits, as shown below, lowering their incentives to play well with others.

Comparison to Music: There's a reason that blockbuster artists like Adele, Beyoncé, and Taylor Swift have eschewed Spotify, and perhaps we can draw a parallel to the news world. Ben Thompson writes in Stratechery, "The problem with Spotify is that at a very fundamental level it treats music as a commodity. You can’t choose where your $10/month goes based on the emotional impact of a song."

 

He makes two great points:

1. Money paid to an aggregator is NOT money paid to an artist, and an explicit purchase makes a fan more loyal, not less.
2. Grossing $10 per customer in a single shot is far more lucrative than pennies from the exact same people when they access your songs in a streaming service.



So it comes down to compensation for those who can demand higher price tags, the Adeles of the newspapers, so to say. In a way, the pay-per-article solves the problem #2, and the pennies per article really do go directly to the papers. It doesn't necessarily create loyal customers, however. So there will always be the Adeles who remain independent, unless you can somehow figure out how to solve the loyalty and price problem (some ideas below).

Twists to the Established Model: Here are some ideas about how to make the newspaper model more efficient, both for readers and publishers. 

1. A Facebook add-on that allows you to instantly buy articles when you run up against a paywall.
2. The pay-per-article platform should price discriminate, charging a few pennies more articles coming higher priced newspapers or award winning authors than lower priced newspapers. Click and purchase data will come handy here, and prices can be adjusted as economics change.
3. A choose-your-own newspaper bundle where the marginal cost of adding each new subscription declines. The papers get paid a % of ad fees (I don't mind having ads, and if that's how authors get paid, I'm for them) from the sites, and the bundler gets the rest, and the reader gets a great deal.

Lastly, while there is a case for writing all news stories for Facebook we know there have been plenty of fake stories on the social network. Facebook, not being a newspaper, doesn't have the obligation to take them down, so there has to be branded news to signal quality, whether that comes in a bundled format or not.

Virtual Reality and Theory of Modularity

If you're anything like me and don't know much about mobile gaming, you haven't paid a ton of attention to virtual reality. It seems like a cool technology for sure, but useful primarily for playing computer games. That's what I thought, at least, until I heard Marc Andreesen on the A16Z podcast in August of 2015. 

After hearing Andreesen, I realized that VR can change lives for many people around the world. Most people's, especially those in war torn or developing countries, "real reality" is nothing to envy. VR can give people an experience that would be nearly impossible in actual reality. Imagine a student from anywhere in the world being able to sit in a Stanford classroom and interact with students and professors as if he were actually there. VR simulations are already being deployed outside of gaming; they already have the ability to create 3D models of patients' anatomy, can make history and science classes come to life, and allow auto manufacturers to test drive a car that doesn't yet exist. 

Despite all the promises of VR, the technology has mostly remained inaccessible to the masses, and Google intends to do something about that. The month of May brought us Google I/O and its most talked about announcement of the mobile virtual reality platform called Daydream. Daydream follows Google's initial foray into VR, which was in 2014 through a cheap, disposable headset called Cardboard. But Cardboard came with a latency problem which could make users sick, a problem  that was solved by higher-end VR headsets such as the Oculus Rift, Oculus+Samsung's Gear VR and HTC's Vive. 

Daydream's introduction fomented a debate about who is likely to win the VR headset race. Daydream, according to Gizmag, is more reactionary than innovative. Gizmag argues that the quality of Daydream will likely continue to lag behind the higher-end headsets, which have multiple controls and are already working on things like positional tracking. "Daydream shouldn't pose much of a threat anytime soon," claims the article. 

What makes Daydream more interesting is Google's announcement that the VR platform will be based on the next Android version called Android N. Consequently, many phones that run Android will come optimized to run VR experiences and be Daydream ready. These phones will be certified by Google and will be required to have various VR friendly components such as "high-quality sensors for head tracking or screens that can reduce blurring by showing images in extremely short bursts," according to The Verge. 

In addition, Google will curate the Play Store to have content optimized for Daydream. In fact, "Google VR head Clay Bavor specifically mentioned Hulu, Netflix and Lionsgate as some of the companies bringing media content to Daydream," according to Variety. Moreover, Bavor mentioned that the Company has already built YouTube from the ground up for VR. The accessibility of Daydream is expected to shift relevance of VR from niche PC gamers to any mobile user who wants to experience various apps in a different way. 

Google's playbook for Daydream is straight out of the renowned Clayton Christensen's Innovator's Solution. Christensen begins by defining interdependence and modularity. He says that "an architecture is interdependent at an interface if one part cannot be created independently of the other part- if the way one is designed and made depends on the way the other is designed and made." He goes on to say that "a modular architecture specifies the fit and function of all elements so completely that it doesn't matter who makes the components or subsystems, as long as they meet the specifications. Modular components can be developed in independent work groups or by different companies working at arm's length." 

Each product can have some components that are interdependent and some that are modular. An iPhone and iOS are interdependent, but the apps on the iPhone are modular.

What kind of architecture is best for VR today? Christensen argues that when a product's functionality is not yet good enough to address customer needs, firms that build their products around proprietary, interdependent architectures enjoy a competitive advantage because standardization in modularity takes too many degrees of design freedom away from engineers and performance cannot be optimized. He goes on to say that, "one reason why entrant companies rarely succeed in commercializing a radically new technology is that breakthrough sustaining technologies are rarely plug-compatible with existing systems of use." So if we think today's VR technology isn't good enough for mobile, then an interdependent structure, or the one Google hopes to create with Daydream, could win out because it will be easily fit and function within all the Android phones.

Modularity, however, becomes the dominant design when products become good enough, and there is a performance surplus from the product. Once the requirements for functionality and reliability have been met, products begin competing on speed of upgrades or responsiveness to customers. With modular architectures, companies can introduce new products faster because they don't have to redesign everything. "Whereas in the interdependent world, you had to make all of the key elements of the system in order to make any of them, in a modular world you can prosper by outsourcing or by supplying just one element." If we think that mobile VR technology is good enough, then those companies that innovate faster and are more responsive to consumer needs, such as Oculus and HTC, would become the dominant players.

The trajectory of product architecture as defined by Christensen, is depicted in the figure below:

The Oculus and HTC VR solutions are almost textbook examples of modular designs. Baldwin and Clark in "Managing in an Age of Modularity" note that modular designers "rapidly move in and out of joint ventures, technology alliances, subcontracts, employment agreements and financial arrangements as they compete in a relentless race to innovate." Baldwin and Clark note that since designers achieve modularity by partitioning information into visible design rules and hidden design parameters, modularity is only benefitial if the partition is "precise, unambiguous, and complete."

When it comes to mobile VR, it will be important for the VR architecture to be designed well to fit the phone's architecture. Here, Google's Daydream presents a truly mobile experience, not one that we initially for PCs and later fitted to smart phones. In that context, it wins by providing accessibility of VR to a broad range of users. As John Nagle of Gyoza Games said, "With the launch of Daydream, Google is again further democratising VR, making it accessible to a vastly broader audience than was ever before possible."

As VR becomes more accessible, Google's platform will be able to provide developers with standards and design rules, allowing for modularity in applications that could be used in a VR setting. Nagle goes on to say, "From a development perspective, including a controller and providing a ‘Daydream Ready’ hardware spec is a great advantage, because it means that we can focus on building great content instead of spending time and money developing and testing on so many disparate hardware platforms."

Therefore, Google wins at first with its interdependent design, even if it doesn't have the best VR solution, just because it is able to make the technology accessible for both users and developers. The VR technology just isn't seamless enough yet with mobile phones to be useful to the majority of smart phone owners. Google's reference device that will allow manufacturers to bring their own headsets will create a much needed standardization in the market, which will allow developers to focus on content rather than on hardware.

As Daydream as a VR platform becomes more prevalent, however, the industry will move towards modularity. That is, phone makers can create their own headsets as long as they meet Google's specs. Developers can create content that will align seamlessly with Android. And the market will move towards a performance surplus as VR vendors innovate quickly to offer performance and the rich content to meet user demand.