Business Model Analysis, Part 9: Outsourcing

This post is part of a series on business model analysis for entrepreneurs. The first post in the series presents a comprehensive list of issues (available as a downloadable PDF) entrepreneurs should consider when designing a business model. Others delve into specific issues; this one looks at factors that determine whether a startup should keep key activities in-house, versus outsourcing them.

The key word in the last sentence is "key." Serial entrepreneur Furqan Nazeeri has argued that startups, because they are resource constrained, should outsource all activities that do not contribute to long term, sustainable competitive advantage. VC Fred Wilson generally agrees, and notes that startups often make the mistake of outsourcing product development due to a lack of in-house skill, but in doing so they sacrifice the crucial ability to iterate the product designs (a point echoed by Vivek Wadhwa). Wilson also says that startups often outsource customer service due to perceived cost savings, but in doing so they forfeit valuable customer feedback.
Another consideration in deciding whether to outsource key activities is the prospect of asymmetry in bargaining between a startup and powerful partners. HubSpot's Dharmesh Shah has warned about the many risks that a startup confronts when negotiating with big companies. Serial entrepreneur and VC Marc Andreesen has likened dealing with big companies to the long, frustrating, and harrowing pursuit of Moby Dick.

I won't try to expand on those insights here. Rather, I'll focus on the microeconomics of in-house vs. outsource decisions, which, in economists' parlance, are choices about vertical integration. According to Yale Professor Oliver Williamson, there are economic advantages to completing transactions between two units within a vertically integrated company—rather than between two independent firms—when the transactions entail high levels of uncertainty, small numbers bargaining, and asset specificity.  With transactions between independent firms, uncertainty makes it difficult to draft a contract that specifies each party’s obligations under any contingency that might arise.  Absent a complete contract, the parties periodically will need to renegotiate transaction terms. If either party is subject to “small numbers bargaining,” that is, if it has few potential transaction partners, then that party may be vulnerable to hold-up when it renegotiates. Finally, if either party’s assets are tailored for a specific transaction type and cannot be redeployed into other uses, then failing to complete a crucial transaction—for example, securing an input required for production—may lead to bankruptcy with little liquidation value.

Startups frequently face the conditions that encourage vertical integration. By definition, they confront high levels of uncertainty. Also, when they target new markets with radical innovations, startups may require access to idiosyncratic assets controlled by only a few potential partners.

However, vertical integration poses challenges for resource-constrained startups, because it often requires major investments. Cake Financial, a service that gave investment advice to consumers based on analysis of their online stock trades, illustrates this dilemma. Cake’s founder had a choice between building software that could extract a customer’s trading data (with their permission) from their online brokerage accounts, or licensing access to the data from a firm that had already developed similar software. Concerned about that firm’s fees and whether it would be responsive to a small startup’s needs, Cake’s founder chose to build the software. This consumed most of the $9 million in venture capital that Cake had raised, and put the startup in a precarious position when demand for its service was slow to emerge and then capital markets slammed shut during the 2008 global economic crisis.

Business Model Analysis, Part 8: Crossing the Chasm

This post is part of a series on business model analysis for entrepreneurs. The first post in the series presents a comprehensive list of issues (available as a downloadable PDF) entrepreneurs should consider when designing a business model. Others delve into specific issues; this one provides an overview of Geoffrey Moore's concept of crossing the chasm.

In his classic book on hi-tech marketing, Moore observed that customer adoption of revolutionary new technology products follows a predictable life cycle. Early adopters, according to Moore, are visionaries seeking breakthroughs; they can imagine the new product’s benefits before they have been proven. Because they are tech-savvy, early adopters can self-assemble complementary hardware, software, and services needed to use the new product, and can cope with its inevitable initial bugs. In the next stage of the product life cycle, the early majority—a much larger group—are pragmatists who will only buy a standardized product that has clearly proven benefits. They demand a “whole product solution”—an easy-to-use, reliable bundle of all necessary hardware, software, and services—supported by a reputable firm.

Moore observed that peer-to-peer references are crucial in driving technology purchase decisions. However, early adopters are not considered to be credible references by the early majority: there is a chasm separating the two groups because they rely upon such different purchasing criteria. As a result, new products that are successful with early adopters often stall when startups try to sell them to the early majority.

Moore’s prescription for crossing the chasm is to target a single segment within the early majority; to engineer a whole product solution with clear benefits for this segment; and to overwhelm the segment with an integrated, intensive marketing campaign. From this beachhead, the firm can then leverage referrals to capture other early majority market segments. Moore likens this strategy to World War II’s D-Day, when the Allies landed a massive force at Normandy as the first step in their invasion of Europe.

Most startups targeting fundamentally new markets will not encounter the chasm until they are a few years old; in the meantime, they will be busy cultivating early adopters. Consequently, seed-stage ventures can probably ignore the chasm risk as Steve Blank points out in Four Steps to the Epiphany. However, a “D-Day” strategy requires plenty of planning, so entrepreneurs should begin to watch for the chasm as their startup matures.

Business Model Analysis, Part 7: Bundling

This post is part of a series on business model analysis for entrepreneurs. The first post in the series presents a comprehensive list of issues (available as a downloadable PDF) entrepreneurs should consider when designing a business model. Others delve into specific issues; this one provides an overview of bundling.

Bundling entails selling, in a single transaction, two or more items that could conceivably be sold separately.  A printed newspaper, for example, is a bundle of news stories, classified ads, comics, obituaries, stock tables, sports scores, etc. Microsoft Office bundles several productivity applications in a software suite.

The ubiquity of bundling is not an accident: the strategy can provide significant benefits, including superior surplus extraction (i.e., capturing a greater share of customers’ willingness to pay), economies of scope, product design improvements, and strategic advantages.

However, pursuing a bundling strategy can be challenging for a resource-constrained startup Most early-stage ventures strain their capabilities to develop and sell a single product, so bundling multiple products from the outset may not be an option. Nevertheless, entrepreneurs should keep the potential benefits from bundling in mind as they design their business models and plan for future product launches.

These benefits include :

• Surplus Extraction. Economists define consumer surplus as the difference between a customer’s willingness-to-pay (WTP) for a product and its price. When a firm offers the same price to all customers (i.e., when it does not engage in price discrimination via negotiated pricing, auctions, etc.), bundling two or more products may allow the firm to extract a larger share of total available consumer surplus — and earn higher profits — than it would from selling the items separately. To illustrate this potential benefit, consider an example with two customers, Jack and Jill, and two products, A and B, each sold by the same monopolist. A and B both have zero marginal cost (as with many information goods), and the firm must offer a single price for each product to all customers. Jack’s maximum WTP is $10 for A and $4 for B. Jill’s maximum WTP is the reverse: $4 for A and $10 for B. If the firm sells A and B separately for $4 each, it will sell one unit of each product to both customers and earn total profits of $16. By pricing each product at $10, it will sell one unit of A to Jack and one unit of B to Jill and earn profits of $20. However, if the firm offers an A+B bundle for $14, it will sell the bundle to both customers and earn profits of $28. Bundling is more likely to increase profits in this way when: 1) the marginal cost of bundled items is low or zero; and 2) the correlation of consumers’ valuations for individual items is weak. Weak correlation means that when customers evaluate the items in a bundle, they don't all have the same favorites. Real world examples of surplus extraction through bundling abound. Some HBO subscribers, for example, value its recent theatrical films highly; others love its original series (e.g., True Blood, Entourage); still others are drawn to HBO's boxing matches or concerts. This varied programming mix allows customers with very different preferences to each justify paying a $10 monthly subscription fee.
• Economies of Scope. Compared to selling items separately, bundling can also reduce a firm’s costs. Firms can realize economies of scope in customer acquisition activities because they can sell the bundle with a single marketing message, rather than two separate ads or two sales calls for two distinct products. Likewise, economies of scope in production are available when integrated designs leverage shared components (e.g., a single screen and battery when combining a cell phone and MP3 player).
• Product Design. Integrated designs may also yield quality advantages through simplification of interfaces, as with Google’s use of a common password across all of its applications and its integration of Gmail into its search service.
• Strategic Advantages. Under certain conditions (described in this academic paper), bundling may allow a company that monopolizes a market for one product (call it "A") to profitably leverage its way into the market for a crucial complement to A (call it "B") which previously was supplied only by independent companies. Products are complements when they are frequently or always consumed in tandem (e.g., browsers and PCs; beer and pizza). By offering only an A+B bundle (i.e., "tying" A and B and not allowing customers to buy A separately), the market A monopolist forecloses access to its customers, denying standalone suppliers of B the opportunity to sell to them. The resulting reduction in revenue weakens the standalone suppliers of B and may even force them to exit the market. Of course, such a strategy can run afoul of antitrust law, as Microsoft discovered when it tied the Internet Explorer browser to its monopoly Windows operating system.

Business Model Analysis, Part 6: LTV and CAC

This post is part of a series on business model analysis for entrepreneurs. The first post in the series presents a comprehensive list of issues (available as a downloadable PDF) entrepreneurs should consider when designing a business model. Others delve into specific issues; this one provides an overview of customer lifetime value (LTV) and customer acquisition cost (CAC) calculations, and how they are used in startups.

As they decide whether to race for scale, startups face a function that relates their long-term payoff—the net present value of future cash flows earned as a result of acquiring new customers during the current period—to their level of current-period investment in customer acquisition. Entrepreneurs cannot observe this function directly; they must estimate its shape based on customer and competitor responses to the firm’s initial marketing efforts and historical information for similar products. Early in a product’s life cycle, with limited data available, these estimates will be imprecise, so entrepreneurs should be wary of overconfidence biases that may lead them to overinvest in growth.

The function relating long-term returns to current-period investments in customer acquisition will have an inverted “U” shape. Up to some point—I* in the figure below—increasing investments should boost a firm’s net present value (NPV), but at a diminishing rate as the cost of acquiring each additional customer rises. Beyond the value-maximizing point, I*, it costs more to acquire additional customers than they are worth. Put another way, if you race too hard, or not hard enough, you will hurt your long-term returns.

Net present value per new customer—that is, customer lifetime value (LTV) minus customer acquisition cost (CAC)—may decline with investment levels for four reasons:

• Broadening Beyond the Firm’s Natural Market Segments. Most products have attributes — features, service quality, brand image, etc. — that match the needs of some customer segments better than others. When racing for scale, a firm may target prospects outside of the customer segments that find its product most appealing. Large price reductions or promotional expenditures may be required to convert these prospects into buyers.
• Prematurely Soliciting Mainstream Customers. Mainstream customers often defer purchases until early adopters have “tested the water” and can verify that a product has an attractive value proposition. If mainstream prospects are solicited prematurely, conversion rates may be low unless inducements are offered.
• Pricing and Promotional Battles. Aggressive moves to capture share may precipitate pricing and promotional battles with competitors.
• Scalability Constraints. The operational strains of rapid growth may degrade product or service quality. This can hurt solicitation conversion rates and raise acquisition costs per new customer. The resulting damage to the firm’s reputation can also put downward pressure on pricing and customer retention rates, further reducing the payoff from racing.

NPV Impact of Customer Acquisition Investments

The ratio of customer lifetime value (LTV) to customer acquisition cost (CAC) is a useful measure of the productivity of customer acquisition efforts. LTV equals the discounted present value of variable contribution—revenues minus variable costs—earned over the life of a typical customer’s relationship with a company. LTV does not deduct customer acquisition costs (CAC). Unless a firm exhibits viral growth or increasing returns to scale (scenarios discussed below), CAC = LTV is the most that a company can profitably afford to invest to acquire a new customer.

Calculating LTV and CAC

Calculating a firm’s maximum customer acquisition cost (CAC) based on the average lifetime value (LTV) of a customer involves four steps:

Step 1: Determine contribution per customer. Variable contribution equals revenue earned less all variable costs incurred in serving a customer in a given year, excluding marketing costs related to customer acquisition. A back-of-the-envelope approach for calculating the average contribution per customer—usually sufficient for providing a rough “reality check” on a business model—simply subtracts a company’s total variable cost from its revenue for the most recent period, then divides the remainder by the average number of customers served during that period.

A more sophisticated approach recognizes that 1) contribution per customer may vary substantially for different customer segments, and 2) the annual contribution per customer is likely to change over the life of a customer relationship. With respect to the latter point, a company may be able to increase its prices over time. Also, the company should be able to collect information about the customer’s preferences and may be able to use that information to cross-sell related products. Finally, over time, variable costs incurred in serving a customer tend to decline as a percentage of revenues for two reasons. First, experienced customers tend to generate fewer customer service inquiries because they “know the ropes.” Second, as a company grows, it typically can improve its operational efficiency and realize volume discounts in procurement.

Step 2: Determine customer life. To calculate the average length of a customer relationship, one can employ the formula 1/x, where “x” is the annual customer churn rate, that is, the percentage of customers that terminate their relationship with a company from year to year. So, if a company retains 70% of its customers each year, then the average customer life is 1/0.3 = 3.33 years. Of course, the average length of a customer relationship may vary widely for different customer segments.

Step 3: Calculate LTV. The annual cash flows per customer calculated in Step 1 are discounted to their present value, using the number of years for the duration of a customer relationship calculated in Step 2.

Step 4: Calculate CAC. A back-of-the-envelope approach for calculating the average cost of acquiring a new customer takes total sales and marketing expense incurred during a period, then 1) subtracts any costs related to retention and usage stimulation efforts targeted at existing customers (e.g., time spent by sales reps calling on existing accounts, rather than prospecting for new customers); and 2) divides by the total number of new customers acquired during the period.

As with the other inputs described above, average customer acquisition costs will vary considerably by customer segment. Likewise, different acquisition methods may have very different costs. Each method will be subject to decreasing returns during a given period as available prospects in the most attractive segments are converted into purchasers and the company is then forced to target prospects for whom the product is less compelling. For this reason, companies employ cohort analysis: they measure the productivity of their marketing efforts—and optimize their efforts accordingly—by tracking, over time, the LTV and CAC of “vintages” of new customers acquired during a given period through different marketing methods.

Step 5: Compare LTV and CAC. In theory, for any given new customer, a company can afford to increase CAC up to the point that CAC = LTV for that customer. Of course, if CAC = LTV for every new customer that a company acquired, it would not generate enough contribution to cover its fixed costs. For this reason, many companies employ a target LTV/CAC ratio. For many software-as-a-service businesses, for example, the target ratio is 3:1.

Calculating LTV and CAC with Virality and Network Effects

When calculating LTV and CAC for businesses that exhibit virality and/or strong network effects, complications may arise:

• Virality. The maximum amount that a firm can afford spend to acquire a customer through paid marketing methods should take viral growth opportunities into account. In theory, in calculating the value of a “seed” customer, one should reflect the LTV of every additional customer who will be subsequently acquired due to free, viral mechanisms that are put in motion by the seed. This could conceivably involve a chain of viral acquisitions that stretches for many years into the future. In practice, it is more conservative to credit the seed customer with only one year’s worth of viral acquisitions. A straightforward way to do this is to multiple the LTV directly generated by the seed customer by the 1.0 + V, where V is the viral coefficient for a new customer of that cohort type.
• Network Effects Generate Value for Other Customers. When a business exhibits increasing returns to scale due to network effects or scale economies in production, acquiring a customer in the current period increases future cash flows from other customers. In calculating LTV, this incremental value should be added to the present value of future cash flows derived directly from a new customer, as illustrated in the technical appendix to Part 2 of this series.
• Variable Costs Depend on Network Density. When networks have a spatial component, the physical proximity of customers may be an important factor in determining variable costs. For example, an online grocery service can achieve much lower delivery costs per customer when a driver’s stops are just a few minutes apart. Hence, to calculate contribution margins accurately, managers need a reliable forecast for network density.
• Two-Sided Networks. Two-sided networks have two distinct user groups whose respective members consistently play the same role in transactions, for example, cardholders and merchants in American Express’s credit card network; job seekers and recruiters in Monster.com’s online recruitment network. To mobilize a two-sided network, platform providers must attract users to both sides—typically simultaneously.
• In this context, LTV calculations can become very complicated; explaining their mechanics is beyond the scope of this post. In fact, marketing scholars have only recently begun to develop statistical models that can be used to estimate LTV in two-sided networks; my colleague Sunil Gupta has done some pioneering work on this front. Consistent with the previous point, these models factor the impact on future cash flows from Side B users in estimating the value of additional Side A users, and vice versa.
• In most companies serving two-sided networks, distinct organizational units will be charged with marketing to the separate sides; these units must coordinate their plans to ensure that overall marketing spending is optimized. In particular, it is important to avoid double counting the profit increase attributable to network effects when separate organizational units each calculate LTV for their respective sides. Complicating matters further, certain marketing programs will impact user acquisition rates on both sides (for example, Monster.com’s Olympic sponsorships, which built awareness among both recruiters and job seekers). Managers must determine how to allocate these expenses across the two sides when calculating CAC.

For more on LTV and CAC analysis, see F. Reichheld, The Loyalty Effect; Blattberg et al., Customer Equity: Building and Managing Relationships as Valuable Assets; Blattberg & Deighton, “Manage Marketing by the Customer Equity Test”; and posts by David Skok of Matrix Partners.

Business Model Analysis, Part 5: Virality

This post is part of a series on business model analysis for entrepreneurs. The first post in the series presents a comprehensive list of issues (available as a downloadable PDF) entrepreneurs should consider when designing a business model. Others delve into specific issues; this one provides an overview of viral customer acquisition dynamics.

A product grows virally when its use spreads through direct, customer-to-customer transmission. Viral growth occurs through four different mechanisms listed below. With the exception of incentives, these mechanisms do not entail any marketing expenditures, so business models that harness strong viral growth can be very attractive.

• Direct Network Effects. To function properly, some products must be used jointly by two or more parties. These products are said to exhibit direct network effects, because their users interact directly. For example, early versions of Skype required both a call originator and recipient to use Skype software. When one party who already has such a product wishes to interact with another who does not, the first party can contact the second party to suggest that they acquire the product.
• Word-of-Mouth. Even if they do not enjoy direct network effects, products can spread virally when a happy customer recommends them to another party, as when a satisfied diner suggests a restaurant to a friend.
• Casual Contact. Like the common cold, some products can spread virally through casual customer-to-customer contact. For example, the free, web-based email service Hotmail grew explosively in 1996 after its founders added a link at the bottom of users’ emails that simply said, “Get your free email at Hotmail.”
• Incentives. Many companies structure incentives that encourage their existing customers to recruit new customers, for example, MCI’s 1990s “Friends-and-Family” plan, which offered reduced long distance rates for calls between MCI customers in a circle of up to twenty members.

Virality and network effects are often conflated and confused, so the distinction between them warrants clarification. It should be clear from the list of mechanisms above that not all products that spread virally exhibit network effects. Likewise, not all users of products with network effects are acquired through viral, customer-to-customer transmission mechanisms. Specifically:

• A new user of a product with direct network effects might sign up based on press coverage or advertising, then discover parties with whom they can interact after using the product. This pattern was evident in the rapid growth of MySpace, Second Life, Twitter, and the question-and-answer service Quora.
• Direct network effects are distinguished from indirect network effects in a two-sided network, in which growth in one side’s user base (e.g., Android phone owners) attracts more users to the other side (e.g., Android application developers), and vice versa. With indirect network effects, the mechanism of attraction is the aggregation of a larger base of users, rather than contact between individual users.

Many startups combine more than one viral mechanism in their go-to-market plan. Dropbox, for example: 1) harnessed a direct network effect when users employed the service to collaborate on documents; 2) benefited from word-of-mouth referrals from loyal customers; 3) acquired customers through casual contact when users emailed links that allowed recipients to download (without installing Dropbox) files stored in the sender’s public folder on Dropbox; and 4) offered a two-way “user-get-user” bonus, that gave both the inviter and recipient an additional 250MB of free storage.

Viral Coefficient

A firm’s viral coefficient is calculated as the number of additional customers subsequently acquired through viral mechanisms for every new customer initially acquired. Startups that rely heavily on viral growth should track their viral coefficient overall and by customer cohort—that is, for each “vintage” of new customers acquired during a given period through different types of marketing program employed by the firm. As shown by the table below, a viral coefficient greater that 1.0 yields self-sustaining growth from an initial “seed”—that is, a batch of new customers acquired in period 1. In the table, we assume that a seed group of 1,000 new customers each purchase one unit of a firm’s product in year 1. These seed customers do not repurchase the product, but through viral means, they attract some additional customers who purchase in year 2, who in turn attract some more customers in year 3, and so forth.

	Number of New Customers
Viral Coefficient	Year 1	Year 2	Year 3	Year 4	Year 5
0.3	1,000	300	90	27	8
1.0	1,000	1,000	1,000	1,000	1,000
1.3	1,000	1,300	1,690	2,197	2,856

When modeling viral growth dynamics for customer relationships that have a multi-year life, it is important to be specific about whether the viral coefficient should only be applied in year 1, or in each year. In some contexts, new customers are likely to quickly exhaust word-of-mouth recommendations or other viral mechanisms (e.g., opportunities to leverage “member-get-member” bonuses).

David Skok of Matrix Partners discusses viral coefficients in depth in this post, and Adam Penenberg’s book Viral Loop provides many examples of viral customer acquisition.