None of the successful company CEOs I’ve worked with seek out innovation for innovation’s sake. They are after something else instead.

In an uncertain environment, with pressure to grow their business, how should they evaluate opportunities?

The last 20 years have left us with new tools and frameworks to evaluate opportunities for business growth. You probably know many of them…

Understanding the customer, future customer, potential customer. Studying Disruption Theory and Blue Ocean Strategy. Determining whether you are (or will operate) in an existing market, resegmented market, or new market. Understanding your business model type, such as direct, indirect, or marketplace. The way you look at trends. Even learning from copycats…

Then there are common ways to bring innovation in house in order to grow. Companies organize hackathons around their industry or tech. Longer-term efforts include the launch of an internal accelerator or incubator program. There are many flavors of these programs. I should know — I’ve operated three different accelerator programs, with three different focus areas, on three different continents. An internal program done well can support early-stage teams, provide access to high-level contacts, access to customers, and expertise. It can also be a distraction and produce minimal results.

What these in-house programs also can offer is a portfolio approach to new projects. But running in-house accelerator, incubator, or innovation programs can be difficult to do at scale. What other approach could we use?

I think in terms of maximizing “shots at goal.”

Dilemmas and distractions

Years ago when I worked in a new group at AT&T that was operating a new service to displace legacy business lines, we needed to operate separately from other divisions. We were culturally and financially at odds with the rest of the legacy business run by other parts of the company. Our main competitors were actually other departments of AT&T. I used to think this internal competition was an anomaly, but after time at other large organizations, from financial services to a university, I see how common it is. My own winning strategy for dealing with this: ignore the competition and focus on customers.

Most do not take that approach. One company I know has run over 30 innovation projects without gaining a single customer. Each project was the brainchild of someone in-house. Each project had a budget. And none of them spoke to customers before spending on R&D and production. Result: years wasted, millions wasted, and jobs lost.

As a friend mentioned, his Fortune 100 company has his team wear jeans and play foozball once a week in order to encourage creativity… I may be too quick to dismiss this behavior. Sometimes good things come just from getting people together in a low-stress environment. But the good things that come may be haphazard and not replicable.

What has irreversibly changed

In the last 20 years since the dotcom bubble, the cost to build a startup or new business has fallen by 100x – 1,000x due to fewer upfront costs, commodification of common processes, and better ways to test concepts. Speed to build has increased by 10x – 100x due to commodification of common tech components and processes and new programming languages. While it’s always going to be difficult to build a big business, getting the small ones started is easier. These changes are here to stay. But in spite of the above changes, if startups themselves don’t work out most of the time, why should we expect corporates to be significantly better?

If we take a process-driven approach, there are new ways that larger companies can test new business concepts and opportunities at scale.

Improve your odds of being right. Reduce your cost of being wrong.

As with the decreased cost and time to build mentioned above, the cost and time to gather useful insights from potential customers has also fallen. Over the past 15 years we’ve come to understand and apply customer discovery interviews as a way to learn. What’s changed beyond customer input in more recent years is the ability to test in rapid iteration, for example by going off-brand and testing multiple product concepts through targeted online ads. The speed of data collection and ability to make modifications is orders of magnitude beyond what earlier versions of this method could achieve years ago. This is rapid experimentation. Start with customer insight, generate five concepts, expand those into 50, test those with ads or low-fidelity MVPs, look for early metrics of success, double down on the more promising ones, and only then short-list the most promising handful for prototyping. Depending on complexity, this rapid experimentation process could take as little as three to six months and be significantly less expensive than betting all on a pre-selected small number of concepts from the start. More shots at goal. Lower cost.

You never eliminate risk, but what is risk after all? Risk is the cost of being wrong. If you are able to lower your downside by decreasing your upfront investment and also increase the odds that you make the right bets, then you have decreased your risk. If you take a portfolio approach to testing concepts and move quickly, you can also improve your odds of building something big.

After my experiences working at Fortune 100 companies, and more recently working with hundreds of startups, I want to bring what I’ve learned to larger organizations. Let me know when you’d like to talk.

You’ve probably heard the term “critical mass” tossed around when discussing startups (especially ones like social media products). The term is usually either used in the form of “we reached critical mass and then things just took off” or “we don’t have critical mass and we can’t grow…” But what is critical mass? How do you calculate it? And what is therefore the number of users that these products need in order to have it? Critical mass is different from “product/market fit,” another loosely defined term, in that there is a number we seek, even if no one admits to knowing what it is. With product/market fit there is no minimum number — your business just ends up growing faster than you can handle.

It’s also good to think of what critical mass does not mean. It doesn’t mean that your business is cashflow positive, that you have good margins, or that you have lots of users. It’s something else — the point at which you have enough users or usage to keep sustain the product.

How did we get here?

The term critical mass does not originally come from social products or startups. It comes from physicists studying nuclear fission. Basic introductory definition: “A critical mass is the smallest amount of fissile material needed for a sustained nuclear chain reaction.” Read at least the Wikipedia article and try out the equations.

Many things impact critical mass in nuclear fission. From Wikipedia again, the following can change the point of criticality: Varying the amount of fuel, Changing the shape, Changing the temperature, Varying the density of the mass, Use of a neutron reflector, and Use of a tamper.

In other words, in fission, the critical mass is not static. It changes depending on those other factors. Here’s the nuclear fission critical mass formula.

Where M is the total mass, m is the nuclear mass, s is 1 + the number of scattering events per fission event, ẟ is the total interaction cross section, ⍴ is density, and f actually is a fudge factor (their words, not mine) that accounts for geometrical and other effects. If the value is 1 or greater, the fissile material with go into a critical mass chain reaction.

Once I saw that fudge factor, I stopped feeling so bad about only trying to come within an order of magnitude when it came to critical mass for a social product.

The above ways to change the point of criticality of fissile material are each pretty instructive too.

• Varying the amount of fuel. The most direct way to think about a minimum amount of stuff needed to attain the critical mass chain reaction.
• Changing the shape. That is, a thin layer of fissile material will not reach critical mass, where the same amount rolled into a ball may achieve it.
• Changing the temperature. This is mostly related to expansion.
• Varying the density of the mass. Less dense: more needed to become critical, if it can become critical at all.
• Use of a neutron reflector. Sending escaped neutrons back at the fissile material to get another chance at causing a chain reaction.
• Use of a tamper, another way to send escaping neutrons back at the core.

You might even start to see some similarities in how we could look at critical mass in a social product.

But of course, the above equation isn’t completely relevant for social products. Humans are different from atoms. It’s tough to make a critical mass calculator without thinking through those differences. But it is a starting place. But when a term doesn’t have a clear definition people start to apply it loosely, so I’m going to rewrite the critical mass formula, with human users in mind rather than molecules.

First, how do we define critical mass for a social product without relying on “you know it when you have it, because then everything goes well” type of circular reasoning? My critical mass definition is the minimum number of users you need to sustain user value in the product with no other product changes. This is product critical mass, not business critical mass. Critical mass in business (when your customers make the business sustainable) will almost always be much higher.

Even if we disagree on how to calculate critical mass, we can say that we don’t have it if a product becomes less valuable in aggregate for its users over time when there is no other change to the product itself. (There are many products that barely changed over years but where value remained, Twitter being a modern example.) So, not having critical mass should result in less action per user over time, even if the total number of active users stays the same. Having critical mass should mean that we maintain our users and most likely grow them — churn will be low and more people are likely to want to join.

What is the point at which user behavior means that the product sustains itself? That is, factoring in how much people use the product as it makes sense to your business (like messages sent, interactions, referrals), plus users lost through churn.

Estimating critical mass for a social product

The purpose is to answer these questions:

1. Do we have critical mass at the current number of users? If not and if all else about the product and usage remains the same, what number of users needs to be added?
2. Do we have critical mass at the current amount of usage? If not, what do we need to change about usage in order to have critical mass at the current number of users?

For now, my inputs are these:

• Percent of users active in time period. We can imagine two different products, one with a high percentage of active users (could be because of recency or other factors that keep people engaged) the other with a low percentage of active users (low need to log in). A higher percentage of active users can sustain a smaller overall population of users. So critical mass would be lower if active user percentage is higher.
• Necessity and expectation of frequency. Some social products (like a microblogging service) only make sense when there are frequent new contributions from many users. Messaging apps or SMS, can have critical mass with smaller engaged populations (the extreme being just two people to send messages back and forth). Other services have much lower, or different, requirements (like a time-based app where users aggregate at specified times of day). Overall, a larger critical mass will be needed when a population is spread around every time zone. A smaller critical mass will be needed when a population is concentrated around a smaller number of time zones. That is, if there is a natural busy hour for activity shared by the user population, fewer people will be needed to attain critical mass.
• Activity per user. Do users browse the content of others or do they also contribute?
• User activity needed to cause a reaction from another user. What does it take and how do you measure and encourage it?
• Percent of users who take an action in time period. Features such as the “like” makes it easy to take an action with a single tap.
• Impact of messages received on messages sent percentage. This is similar to the probability that a fission event creates other fission events.
• Other users’ impact on retention or churn. Likelihood to remain active if user receives action from another in time period
• Amount of value from single-player mode or other way to get value without user interaction. This enables the product to attain critical mass with a smaller number of users.
• Unaccountables. Time, place, UI / UX, subpopulation peculiarities, and other issues that act unpredictably. If the physicists used a fudge factor, I will too. This is also why I’m happy estimating this within a factor of 10x for now.

I’m not going to write a formula now, because I’m still thinking through the variables and am figuring out how to weight them. What I’d really like is some actual data to test against. Want to contribute? Please get in touch here.

But in general, I’m expecting something like these variations:

• Messaging app like texting: two people minimum in the extreme example. This is the one that will cause the most disbelief. Remember, we’re not looking for what sustains the business financially. Depending on your revenue model, those numbers can be dramatically different anyway. Instead, we’re talking about what sustains the product. In this extreme case, just two people could sustain usage and get value from the product and then go on to invite others in.
• Messaging app like Slack: 10x the above for one group. That’s enough for value from the small crowd, info from the group leader, plus enough people you want to talk directly to for one-to-one messages.
• Broadcast / microblogging service like Twitter: 1,000x the above. That’s enough for value from the crowd plus enough people you want to message directly.

The above are the best case scenarios (lowest critical mass numbers), given the variables. In most cases, where the product still needs work, the numbers will be much higher.

But I’m making the assumption that there is some predictability here that leads to better understanding users and systems. I might be wrong and there typically is not enough that can be predicted. But if we can improve our guesses, then talking about critical mass will shift away from being a loose concept and could carry more value as something at quantifiable, at least within bounds. I’d love to look at actual data of social products at their critical mass tipping points. If you have anonymized data you’d like to share, please contact me so we can refine the theory alongside actual results.

How can you reach critical mass for your product?

You can’t get there in one step, so here are a series of steps that help:

• Increase retention. Critical mass requires high retention for different reasons than typical, such as the increased costs of acquiring new users. Focusing on critical mass, you want retention to be high because repeated usage by the same users is better than occasional usage by different groups of users.
• Increase engagement. In the social media examples above, fewer users that are engaged can make up for more users that are not engaged. Because some products depend on the relationships between users, a small group of highly engaged users produces an effect that is preferable to a large number of unengaged users.
• Understanding the business you’re in. In marketplace services, a balance between supply (for example sellers on ecommerce sites, or drivers on rideshare) and demand (buyers on ecommerce or passengers on rideshare) is so important that these services are often careful how they launch. A marketplace startup will have an easier time focusing on one product category or one geography rather than launching with everything, everywhere.

I just finished teaching my tenth college course over three and a half school years. Across those 10 courses, I’ve had approximately 350 students in my classes.

During and after every semester I try new formats and reflect on what I should change, based on what’s working and who happens to be in each class. After 10 courses, it’s time to more seriously reflect on common areas of trouble for students. I make the claim that if these problems could be removed, then everyone would learn more quickly and more overall.

Don’t mistake this for a complaint. I’m presenting what I observed over these 10 groups of students. And don’t mistake this for a indictment of millennials. Overall, I’m impressed with this generation. Instead, let’s look at the point at which I see them – as students in a top university before they have started to work.

Lack of math skills.

This was the biggest surprise for me. Around half of the typical class population is not comfortable with math – and here I mean business math. Nothing sophisticated, no calculus or number theory. Just comfort with using addition, subtraction, multiplication, division, exponents, and percentages. Ability to do some of that in your head.

Ability to memorize formulas and apply them in other situations. This subcategory of the above turned out to be a huge problem. When I was a new professor I spoke to a colleague who had just given a pop quiz on topics that the class had discussed for a month. Everyone in the class failed. My colleague must be a terrible teacher, I silently concluded. I decided to try the same thing and handed out a pop quiz based on topics the students had confidently discussed for a month. Class average: 48%. I was shocked, but showed the answer key, went through the questions again, and told them to expect another pop quiz sometime. One month later I gave the exact same quiz. Class average: 70%. That result led to one of the biggest changes I made in university teaching. I call it “tricking people into learning.”

Estimated time for a student to improve to adequate level: 20 hours.

Belief that they don’t need to memorize and internalize formulas and definitions.

Without the memorization, you often fool yourself into believing that you understand and can apply the knowledge in different situations. This is one area that should have the biggest ROI on student time.

Estimated time for a student to improve to adequate level: 2 hours.

No knowledge of Excel.

If there is one tool that you’ll use after graduation, it’s Excel. For the record, I love Excel. Sure, there are more powerful programs that do similar things, but for accessibility and universality, Excel still wins. One of the good random things that happened to me was for me to find an Excel manual during an early management consulting project that was going slowly. I spent a good part of each day reading it and trying out formulas. Finding that book easily saved me hundreds of hours over the years. Now, while I do teach within a business school, perhaps half of my students come from other areas of the university – engineering, communications, cinema, arts and sciences. The lack of Excel knowledge, and the pain with which they learn it, are just too extreme. Even the ones who have taken accounting classes, a student told me, do their accounting spreadsheets by hand. I do not believe that that hand work is intentionally there to guide them into better understanding of business accounting.

Estimated time for a student to improve to adequate level: 10 hours.

Ability to stay focused in class without devices.

I have tried every combination of device usage in my classes including no usage, total usage, usage only when we read a case study and on and on. Hands down, zero usage wins. Note that I teach undergrads and this finding may not hold for grad students. I also teach an elective that is more quantitative and based on using online tools. In that class I deal with laptop usage for class. But in my other class, which is a required class for the entrepreneurship minor, I have banned devices and laptops once and for all. After doing so, students become better at thinking on their feet rather than simply looking up a response, they have better discussions with other classmates, and they don’t need to avoid online distractions because they’re not online.

Estimated time for a student to improve to adequate level: n/a.

Ability to communicate well.

In general, the students are pretty good at standing in front of the class and presenting their projects. This is something I only had to do once in my time as an undergrad. For today’s students, in-class presentations are pretty common and their confidence shows. The area they could improve, or eliminate problems from, is in written communication.

The first area is email. I receive a good number of emails which start with the word “Yo,” or “Hey,” or which are tough to read because, I have to hope, lack of effort rather than lack of ability.

The other area is writing longer papers. Lack of proofreading, spelling mistakes, using voice-to-text (it shows) make otherwise smart students look pretty bad. The proofreading and spelling mistakes are easy to fix with software. Being able to write well though… Commit to reading lots of good books and articles and writing thoughtfully over years.

Estimated time for a student to improve to adequate level: 1,000 hours.

Other Tangents.

The most common question students bring up when we go off on larger issue tangents. Is a university degree worth the high cost of tuition? I reflect that I could take one half of the number of students I currently teach, charge them half as much a credit as the university does, and provide a better learning experience while also earning more. This is a reality that the American university will be forced to deal with in the next generation. It’s going to be fascinating as the change happens. What will replace the university network though I don’t know.

I am not a university academic. Becoming a professor was entirely accidental for me – I never applied for the role, but because of previous work experience was asked to teach. I’m glad I did. Teaching has changed the way I express ideas and think about their impact.

New changes. Since there has been so much similarity in the way different classes of students struggle with my classes, next semester I will distribute a list of likely problem areas. Avoid these and you will improve your chances dramatically. History suggests only a 20% class body follow-through on the offer, but I’ll take it.

In light of Mark Zuckerberg’s Congressional testimony (and Facebook’s own natural monopoly) I reflected on some of the monopolies I’ve known. Given what Facebook does I focused on communications companies, which historically were somewhat close to what social networks of today provide.

AT&T’s government granted monopoly

We trace this back over 100 years to the Kingsbury Commitment of 1913. That agreement set out the rules for AT&T’s place as the US’s national carrier and how it needed to work with local service providers. In the early days of communications technology build-out, national governments often stepped in to guide (or control) what telecommunications companies were able to provide which services and for what benefits of the population.

AT&T however was known as a “gold-plated” monopoly. Phone service in the US extended to rural areas that were more expensive to build out (not a condition seen in every other country) and the degree of innovation coming out of the company was dramatic. Some of their innovation portfolio includes dial tone, mobile telephony (car phones originally), the transistor, Unix… This gold-plated aspect led to some interesting experiments in employee training, including a foray into teaching the humanities (see The Organization Man Goes to College: AT&T’s Experiment in Humanistic Education). This investment in employees is comparable to Peter Thiel’s argument that companies like Google are also monopolies — and that’s why they take care of their employees.

Telecommunications hardware used to be much more locked down than today. It used to be illegal, then uncommon, to own a landline phone in the US. AT&T rented them. The monopoly granted AT&T allowed the company to regulate any devices that connected to their network — a term that was interpreted loosely. Legal action by AT&T to enforce its monopoly dominance included preventing a company called the Hush-a-Phone (basically a cup that fit on top of your telephone handset in order to shield the speaker’s voice) from selling its products. Why go to such lengths to prevent this physical product from existing? If you have near total control, any little step away from that is problematic.

Interestingly, Peter Drucker argued against AT&T’s monopoly breakup in 1984, but on behalf of national security. The national security and telecommunications infrastructure and equipment issue is still a concern today (actually part of an ongoing concern for at least the last 20 years) with Chinese equipment provider Huawei.

The swing of general communications technology moved from government enforced monopoly to decentralized.

China Telecom

China Telecom has a much shorter history than AT&T and being based in a different country with different needs and at a different time, competition to this monopoly came in different forms. An early competitor was China Unicom, which in the mid-1990s had low single-digit market share overall and market penetration in just a few urban areas. The company was dependent on China Telecom for national interconnections (which sometimes was cut). Starting in 1998, China (as did other parts of the world) started to allow more companies to offer telecommunications services, initially through (lower quality) Voice Over IP  (VOIP) calls. These early calls were not like the Skype you later came to know. Rather, the caller would dial the number of a local gateway and then their destination phone number (basically operating like a calling card except that the voice traffic was routed at least partly on an IP line). Later enterprise implementations were seamless and did not require the caller to dial multiple phone numbers. I remember various Chinese government agencies during this time exploring VOIP network rollout, some with official licenses, some because the founder’s relative was so-and-so and they were operating in the grey market (many long stories there).

Impact to the consumer — the official and later market determined price of a phone call dropped significantly. Back in 1997, the official rate to call from China to the US (something very few people ever did) was US$1.30/minute. During that time I knew someone who was arrested for offering cheaper non-official services. By 2000, the rate had fallen to around $0.10/minute. It’s been basically free for years now.

In China, total subscribers (both fixed and mobile) was 7 million in 1990. The number of just mobile subscribers in China is now 1.6 billion. Yes, that’s more than the total number of people in China. More and more countries have mobile phone penetration of greater than one per capita. Oh, and Unicom later went on to build an enormous VOIP and mobile network.

Other countries have dealt with communications technology in different ways. In Myanmar (then Burma) people went to jail in the 1990s for illegal possession of a fax machine or a modem (the Internet was only made legal there in 2003). These actions (as elsewhere around the world) were more based on protection government communications control than consumer access.

The above were state granted and regulated monopolies. However, we could also think about how else the initial national communications infrastructure might have been built out around the world.

Now back to Facebook. One of the interesting exchanges from the Senate hearings was this one between Zuckerberg and Senator Graham:

Graham: Who is your biggest competitor?

Zuckerberg: Senator, we have a lot of competitors.

Graham: Who’s your biggest?

Zuckerberg: I think the categories — did you want just one? I’m not sure I can give one. But can I give a bunch?

Graham: Mmhmm.

Zuckerberg: There are three categories that I would focus on. One are the other tech platforms, so Google, Apple, Amazon, Microsoft; we overlap with them in different ways.

Graham: Do they provide the same service you provide?

Zuckerberg: In different ways, different parts of it, yes.

Graham: Let me put it this way: If I buy a Ford and it doesn’t work well and I don’t like it, I can buy a Chevy. If I’m upset with Facebook, what’s the equivalent product that I can go sign up for?

Zuckerberg: Well, the second category that I was going to talk about —

Graham: I’m not talking about categories. What I’m talking is the real competition you face.

Other articles don’t focus on this point, looking at social networking businesses across different markets and usages. If you think of just the casual, social pieces of peoples’ lives, what is the biggest social networking competitor to Facebook?

Sources of Monopoly Power

The list of sources of monopoly power in which Facebook is strong includes: no substitute goods, network externalities, and increasingly technological superiority and manipulation.

Facebook has an advertising revenue model. Therefore, anything that interferes with the company from earning ad revenue from their user generated content is likely to be seen as a threat. In 2008, Power Ventures, a social network of social networks was sued by Facebook on multiple counts (another interesting take is here). Part of the case is still pending. What did Power Ventures do to draw Facebook’s attention? They enabled users of Facebook and many other social networks to see all of their feeds in one place — and without the original social network generating, or being credited for, the advertising. Good for the users (convenience) but bad for the ad-based companies.

In its 2017 annual report, Facebook shows 184 M daily active users in US and Canada — the smallest number by any of their regions, which are all more populous. By revenue, however, it’s the reverse. The US and Canada, having more developed and profitable advertising markets and what seems to be more frequent usage, generate the highest per user revenue (over $84/user in the US and Canada for the 2017 annual report’s most recent four quarters, p 38). Facebook’s social networking ad revenue market share approaches 80% in the US.

I won’t go into the details on the Cambridge Analytica scandal, with which anyone reading this far probably already has background. If it weren’t for Cambridge Analytica, when would Zuckerberg have testified to Congress? It would have come eventually.