Categories
Marketing Education

Innovations Diffuse, but People Adopt Behaviors: How to Use Misinterpretations of Diffusion of Innovations to Your Advantage.

Playing “Oregon Trail” as a kid made me grateful for cross-country transportation innovations.

#

Everett Rogers was not an engineer. That might seem odd to most technologists. Since 1962, they’ve used his groundbreaking idea – The Diffusion of Innovations – to track the introduction and spread of revolutionary technologies through a population. Iconic examples of the so-called S-curve permeate techno-lore: Refrigerators, color televisions, microwave ovens, dishwashers, personal computers, mobile phones, and the Internet itself. Individual entrepreneurs use it to justify hope in the inevitability of all innovation and the eventual success of their idea, product, or service. They’re reading it incorrectly. Rogers was a professor of Communication Studies – exploring how people accept change in societies. He never claimed all innovation would succeed. In fact, innovation only seems inevitable in retrospect. Many innovations never achieve 100 percent of their potential market. But more to the point, technologists are not paying attention to a simple point of language: Innovation is not the same as a single idea, product, or service. The innovation may succeed. You may not. Therefore, while the core ideas behind the Diffusion of Innovations may be helpful for innovators in general, they offer little value to help ensure the adoption of your specific product.

Let’s use Diffusion of Innovations as a starting point to something more useful for individual entrepreneurs.

 

Introduction

The Diffusion of Innovations is a representation of the natural process of the spread of innovations throughout a population. The word “natural” is intentional here – the resulting diffusion results in a so-called normal distribution over time – aka, a bell curve. Rogers divided this bell curve into five segments. The largest two groups (the “early” and “late” majorities) are each one standard deviation from the center and represent about two-thirds of the entire population. On either “tail” are innovators and early adopters on the left and laggards on the right. The critical point comes from Rogers’ background as a communications scholar: People use different decision-making processes at different stages of adoption. Early Adopters are more apt to seek out relative advantages and have a lower threshold to trying something new. Laggards, much less so.

Diffusion of Innovations

You can look at the diffusion process in two ways – the first as a bell curve distribution, and the second as a cumulative measure. The first view defines the behavioral categories. The second view creates the S-curve.

This idea has been around so long because it seems to match the world around us. While the time scale of diffusion may change for any particular technology (making the curve shallower or steeper), the “S-curve” we see on charts that track these data seem remarkable consistent.

But not always.

Technology Adoption Evidence

Our World In Data is a fun tool to use. When you do, it seems like it confirms Rogers’ theory – we see plenty of what look like S-curves. We also see an acceleration in technology adoption in recent years (steeper curves). But look more closely. The actual data tell a more complex story the Diffusion of Innovations would suggest.

A funny that happens when you begin to test your assumptions with actual evidence. While the data collected here are incomplete, we notice a few critical features:

Feature #1: Diffusion doesn’t always follow a smooth line.

Sometimes, due to non-market forces (World War II comes to mind), diffusion can slow or stop altogether. Those seeking to introduce an innovation would be wise to mind the non-market as those factors may dramatically change the path and the rate of diffusion.

Feature #2: Not all innovations reach 100% of a population. In fact, most do not.

These charts track technology adoption, but any “innovation” will do. Sometimes, the innovation is a product that some people will never afford. Sometimes, that innovation conflicts too strongly with a cultural norm. Sometimes, that innovation doesn’t catch on for no good reason (the Linux desktop is one of many examples). The bottom line is that defining the “population” is a critical step in understanding the shape of the diffusion curve. Only in rare cases is that number “100%” of a population.

Feature #3: Some innovations are overtaken (prematurely) during their diffusion cycle.

Innovations such as eBook readers and cell phones (both overtaken by smartphones) are good examples. As the pace of change in many areas accelerates, new products, services, or ideas may stunt the growth curves of others. What’s more, all charts like these fall victim to survivorship bias – in other words, we only see the winners. The chart would look very different with the losers included. Innovators would be wise to watch for disruptive forces outside their direct competitive set.

Innovators use Rogers’ work, as well as extensions of his ideas such as Moore’s 1991 book Crossing the Chasm, to help them choose promising avenues for investment, produce reliable sales / voter / opinion forecasts, and (most importantly) choose communication strategies most likely to work at different stages of the diffusion cycle.

But there is an important distinction to make here: An innovation is not a single product. That may seem obvious on its face, but many leaders fail to make the logical leap. To put it more simply: Buying your first “microwave” is not the same as buying a “Whirlpool brand microwave”. The Diffusion of Innovations theory may provide some guidance on features and messages most likely to resonate at a given point in time, but the theory provides little in the way of actionable advice to a specific entrepreneur within a broader innovation category.

Let’s do that now.

We’ll use Diffusion of Innovations as a starting point to help provide actionable guidance for a manufacturer of a “Keto” snack bar product line. In other words, we’ll start with the general advice of Rogers’ theory and work down to specific advice for an individual brand. Why do we need to do that? Diffusion of Innovations helps explain what happens, but not why it happens. Like many great theories, we understand the results long before we understand the underlying mechanisms. Sixty years later, that’s finally changing.

 

Stop using the words diffusion of innovations. Start using the words adoption of behaviors.

How we use words can shift their meaning, and by extension, how we act on them. This is especially true with theories that have entered the general public consciousness. Diffusion of Innovations is no exception. Let’s have a look at each word, what it means precisely, and what words we might choose instead to help individual innovators.

Diffuse vs Adopt

The word “diffusion” and its root “diffuse” weren’t communication terms originally – they were physics terms. To diffuse is a natural process so long as the correct conditions are present. For example, high density gasses will diffuse (spread out) into a fixed volume of low-density gas as soon as a barrier between them is removed. Diffusion (at least in physics) is inevitable.

Diffusion of Innovations, in contrast to the gasses example above, are not inevitable. Many so-called innovations fail to “catch on” – especially in the light of what the word “innovation” truly means (more on that in a moment). By using the word diffusion, we are making an unconscious assumption that our product, service or idea has a certainty to it – we simply need to remove the “barrier” between our innovation and the available market. That assumption is a fatal error for entrepreneurs.

A better word to use is adoption.

Instead of an inevitable process, adoption is a conscious choice by individual people. Adoption is by no means inevitable. Yes, the Diffusion of Innovation theory provides general guidance for the mindset of groups of people at different lifecycle stages, but that doesn’t tell us much about how to change the behavior just one person. We’ll get to a better idea for that process later. At this point, it’s better to recognize that the process of adoption takes active work on your part as the entrepreneur.

But even more problematic is our general understanding of the word innovation.

Innovation vs Behavior

This is a very simple definition of innovation, but it is a powerful one. Notice something about it? The “product” (aka “technology”) is third in the list. Processes and modes of thinking also are innovations. In other words, the idea of representative democracy is just as valid an “innovation” as the technology of the microwave oven.

Many technologists conflate the words innovation and technology to mean the same thing. But technology is a sub-item within innovation. It may seem like a philosophical point, but it’s not: Innovations cannot exist without a human acting on them. To put it another way and paraphrase the well-known saying: If an innovation falls in the forest, it simply doesn’t exist.

Therefore, innovation, at its heart, is behavior.

That’s especially important when we begin to combine the words “adoption” and “behavior”. Innovations do not diffuse. It only appears that way on the outside looking in. Individuals adopt a new behavior. The first is inevitable. The second is not.

Think of our example of the Keto snack bar. The snack bar was an innovation at one point in the past – but not so much today. That behavior (eating a pre-packaged snack bar) has reached nearly 100% of its target population. No, the innovation here is a process and idea behavior. The process is the Ketogenic diet, and the idea is the concept of protein balance in human macronutrient consumption. There may be some technology in creating a shelf-stable “Keto” snack bar, but that technology isn’t critical to the behavior change process (although it may be a critical part of the process of whether you could produce the product at all).

Let’s use some data to help us illustrate the point:

Keto popularity

We’re using our standard comparison formula, benchmarking “Ketogenic diets” against terms that help us highlight changes. Over the past three years, Keto (the blue line) is moving up steadily while the benchmark items remain flat or cyclical.

For our Keto snack bar entrepreneur, the true behavior to adopt is the Ketogenic diet. Given that, what conclusions could our innovator draw?

  • Keto’s so-called “diffusion” won’t help any particular product other than providing a larger market. In other words, just because Keto is “hot” doesn’t mean that people will buy a Keto snack bar. We can see that in the chart – the top-ranked Google Keto product company fails to get any “boost” from general interest in Ketogenic diets.
  • The innovation of Keto (as a scientific concept) isn’t as important as the change in behavior. Just because people are talking about Keto does not mean they have changed behavior (in fact, many of those who talk about it the most don’t actually follow it – or even know quite what it is). Our entrepreneur should focus on finding evidence of behavior change. Search behavior is one way to do that, but diet survey data is better.
  • To that point, what if people are interested, but have not yet changed behavior? Shouldn’t our Keto snack bar entrepreneur work to educate the market on Keto to help boost its own market? No. Not if they don’t want to run out of money. The “education” process is slow and thankless. And per point #1, just because people follow the Keto diet, doesn’t mean they will purchase your product. Our entrepreneur should watch the market, follow the adoption path, and focus on linking the prior behavior before the Keto diet (snack bar consumption) with the new behavior change.

Words matter. Diffusion of Innovations happens at the population level. Change of behaviors happens at the individual level. You always sell one to one.

 

Stop planning to address a market opportunity. Start making the most of feedback loops.

One of the most persistent criticisms of Rogers’ work is that it assumes a sender-receiver model of communication. A technology diffuses into a population, the population does not diffuse into the technology. To put it more simply, the people buying microwave ovens didn’t change the microwave oven as a technology.

That’s ridiculous, of course, and even more so now with software-driven products. The true innovation process is iterative and dynamic, with the innovation itself morphing as it is adopted by larger groups of people. Much of the bias in Rogers’ thought process comes from the dominant view of the communications process at the time – a mechanistic view of “sender-channel-receiver” born of telephone communications in the 1940s and 1950s.

This process isn’t limited by the type of innovation, but it is most visible in “idea” or “policy” innovations. Family leave policies at modern corporations are a clear example. The early innovation was guaranteed family leave for mothers for a certain amount of time. That was extended early on to adoptive mothers. Then to fathers (albeit for a shorter period of time). Then to fathers for a longer period of time. Then to same-sex couples. And so on. As the original innovation “diffused” within the corporate population (restated: as individual corporations adopted the policies), and people used them, the original innovation changed both in scope and intent.

Behavior patterns also change as technologies as they are adopted: Smartphones are an obvious example.

Technology Evolution Smartphones

Innovations evolve over time based on user behavior. In the case of smartphones, their evolution followed other innovations and supplanted (or stalled) others.

The early behavior “making a call outside of a fixed location” morphed into “use a handheld and mobile internet access device” not only because of technology improvements (of which there were many), and because of price reductions (giving more people access), but most importantly because of behavior changes from people adopting cell phones in greater numbers. It was the true genius of innovators such as Steve Jobs to recognize that the technology didn’t drive behavior, behavior drove technology. The faster the innovator can respond and implement adopter feedback, the more successful that innovator will be in the marketplace.

How do multidirectional feedback loops help our Keto snack bar entrepreneur?

The basic science of the Ketogenic diet evolves (as all science does) as more people adopt the behavior change (the Ketogenic diet) and researchers learn more about the biological processes behind it. That translates specifically into the quantity and proportion of macronutrients you should eat, and at what times of the day they are needed. As you could imagine, how much to eat and when to eat it would have a material impact on our Keto snack bar product.

An important reminder: This isn’t the behavior change process difference of an “early adopter” versus a “laggard” in Rogers’ framework. This is a feedback loop within the innovation itself. Specifically, the Keto snack bar entrepreneur must be ready to adjust their product in four unique ways:

  1. Formulations: As the science evolves, the basic nutritional requirements may (will) change, and therefore, the product itself may need to change. This is more than a “brand extension” strategy (e.g. new flavors and package configurations).
  2. Key messaging: Again, this is different than “early adopter” versus “laggard” messaging strategies, as language around “Keto” changes, packaging and marketing language also will need to evolve. Consumers (whatever their “lifecycle stage”) are likely to be confused by changes in dietary advice.
  3. Buying channels: Only specialty retailers carried “Keto” products five years ago – now, nearly all of them do. Selling products in mass retail channels versus specialty retail channels all present different challenges. Generally, new channels open over the course of the adoption curve as larger retailers see the opportunity, but with “long tail” channels such as Amazon, large retail can begin selling products very early in their adoption process.
  4. Pricing models: Subscription models are a parallel innovation that has changed the way supplements and diet products are sold. Staying true to any diet (Keto included) is easier with consistent availability and delivery of food options. A subscription for snack bars can help our entrepreneur prevent people from “abandoning” the behavior.

Doesn’t that sound like 4Ps (Product, Price, Place, and Promotion) of the marketing mix? It should. It is the core principals of Marketing 101 that are better guidance for our entrepreneur than the Diffusion of Innovations framework. Entrepreneurs must maintain flexibility in their marketing mix to respond to innovation/adoptee feedback cycles, many of which can happen with blinding speed and unpredictability.

Following an innovation is like weather in Minnesota. If you don’t like it, wait a few minutes. It’ll change.

 

Stop equating innovation with a single technology. Start accepting that multiple behaviors compete to solve the same underlying need.

To this point, even though we’ve used different words (behavior in place of innovation and adoption in place of diffusion), we’re still making the assumption that people adopt one behavior to satisfy one need.

  • How do you get around? Adopt the automobile.
  • How do you keep food cold and fresh? Adopt the refrigerator.
  • How do you stay connected on the go? Adopt the smartphone.

One need. One solution.

But our language shift has another benefit. When we stop thinking about technology and start thinking about behaviors, we open our mind to the flaw in the one-to-one logic: There is more than one way to address a need. That may seem obvious on the surface, but it gets a bit trickier unless we use an example. The “automobile” is as good as any, but you could use any technology, product, service, or idea. Simply think about the need being addressed rather than the innovation solution. When you do, the picture gets more complex:

Technology Behaviors - Transportation

A funny thing happens when you look at the need and not the technology. People adopt multiple behaviors to solve the same problem. (In this chart, we’re ignoring “walking” which was the dominant mode of transportation until humans domesticated the horse.)

Charting the diffusion of a single innovation is ridiculous. The automobile is just one mode of transportation for a small set of the total number of transportation situations. The aircraft is the fastest way to cross a large country, but perhaps not the most fun. By contrast, a scooter or bike might be the slowest way to do the same job, but it might be much more enjoyable (for the right person). Harvard researcher Clay Christensen puts it even more simply: Customers hire products to do jobs for them. In this case, people are hiring the automobile to get them from place to place … but that’s not the only “product” they could “hire” to do the job. When you change the focus of the chart from one innovation to one underlying need, Christensen’s point becomes obvious.

In marketing, we would say that transportation has a broad competitive frame, but nearly all underlying needs have a variety of ways you could address them. At a high level, we can simplify those options into three major categories:

  1. One option can overtake another option. Horses are lovely animals, but they’re not a serious option for urban transportation. (Deep wilderness is another matter.)
  2. One option can coordinate with another option. People may use a car to get to the “park and ride” in the suburbs and a light rail train to get to work in the city.
  3. One option can coexist with another option. People can use any or all of the transportation options on our chart. Using one doesn’t mean we cannot use another one at a different time – even for the same purpose. We may ride our bike to the grocery store one day, and then take our car the next.

But do all innovations follow that pattern? Nearly every household in the United States owns a refrigerator. The only thing left to do now is build better and better refrigerators. Right?

Perhaps not.

The diffusion of innovation is really just an adoption of a behavior to satisfy a need – in this case, keeping food fresher for longer. But what if there were a different way to do that? We’re not talking about food replicators from Star Trek or anything so 24th century as that. What about Uber Eats? Do we still need the original innovation (the refrigerator) if a new innovation presents itself that’s more useful in some way? What’s fresher than “at your door in 15 minutes?” It may seem like a silly premise: People will keep refrigerators and adopt real-time food delivery. Who is going to stop having a refrigerator in their house?

Consider this: By some estimates, Americans waste over 30-60% of everything they put in their refrigerators. Think nothing of the immorality of wasting food when so many people live with food insecurity, that level of waste costs the average household over $1,600 each year. Is the ability to store food for a longer time really a useful innovation? What if people started thinking that real-time delivery was a more useful behavior than storing cold food in a huge box in their kitchen (the largest single appliance in most homes), and perhaps do something else with that space … and save $1,600 every year to boot?

We shouldn’t need to do the same mental exercise for smartphones. The pattern is self-evident.

Let’s return to our Keto snack bar company. Understanding their challenge in a broader sense of behaviors and needs provides better guidance than “following the wave” of diffusion. Ketogenic diets compete with a variety of other diets and trends to solve the same underlying needs: weight loss, fitness, health, and wellness. Atkins, vegetarianism, veganism, organic, local, raw, calorie counting, and countless other legitimate (and illegitimate) diets all attempt to address the same need. In Christensen’s language, people are hiring the diet to help them lose weight. If the diet fails to deliver, they’ll fire that diet and hire another one. Or they mail keep one diet and hire aspects of another one. The needs are simple. The behaviors adopted to address those needs can be complex.

How should our snack bar entrepreneur think about that complexity? Some of the diets we mentioned coordinate well with Ketogenic diets (Atkins); some won’t make much of a difference if a person adopted aspects of both (local or raw), and some could conflict (calorie counting). As the Keto snack bar innovator, what can this competition of ideas tell us about product development strategies?

  • Coordinate: You may be able to convince Atkins followers to eat your snack bar with some subtle shifts in messaging.
  • Neutral: Convincing “raw food” adherents might be challenging depending on your snack bar formulation, so you may want to consider a new offering that uses only raw ingredients. Also, you could consider licensing production in high-density urban cores where “local food” adherents congregate.
  • Compete: Calorie counting doesn’t fit well with Ketogenic diets (calorie densities in fats). A possible solution could involve offering a smaller snack bar size, therefore reducing the calories per bar and increasing the chances a calorie counter might try it.

This isn’t a matter of trying to ride three horses, but rather it is a deliberate diversification strategy. Adoption of complex behaviors is not predictable by some bell curve – your intuition already tells you this. (How many diets do you see come and go? With the same person?) As an innovator, you would be wise to accept this reality and plan for multiple possible paths to the same end goal for the consumer.

 

Stop trying to predict adoption with lifecycle phases. Start using q.

The bell curve shown in the Diffusion of Innovations theory simply gives us a macro view to the adoption process. It is not showing us what’s happening at the person to person level. In fact, Rogers’ admitted the mechanism of adoption largely is a mystery.

It’s not.

There is an old saying in marketing: If word of mouth was perfectly efficient, there would be no need for advertising. The number one source for information about a new product or service, and the most reliable form of persuasion, and the biggest predictor of your behavior, is your social group. This truism is not the result of modern social media, although these networks give us a powerful and measurable tool.

We’ve all had this experience. You signed up for Netflix once a certain number of your friends, coworkers, or peers signed up. Just one person usually isn’t enough to trigger your behavior. But there is a certain number, we’ll call it q, that tends to tip the scales in your decision-making process. The research shows this adoption process isn’t necessarily tied to the “phases” in Diffusion of Innovations. Even if you would consider yourself a “laggard,” if enough people in your social circle adopt a new kitchen gadget (hot pots come to mind), you’re likely to do it too. Social pressure is that powerful.

Complex Contagion

On the left is a simplified representation of Centola and Macy’s original work on Complex Contagions. They argue that high-risk adoption behaviors (e.g. a new diet) requires confirmation from multiple sources in your social network – a “wide bridge” of support. Dean Eckles from MIT expands and revises these conclusions with theoretical models combining short and long social ties. The math isn’t as important as the conclusion: Humans are social monkeys, and social ties are the mechanism of adoption.

Put simply, this is the idea behind “virality” in the spread of ideas and behaviors. There’s plenty of math behind this (explore graph theory and complex contagions if you’re interested in learning more), but here are the basic ideas:

  • Adopting a behavior change (read: innovation) is costly to the early adopter, and less costly to someone who waits.
  • Behavior changes lack credibility until enough people in your social network adopt them.
  • Simple “awareness” of a behavior change option usually is not sufficient.
  • There is a difference between “uncontested” and “contested” behavior changes. Uncontested behavior changes don’t have an alternative other than the status quo – antilock brakes on cars are an example. Contested behavior changes have competition – diet changes are almost always contested by other diet options beyond the status quo.

The variable q simply represents the fraction of people in your social network who have adopted a new behavior. When q gets large enough, you are highly likely to switch as well. The threshold simply is lower for uncontested changes (q can be smaller). Centola and Macy’s work finally starts to address the mechanisms behind behavior change in a way that we can act on them as innovators.

That takes us back to our Keto snack bar example. Instead of using a “lifecycle stage” with its generic population percentages, it is more useful to think about the total addressable market (TAM) for the Keto snack bar as a complex, contested contagion where the q variable (the fraction of those people in a social network) must be high in order to “convert” a new buyer to adopt the Keto diet.

In this case, our Keto snack bar entrepreneur would be well advised to concentrate efforts on distinct social networks of existing Keto adherents, following the spread of converts in their social networks. In other words, instead of attempting to influence q, the entrepreneur simply is discovering concentrations of people and taking advantage of the progression of the contagion at work.

This is not the same as the oft-discussed tactical approach of “creating a video and hope it goes viral,” or an “influencer strategy” on Instagram – this is a hyper-focused strategy based on individual social networks. Later in the process of mass contagion (when the behavior spreads to enough “nodes” in the graph) the entrepreneur can pursue mass advertising channels and big-name influencers. But in the beginning, better (and more cost-effective) tactics include:

  • Targeted social network advertising based on defined interests and engagement history. Specifically, you want to target those people who have adopted the Keto diet, not simply those discussing it. You may find these people at Keto-focused fitness centers, nutritionist client bases, or the scientists themselves.
  • Once you find those people (nodes), focus on incentives for them to choose your snack bar. Those could include special offers, but they are more likely to include in-depth information about how this snack bar fits with their lifestyle choice.
  • The spread of the snack bar, then, will spread along with conversions to the Ketogenic diet from people in the original node’s social network. In other words, it’s a piggy-back strategy.

This flies in the face of “big advertising” and “mass awareness” campaigns. When you’re talking about behavior change (aka innovation), simple awareness is not enough. People change behavior because people around them change behavior. Our Keto snack bar entrepreneur will spend less on advertising, and they will spend more time mapping individual influence networks.

When you’re trying to innovate, individual people and their personal networks matter more than mass markets, so-called influencers, and brand awareness.

 

Conclusion

Rogers’ work on the Diffusion of Innovations, as well as follow-on work by dozens of other scholars and practitioners, has been groundbreaking work. It helped remove the spread of innovations from the realm of alchemy and into a testable social science. However, in the intervening years, we have learned much more about how to apply Rogers’ original work – like most great theories, its best value isn’t in its precise conclusions, but rather in the new questions it forces us to ask. It is to that end we make the following suggestions to individual entrepreneurs hoping to succeed with their own idea, product, or service:

  1. At the population level, innovations appear to diffuse. However, that masks the underlying reality. All diffusion is the result of an individual person adopting a new behavior in place of an old one. Concentrating on the individual person’s adoption process – and the mechanisms to influence it – will be more effective than generalized lifecycle phases.
  2. Innovations do not remain static as they diffuse into a given population. They evolve. They always evolve. Smart innovators must remain flexible, ideally designing their complete marketing mix (the “product” included) to follow those evolutionary paths.
  3. An innovation may never reach 100% of its target audience. In other words, contrary to popular examples, many innovations fail to fully diffuse. That’s largely due to other innovations competing for attention and adoption. The original innovation may be useful, but it will lose out to another more useful innovation – or it may be adopted alongside another innovation. Smart entrepreneurs will understand that usefulness is the key in satisfying the underlying problem or opportunity for that person and plan multiple winning strategies based on customer needs, not technologies.
  4. Lifecycle phases are poor predictors of behavior change. Specifically, the rate of change is not obvious from the bell curve. A better predictor is a person’s social network and how many of those people have adopted the change. Smart entrepreneurs will focus on surrounding target audiences and take advantage of the social pressure from those who already have adopted the change.

The aim here is not to discredit Diffusion of Innovations, but rather to build on it to help individual entrepreneurs see the difference between the success of an innovation at the population level – and their idea, product, or service at the individual level.

To give yourself the best chance, focus on the need first, the behavior second, and your offering third.

###

Categories
Marketing Education

Learn the true influencers for CRISPR using Brand24 (not Gartner)

Need more details? Check out my full article here.

Categories
Marketing Education

Using Google Trends to Build Your Own Hype Cycle (Today’s Example: CRISPR)

Need more details? Check out my full article here.

Categories
Marketing Education

Five Reasons You Should NOT Use ANY Gartner Hype Cycle

Need more details? Check out my full article here.

Categories
Marketing Education

Unhyping the Hype Cycle: Five Secrets to Building an Attention Dashboard for Any Innovation

Photo by Jongsun Lee on Unsplash. SQUIRREL!!! added by the author.

If we could only predict the next boom in cryptocurrency. Or when gene therapies might show up at your local Walgreens. Or when planes will fly themselves. Timing any of these correctly will make individuals rich and shareholders giddy. Timing them wrong will bankrupt investors and sink promising product launches. The most popular model for understanding technology trends such as these, and cashing in when the time is right, is the Gartner Hype Cycle. If ever there were applications where its usefulness should apply, it would be cryptocurrency, gene therapy, and self-flying planes. Unfortunately, the Gartner Hype Cycle can’t help us understand any of these trends. The only enlightenment it can offer is the simple truism that “hype exists.” Sorry Gartner, but PT Barnum proved that theory 100 years ago.

Let’s see how we can do better.

Introduction

Anyone who has sat through a corporate PowerPoint presentation or Investor Pitch Contest has seen the Gartner Hype Cycle. As testament to Gartner’s excellent marketing, most corporate followers can recite it by heart: After a trigger event, hype grows quickly, peaks, crashes just as quickly, and then slowly reaches a balanced equilibrium. It’s so reasonable and intuitive that most of us rarely question its basic premise.

Since 2009, Gartner has published a so-called “Hype Cycle” for all manner of markets, technologies, and trends. In fact, the model has become so popular that “analysts” now publish nearly 60 different hype cycle charts each year – from Supply Chains to Digital Marketing to Government Technology. Here’s the gotcha: When we apply even the most modest level of scrutiny to it, each assumption that underpins the Gartner Hype Cycle falls apart.

Great marketing. Bad science.

But what’s the alternative? In the early stages of any new technology, “hype” seems to be all there is to go on. Entrepreneurs need some mechanism to understand if a hot topic has potential, or if it’s just hot air. Investors need to know when to jump into the water, and when to sit on the beach.

What follows is the answer.

In five easy steps, we’ll dismantle the flawed logic of the Hype Cycle and replace it with a bespoke Attention Dashboard – a set of building blocks you can use to generate meaningful insights on your own. No fancy charts. No breathless anticipation of an analyst’s opinion. No massive checks written to consultants. Less branding and marketing spin. More data and useful conclusions.

To help illustrate the process, we’ll use an emerging technology clearly in its commercial infancy: clustered regularly interspaced short palindromic repeats, or CRISPR, for short.

Step 1: Stop using the word hype. Start using the word attention.

There is something about hype that seems both unseemly and unethical at the same time. Hype is disconnected for substance, truth, and reality. Hype is a deception. And the worst criticism of all: Hype is marketing.

Definition of Hype and Attention

Source: New Oxford American Dictionary, edited for clarity. Hype and Attention mean much the same thing, but hype is an emotional trigger on our lizard brain. Attention is a neutral (even positive) word, allowing us to engage our rational mind.

While not wholly false, when we use the word hype (with all its emotional baggage), we ignore its usefulness. Whether positive or negative, hype draws attention to a product, service, or idea, helping to break through the background noise of daily life and shine a spotlight on it – if ever so brief and fleeting that light may shine. Why are businesses, politicians, and advocacy groups so willing to risk the criticism that comes from hype? Because the biggest risk to achieving your goals isn’t being criticized, it’s being ignored.

Hype or Attention?

Let’s make a helpful edit to this Gartner Hype Cycle chart. If we’re interested in the promise of CRISPR, it’s more useful to think of it as “attention” rather than “hype.” Attention draws interest. Hype turns people off.

Gartner was smart not to brand its “Hype Cycle” as the “Attention Cycle” – the word hype carries more emotional energy, making you feel as though you are privy to forbidden knowledge of how the world works. But while it may be smart branding, and while it may drive business for consultants, using the word hype leads to poor decision-making. Most decent people want to avoid hype, but they’re just fine with attention – especially when they believe in what’s getting the attention.

As an entrepreneur or investor exploring commercial opportunities for CRISPR, the word “hype” clouds our thinking. Using that word implies that the promises of the technology won’t materialize – that CRISPR is doomed to the dustbin of promising, yet ultimately failed, therapies. But that’s an oversimplification. Some specific therapies may not pan out. Others may provide only marginal improvement over current therapies. Still others may be revolutionary, transforming our approach to entire branches of medicine. Like many emerging technologies, CRISPR is pluripotent, meaning that it can morph into many specific applications as it develops. In fact, many versions of the base CRISPR science already are diverging to meet different therapeutic needs. Each of those will have different commercialization pathways. Some may end up meeting the literal definition of hype, but many will not.

What you call something matters. Choose your words carefully.

Step 2: Start focusing on what you want. Stop chasing squirrels.

It may seem tempting to track a wide variety of “emerging technologies” on the same chart, but it doesn’t help you understand them. In fact, it doesn’t help you understand any of them. You’re like a dog let loose in the backyard with 20 squirrels running in every direction. Unless you can pick one, you’re not likely to catch any of them.

It’s the same dynamic with emerging technologies. They’re all squirrels. They all dart around the marketplace at random. If you’re trying to catch “Smart Dust,” “Connected Home,” or “Biochips” by chasing them all, you’re just like our poor dog – exhausted and frustrated. Yes, those technologies may be “emerging,” but each one of them follows different market forces, varying user applications, and independent technology development paths. Plotting them all on the same curve implies they all will follow the same “hype” (read: attention) path, and more importantly, that none will fail or that none morph into something else entirely. That’s a ridiculous assertion on its face. We have examples of many technologies failing to ever reach the “plateau” in anything resembling their initial form. “Cold Fusion” immediately comes to mind, but give yourself 30 seconds, and you could come up with a dozen other examples.

 

Technology Topic Clusters versus Hype Cycles

Another helpful edit – let’s track related, not unrelated, technologies. But wait … where’s CRISPR on the “Emerging Technologies” list? By almost any measure, CRISPR is “emerging,” yet we don’t see it. The Gartner Hype Cycle is simply a random list of whatever they want you to notice, not what’s important to you.

By definition, emerging technologies change quickly as they mature. In other words, we have non-comparable data points plotted on the same graph. It’s not about comparing apples and oranges. It’s about apples and orangutans.

Instead, let’s zero in on what matters to us. We’re interested in tracking attention specifically on “CRISPR” as a therapeutic technology. To do that, we’ll want to track “CRISPR” as a term (obviously), but also a few other related terms in a topic cluster:

  • Gene Therapy: CRISPR falls within this general category. Tracking this topic will ensure you catch other, related developments in the field as they emerge.
  • Pfizer: While not specifically focused on CRISPR, Pfizer is a major player in pharmaceuticals, and is likely to play a major role in the future of the technology. Including it ensures we don’t ignore organizations who can (singlehandedly) drive the conversation.
  • CRISPR Therapeutics: In addition to a major player, you’ll want to include a smaller organization solely focused on the topic area. You’re watching to see if their fortunes rise and fall with changes in attention … or if they do not.
  • Coronary Artery Bypass Surgery: This is an excellent benchmark – a well-understood and mature healthcare topic. No, it’s not used to treat the same conditions as CRISPR might treat, but it provides a reference point for attention. Attention in “Coronary Artery Bypass Surgery” is stable over time, whereas attention in CRISPR ebbs and flows. Comparing two topics in the field helps us understand if attention in CRISPR is high because attention in all healthcare is high, or if the variation is unique to CRISPR.

You can use the same formula with your own topic of interest: (1) Choose the specific term; (2) Add the generic technology/topic; (3) Add a major player in the field; (4) Add a minor player or startup; and finally, (5) Remember to include a benchmark for comparison. Think of it as a study group with “controls” included. It’s good research design. Consulting firms tend to struggle with good science.  Now, you know how to do better.

If you’re not careful, you’ll end up chasing squirrels … and never catch one.

Step 3: Start quantifying attention. Stop fretting over hype.

The Gartner Hype Cycle uses “expectations” as its dependent variable (on the Y-axis) as simply another word for hype. At first glance, it seems intuitive – expectations can rise and fall over time. But take a more careful look at the chart – is the expectation at the “trigger” at the same level as the “trough”? Not exactly. The trigger is more accurately a representation of “no expectations” and the trough is a representation of “negative expectations” – or at least lower than the initial condition (that’s what “disillusionment” means.) At the very least, we should draw the chart like this:

Sentiment in the Gartner Hype Cycle

Attempting to correct the scale (and the zero point on the Y-axis) highlights a more pressing issue: The measurement of “hype” as a variable at all. What is hype, exactly? How do you measure it? Is it a measure of the number of articles written within a certain amount of time? Is it a measure of the tone or sentiment of those articles? Is it the authority of the publications? Or the authority of the authors? Or the size of their Twitter following? Without a standard quantification for your variable, you don’t really have one.

Quantifying Hype 

Whenever you see a chart, be wary of undefined terms or vague ranges masking as quantified variables. In the first case, we’re given no explanation for “expectations” on the Y-axis other than “it goes up and down.” The time series ranges in the legend don’t correspond with any formula – they’re just an analyst’s guess. Ignore them.

In years past, if we wanted to quantify attention, we needed to make educated guesses based on counts of media articles, exposure to advertising, or some other indirect measure. Today, we can use actual behavioral data to track the pattern of changes in attention. Google Trends is a good place to start (and it’s free), and it’s a powerful tool if you know how to use it. In simple terms, Google Trends shows you the change in search volume on a topic (or topics) over a period of time. It will not show actual search volume (in number of searches), but it will show you much more important information in addition to change over time: relative search volume. If you compare multiple terms (up to five) on the same chart, you can see how they compare to each other. This is particularly useful when you benchmark a new term against one you understand quite well.

CRISPR Attention 2016 to 2019

Hmm. Notice something about the chart once we use actual data from Google Trends? Do you see anything like the curve represented in the Gartner Hype Cycle? No? More on that later.

The blue line represents the topic “CRISPR” over a three-year period. As you can see, interest spikes and wanes over time, but remains high as it relates to the terms and topics we’ve chosen as controls. From this data alone, what conclusions could we draw? Pfizer (the company) and Coronary Artery Bypass Surgery (the topic) draw more consistent attention than CRISPR, although CRISPR occasionally spikes. Attention from CRISPR seems to drive a bit of interest in gene therapy more generally, but it does not create a “halo effect” on any specific startup company. If I were an angel investor, I might tentatively conclude to hedge my bets on several CRISPR-related startups rather than just one. What would have learned from a “static position” on the Gartner Hype Cycle (had CRISPR even been on it)? Not very much. And what you would have learned would have been wrong.

Is Google Trends the only source for attention data? Certainly not. Engagement on social media (likes, comments, and shares) is another one. As is checking the prices for keywords in a pay-per-click auction (Google, Amazon, and LinkedIn all use these real-time systems.) What you want to avoid are “counting” metrics such as “the number of media articles written on a topic.” The media might be paying attention, but what you really want to know is if everyone else is paying attention.

Bottom line: Never settle for pure intuition without bothering to confirm it with data … especially if the data is accessible, accurate, and affordable.

Step 4: Stop imposing arbitrary phases. Start fixing time as a truly independent variable.

Time is the usual suspect for the independent variable, appearing reliably on the X-axis of most charts. But notice something about this one. Time isn’t the variable; phase is the variable – e.g. “Plateau of Productivity.” Be careful, time and phase are not interchangeable. Using one in place of the other isn’t unheard of, certainly. The more-familiar S-curve of Diffusion of Innovations uses product lifecycle stages instead of absolute time measurements as well. However, swapping time for phases is a mental trick that can fool you if you’re not careful.

Gartner uses branding not actual time

Never confuse phases for actual passage of time. When time is an independent variable, it can dramatically impact the shape of the “curve” in the plotted data – compressing or extending it. When you’re making a decision, the phase isn’t as important as the velocity through that phase. The only way you know that is by plotting actual time.

We’re used to seeing time on the X-axis. Quantum mechanics aside, time is a fixed, periodic variable. It progresses in an orderly fashion, and at the same rate, from the past into the future. It’s so obvious that we don’t stop to think about it. When you exchange time for phases, you risk being tricked into thinking that phases act in the same way as time: Linear, with fixed intervals, and progressing with consistent velocity. Emerging technologies never evolve in an orderly fashion – they are riddled with fits and starts, sometimes accelerating quickly and sometimes languishing for years. By swapping time (predictable) for phase (unpredictable), we hide the most critical information.

Any time you think about trends, a snapshot in time isn’t as important as the velocity of the movement and what triggers those movements into motion. That’s how investment decisions are made. The arbitrary, non-measurable phases in the Gartner Hype Cycle are not truly independent variables at all.

To see for ourselves, we only need to go back to our example:

Google Trends for CRISPR

Instead of “one” trigger to “start off” the “hype cycle,” we notice at least two spikes in attention in a reasonably stable volume of attention over time. The better question to ask is what triggered those changes?  Although we cannot know what will cause attention to spike in the future (sorry, Gartner), we would be wise to see what drove attention in the past. Luckily, our chart gives us some clues – two spikes in particular: One on or about May 1, 2018 and another in December of that same year.

Sticking with Google (although you could use any news aggregation tool), a time-limited search quickly reveals the answers:

Spike #1: Scientists can use CRISPR to edit genes. Should they?

Spike #2: Marc Thiessen: Gene editing is here. It’s an enormous threat

In the first case, it was a national news program (60 Minutes) that brought the technology to a wide audience. Were there stories about CRISPR in the media before? Of course, but none quite captured the broad-based public attention that 60 Minutes could generate. In an era of “micro-influencers” it seems that big media can still drive big interest.

In the second case, a Chinese scientist claimed to do what the 60 Minutes program hinted at – using CRISPR technology to edit human infant genomes. Whether he actually did it or not is a matter of debate, but news outlets pounced. As with other emerging technologies, the negative attention of a “worst case scenario” can drive interest. It’s notable that positive developments (potential therapies for debilitating disease) failed to result in a similar spike in attention.

Again, if I were that angel investor, I might reconsider my micro-influencer strategy and be prepared to fund my startup to go for the national news. I might also focus on those companies who could court negative attention – either by doing what people fear (or shouldn’t fear in this hypothetical investor’s opinion) or by preventing it.

Beware of the “phase trick” on any chart. Keep time independent of value judgements.

Step 5: Stop believing in generic cycles. Start discovering true attention drivers.

We may admit that we can’t quantify hype, that we can’t predict timing, and that we must stick with comparable data points, but the shape of the Gartner Hype Cycle curve still seems to match our everyday experience. After a trigger, hype rises quickly, peaks with a hysteria, crashes hard, and finally, slowly regains a reasonable middle ground. It all seems so intuitive.

Like so much of our intuition, it turns out that we’re wrong.

When scholars attempted to replicate the hype cycle (here, here, and here as just three of many examples) the resulting curves looked nothing like the intuitive story the curve attempts to tell us. In one particular example, researchers used actual quantitative data (gasp!) to plot the actual “hype cycles” for three emerging technologies: Internet Telephony, Gene Therapy, and High-Temperature Superconductivity. Here’s what they found:

Hype Cycle - Actual Data

See a pattern? No? Precisely. When scientists put the Hype Cycle – or any hypothesis for that matter – to the test, they’re trying to disprove it. Having an idea is just fine, but that’s just the first step. You must test it, empirically, to decide if your idea actually represents something about the natural world. If it does, you don’t accept it, but rather you continue to try to find its flaws. If it fails, you discard it. That’s how science works. Clearly, that’s not how consulting works. Here’s what one group of researchers said in the introduction to their 2013 paper (emphasis mine):

Given the model’s proclaimed capacity to forecast technological development, an important consideration for organizations in formulating marketing strategies, this paper provides a critical review of the hype cycle model by seeking evidence from Gartner’s own technology databases for the manifestation of hype cycles. The results of our empirical work show incongruences connected with the reports of Gartner, which motivates us to consider possible future directions, whereby the notion of hype or hyped dynamics (though not necessarily the hype cycle model itself) can be captured in existing life cycle models through the identification of peak, disappointment, and recovery patterns.

Let’s translate academic English into standard English: Generic hype cycles are bullshit.

To better understand changes in attention and tone surrounding CRISPR, we must understand who to pay attention to, and who we safely can ignore. We could stick with Google here and set up a “News Alert,”, but it’s more efficient to use a media tracking tool. Brand24 is a functional, low-cost option, and we will use it here. Why use an analysis tool rather than a simple raw news feed from Google? A few reasons:

  1. Media attention shifts quickly – you may have very few results for many weeks, and then a flood of them in one afternoon. You want a tool that can help you see the trend line.
  2. Influence on attention tends to aggregate – as you already know, certain media outlets, companies, universities, and individuals disproportionately impact what gets attention and when. You need a tool that can show you the “hidden” influencers in the network.
  3. Attention tends to multiply iterative and duplicate results – not all attention is created equal. The 60 Minutes piece, for example, spawned thousands of “reactions” from less-influential sources. While important in aggregate (to know what’s getting attention), the flood of information can overwhelm your analysis, leading you to follow the wagging tail, not the happy dog. Your need a tool that can analyze the origin of the topic in real time.

Let’s use Brand24 to collect some basic data over the past 30 days:

CRISPR data from Brand24

Huh, the “influencers” aren’t who we thought they’d be.

Again, our intuition fails us on multiple levels. Is Twitter the biggest source of attention (aka “news”) on CRISPR? No. It’s YouTube. Are the traditional media outlets driving the conversation? No, not recently. It’s the science-explainer YouTube channel Seeker. Although MIT Technology Review clocks in at the number three position, the usual “big media” suspects aren’t present.

What about tone and sub-topics? When you examine the most influential sources in the recent conversation, you notice something unmistakable – they’re all explainers. In other words, although there has been plenty of media coverage, the public remains largely ignorant of the underlying technology. As that angel investor, I would be wary of sentiment and opinions from the general public – they’re interested, but they don’t know enough (yet) to buy a commercially-available product. That investor might be wise to stick to investing in startups offering technical products to an informed, scientific audience – at least until public knowledge grows.

Attention doesn’t follow a set pattern. Take the time to learn what drives changes in attention for your specific topic.

Conclusion

The Gartner Hype Cycle is a brilliant piece of marketing, and it’s a boon to consultants everywhere, but it’s not going to help you make decisions. Specifically, it fails in five key areas:

  1. It chooses a loaded, emotional word (hype) instead of a variable you can objectively quantify (attention). Instead of the word hype, use the word attention to eliminate negative bias. You can always track positive and negative attention (sentiment) separately if that’s meaningful.
  2. It fails to quantify attention – simply using the word “hype” distracts us from recognizing this error. Instead, use a quantitative measure of behavioral data, such as the number of web searches on a given set of topics.
  3. It fails to stabilize time as the independent variable, using branded “phases” to mask the reality of changes over time. Instead, use actual time. Specifically, choose a time scale that makes sense for your area of interest. For fashion trends, that might be a few months. For medical devices, that might be a few years.
  4. It fails to measure the correct data, grouping non-comparable trends and technologies onto the same graph until it’s impossible to see the meaningful pattern behind any one of them. Instead, choose your topic cluster (and comparable / benchmark topics) carefully and intentionally.
  5. It fails to properly identify the shape of the curve itself, relying on our “intuition” and a “reasonable-sounding story” to trick us into not looking deeper for the real patterns. Instead, use your own data to plot the “attention cycle” unique to your area of interest. When you use your own data, you’ll begin to discover the true influencers and possible triggers. Hint: They never follow a predictable pattern. You’ll need your own data.

In other words, use this:

Attention Dashboard for CRISPR

Real, actionable, accurate, and reliable data on a topic that matters to you.

Not this:

Issues with the Gartner Hype Cycle

The next time you’re in a meeting and someone whips out a version of this chart, feel free to respond with this one. You’re welcome 😉

As we’ve demonstrated, you don’t need expensive consultants or sophisticated software to build an “Attention Dashboard” for the topic that interests you. In a few easy steps and for less than $50, we were able to draw useful conclusions about the past, present, and future of CRISPR – far more than we could learn by studying the so-called Hype Cycle, or by listening to consultants wax philosophical to the tune of thousands of dollars per hour.

If all else fails, you’re safe to bet against anyone who still believes in hype.