(Almost) Everything You Wanted to Know About Building Great Products But Where Afraid to Ask

One Startup’s Bid to Take its Product from ‘Minimum Viable’ to ‘Habit-Forming’ in 18 Months

Craig Ettinger
34 min readMay 3, 2016


This post is designed to provide, in a single writeup, an overview of some of the more valuable product development constructs and frameworks, particularly for companies attempting to launch a two-sided platform business or one aimed at leveraging network effects. Specifically, I’ll be covering:

  • Minimum Viable vs Minimum Lovable Products — is there a difference
  • Product-Market Fit — what exactly is it and determining if its been achieved
  • Platform Business Model (network effects and critical mass) — how to overcome the “chicken and egg” and “ghost town” challenges when attempting to launch a platform business
  • Habit-Forming Products — how to build them using the Hook Model

I’ve woven the story of my own startup throughout the post to hopefully make it a more interesting read (i.e., not overly clinical), and therefore more effective at conveying key concepts.

My involvement with product development began about 12 years ago when, in 2004, I was part of a team that digitized 80 years of TIME magazine and then introduced a paid archive to give consumers easy online access to all of that rich historical content. Two years later I helped launch TIME’s first mobile website (back when .mobi domains and “WAP” sites were all the rage). While at the time I didn’t exactly know what I was doing from a product management standpoint, those early experiences sparked within me a real fascination with and passion for bringing products to market.


If you know anything about successful startups, you probably know this: they are based on a founder identifying a problem that its intended customer base has and then creating a product or service that helps solve that problem. They’re formed to serve an unmet customer need — to “scratch an itch”.

In the fall of 2014, after leaving my job as general manager for TIME.com, I began working in earnest on a startup idea I’d been considering based on the following thesis:

There’s a growing flood of movies, music, TV shows, books and now podcasts and apps coming at us. And, while our entertainment options are nearly limitless, our available leisure time is not. Now, more than ever, we need help finding the needles in the entertainment haystack. Unfortunately, the existing analog or digital tools we rely on to be our guides can be inefficient, unreliable, or often leave us trapped in a filter bubble. Therefore, a clear need exists for a product designed specifically for quick and easy access to quality (reliable, relevant, interesting) entertainment recommendations. Not reviews. Not ratings. Recommendations.

Fast-forward about 16 months. With a ton of hard work and help from friends and family (including some as investors), a small digital agency in NYC, a couple of stellar Austin-based engineers, and some terrific marketing interns, I launched and have been growing an iOS app called Tastebud.


In rolling out Tastebud, I’ve aimed to apply the build-measure-learn principles of Eric Reis’ Lean Startup methodology. At its core, Lean Startup is about reducing the uncertainly inherent in new product development. Its name is a bit misleading, because it can (and should), in fact, also be applied inside of large, mature companies where innovation is often stifled due to a strong aversion to risk.

At the heart of Lean Startup’s build phase is the well-known but frequently misconstrued concept of minimum viable product (“MVP”). An MVP is that version of a new product that allows a team to collect the maximum amount of user learnings (supported by data) with the least effort. Another way to look at it is this: the MVP is ready to be shipped when it contains just enough functionality to allow it to resonate with early adopters — users who are visionary enough to “fill in the gaps” on missing features if the product truly solves a real problem. MVP enables us to answer the critical question: Are we on the right track or not? It helps us avoid building products that nobody wants by allowing us to gather the facts before it’s too late to make key adjustments or pivot.

A couple of years ago the term minimum lovable product (“MLP”) started cropping up. The problem is there doesn’t seem to be any consensus around the definition of this oxymoronic term (isn’t love inherently maximum?!!).

When attempting to explain MLP, many product people like to reference this quote by entrepreneur and President of Y Combinator, Sam Altman:

“It’s better to build something that a small number of users love, than a large number of users like.”

Taken out of context this quote doesn’t really explain anything. What if the small number of users is only one and the large number of users is 1 million? Also, that quote could very well describe a single product, rather than two different products, that has a fairly standard distribution of customer satisfaction where many like it and a small number truly love it.

If you search Google for “minimum lovable product”, on the first page of results you’ll come across several blog posts or presentations from people who (I won’t name names), in the act of making their case for MLP, have either deliberately or inadvertently misrepresented the meaning of “minimum viable product”.

You’ll see a SlideShare presentation called “The Minimum Lovable Product (forget MVP)” that contains this slide:

This slide implies that MVP is about building products that customers will tolerate.

You could ask 100 product managers for a definition of MVP and while you might get 100 different answers, not a single one would imply that companies should aim to launch a product that customers will merely tolerate (except, perhaps, for the product lead that brought us the barely functioning Obamacare website in 2013). If I were MVP, I would sue the creator of this slide for libel and defamation of character!

You’ll also see a blog post titled “So long MVP. Hello Minimum Loveable Product” that while offering some valuable insights on product development, ends up undermining itself by putting forth the following flawed “cake” analogy aimed at distinguishing an MLP from an MVP:

Meriam-Webster defines viable as capable of succeeding. The cake in the above image not only looks bland, but also dry. A tasteless cake that’s hard to swallow is clearly not a cake that has much ability to succeed.

A better example of an MVP cake would be one that’s simple yet still looks appetizing. Perhaps its dusted with powdered sugar or drizzled with icing — not entirely “featureless” but far from the elaborately decorated cupcake pictured above.

Also, what if your customer’s favorite Sesame Street character happens to be Elmo (as it is for many kids these days), or Abby Cadabby or Zoe (favorites among girls)? The site of Cookie Monster might actually leave your customer in tears. Not such a lovable product now! Would it not be better to get an initial read on which character(s) is most popular using simple cupcake toppers before going all in on cupcakes topped with a meticulously hand-crafted version of a single character?

All of this assumes, of course, that we’ve already somehow established the fact that cake and not ice cream, for example, is the treat or dessert that’s going to best satisfy our customers’ sweet cravings.

As I continued through the Google search results, I was ultimately led to what appears to the first instance of the use of the term “minimum lovable product”. It appeared in January 2013 in Henrik Kniberg’s fantastic overview of the product development process at Spotify:

On the one hand we don’t want to build a complete product before shipping it, because that would delay our learning. We can’t be sure that we are on the right track until we’ve delivered real software to real users, so we want to get there as quickly as possible. On the other hand, we don’t want to release a useless or embarrassing product. Even if we say it is a beta or alpha, people expect great software from Spotify and judge us by what we release. So the squad needs to figure out the smallest possible thing they can build to fulfill the basic narrative and delight the users. We need it to be narrative-complete, not feature-complete. Perhaps a better term is Minimum Loveable Product

To Kniberg, MLP and MVP are one and the same. Both describe the simplest version of the product that is narrative-complete (i.e. solves an existing problem) and capable of delighting users without being feature-complete. He simply suggests that MLP might be a more suitable name for the concept.

But even if we understand what we should be aiming for in an MVP, it still requires us to make the tough judgment calls as to which features are essential and which can be left out. As Reis himself says, “Minimum viable product is an attempt to get startups to simplify, but it is not itself simple.“

Prior to launching Tastebud, we ran a private beta with close to 100 test users. Based on feedback received (mostly via a great tool called Doorbell), we made some fairly significant design/UX changes to the app midway through the beta. These included changing the way transitions worked as the user scrolled through their feed of recommendations, as well as enabling the feed to be filtered by category (movies vs. music vs. books).

One change we didn’t make, however, despite several testers having asked for it, was adding in the ability to comment on other users’ recommendations or requests. The design, development, regression testing and QA required to properly implement commenting would have added at least two additional weeks to our target launch date. And, we felt that the app’s like/favorite button was “good enough” in terms of allowing users to provide (and for recommenders to receive) a measure of feedback.

Tastebud’s second App Store update included commenting

As it turned out, we didn’t launch commenting until the second app update — two full months after our public launch in May 2015. In total, we’ve pushed out 12 updates (about one per month) to Tastebud since first going live in the App Store. In addition to including bug fixes and performance enhancements in nearly every update, we’ve shipped 14 new features or upgrades to existing features. These include everything from improving find-a-friend functionality to introducing light game mechanics via user scoring and a leaderboard to adding the ability to play in-app song samples and movie trailers directly from recommendations.

The monthly updates to Tastebud reflect a continuous process of iteration, which is critical when following Lean Startup and beginning with an MVP. Reis points out that many founders have difficulty getting past the “fear of the false negative”. This is the concern that users will mistakenly dismiss a new product or service because the MVP was just too limited and had they simply seen the fully featured creation, in all of its delightful glory, surely they would have embraced it. Reis explains that the right way to overcome this fear is not by launching a fully formed product (especially not without prior testing), but rather to fully commit to iteration. Or, as our friend Henrik Kniberg says, “We [at Spotify] ensure that our products go from being great at launch to becoming amazing, by relentlessly tweaking after launch”.


Ultimately, however, if a startup is going to be successful, it needs to demonstrate product-market fit, a term that may be as difficult to define as it is to achieve. Loosely, product-market fit suggests that if after numerous iterations, you’re still having difficulty acquiring and keeping users, you’ll need to strongly consider whether it makes more sense to make a significant pivot or abandon the product altogether rather than to continue down the current path.

The term product-market fit was coined by Mark Andreessen in a 2007 blog post that asked the question, “What’s the most important factor in a startup’s success or failure — team, product, or market?” Andreessen argued that market is what matters most because a great market — one that’s large, growing and has lots of real potential customers (i.e., a problem that needs solving) — will “pull” a good enough product (it just has to basically work) out of the startup, regardless of how good the team is (just needs to be baseline competent). “Conversely,” says Andressen, “in a terrible market, you can have the best product in the world and an absolutely killer team, and it doesn’t matter — you’re going to fail.” With this in mind, Andreessen argues that the only thing that matters for a startup founder is getting to product-market fit, which means, “being in a good market with a product that can satisfy that market.”

When considering the launch of Tastebud, I saw a market for consumer entertainment content that was massive. U.S. consumers alone spend nearly $60 billion annually on movies, music, books and TV. Roughly $8–10 billion is spent annually in Apple’s iTunes on digital downloads of movies, music, books, TV shows and apps, making iTunes larger than 1/3rd of FORTUNE 500 companies. And, as we continue to move toward unfettered access to nearly all of that content, consumers increasingly need help, real help, identifying the gems, separating the good from the bad, the relevant from the irrelevant. Tastebud was going to satisfy this need. I was going to demonstrate product-market fit.

But, how exactly does one know if they’ve achieved product-market fit?
Andreessen says you can always feel when product-market fit isn’t happening and when it is happening. Not quite the precise yardstick you’d expect from one of the world’s most successful computer science engineers.

Ben Horowitz, Andreessen’s long-time business partner and author of the fantastic book, “The Hard Thing About Hard Things”, had this to say about product-market fit:

Some companies achieve primary product market fit in one big bang.
Most don’t, instead getting there through partial fits, a few false alarms, and
a big dollop of perseverance. “ “I am sure that Twitter knew when it
achieved product market fit, but it’s far murkier for most startups. How many customers (or site visits or monthly active uniques or booked revenue dollars, etc.) must you have to prove the point? It’s usually not
black and white.

Product-market fit is often expressed in terms of the characteristics that are associated with it. With strong product-market fit one typically sees low customer acquisition cost, high customer loyalty, high word-of-mouth/referral, and high lifetime customer value. With weak or no PMF one typically sees high customer acquisition costs, limited word-of-mouth/referral and high customer churn.

Product-Market Fit (“PMF”) Indicator

Entrepreneur and angel investor, Sean Ellis, who helped coin the term ‘growth hacking’ and was the first marketer at Dropbox, has tried to remove some of the ambiguity around product-market fit by coming up with what I’ll refer to as the “PMF Indicator”. Ellis’ PMF Indicator measures the percentage of a product’s existing users that would be “very disappointed” if they could no longer use that product. After comparing results across 100 startups, Ellis found that those companies that struggled for traction had a PMF Indicator under 40%, while most that gained strong traction exceed 40%. Ellis therefore concluded that achieving PMF requires at least 40% of users saying they would be “very disappointed” without your product (*note: A PMF Indicator of 40% or greater should be viewed more as a prerequisite for achieving product-market fit rather than the decisive measure for having actually achieved it).

Net Promoter Score

Another metric that can be used as a proxy for gauging product-market fit is Net Promoter Score (“NPS”). NPS is based on a single question put to a company’s customers: How likely is it that you would recommend my brand/product/service to a friend or colleague? This simple yet powerful measure of customer satisfaction was first proposed and later trademarked (along with Bain & Company and Satmetrix) by Frederick Reichheld in a 2003 Harvard Business Review article.

Perhaps the most important form of marketing for a consumer product or service is word-of-mouth (online or offline). Having a truly great product/service — one that has, by definition, achieved a high degree of product-market fit — is what turns otherwise-ordinary customers into evangelists who will put their own reputation on the line to recommend that product/service. Earning a relatively high NPS is evidence of having achieved product-market fit.

Again, we measure NPS by capturing response to a single question: How likely is it that you would recommend this product/service to a friend or colleague? Customers would be presented with a simple survey:

Promoters (score 9–10) are loyal enthusiasts who will keep using the product/service and refer others, fueling growth. Passives (score 7–8) are satisfied but unenthusiastic customers who are vulnerable to competing products. Detractors (score 0–6) are unhappy customers who can impede growth through negative word-of-mouth.

NPS is then determined by subtracting the percentage of customers who are Detractors from the percentage of customers who are Promoters.

The worst score a product/service can get is -100 and the highest score is +100. Generally speaking, any score that is above zero is good, anything above +50 is excellent, and over +70 is considered world-class. As a point of reference, Apple’s iPhone has a NPS of 63, Samsung’s is 54, HTC’s 32 and Nokia 30. Net Promoter Scores, however, vary widely by industry so its best to compare a NPS against direct and indirect competitors within one’s industry.

Of course, NPS is best used as a means for continuous optimization of the customer’s relationship to and experience with the product or service. Adding a second question to the survey — What can we do to earn a 10? — will provide you with actionable feedback and you can segment your customers for follow up: Promoters to request referrals and testimonials (and in the case of mobile apps to rate you in the App Store or Google Play); Detractors to let them know you’ll do your best to address their concerns, and with Passives to see what you can do to ensure they move up to promoters rather than dropping down to detractors.

While just about any online customer engagement tool will allow you to conduct a Net Promoter Score survey, several companies offer turnkey NPS solutions for digital products. Among the more popular are: Wootric, Promoter.io, Delighted, and Apptentive. I plan to begin formally measuring and monitoring Tastebud’s NPS once the app has gained a bit more traction with users.

One last point on product-market fit: keep in mind that just because a product has attracted tons of users early in its life, it doesn’t mean that it has achieved product-market fit. In the app market, for example, exogenous events such as significant press coverage, front page promotion on Product Hunt, partnership with a celebrity, or a quick burst of ad spending can all lead to significant downloads (sure, PR hits and promotion on Product Hunt are often earned due solely to product excellence, but they can also be obtained by knowing the right people and other quasi-manufactured means).

Live video streaming app Meerkat serves as a perfect example. It got off to an explosive start when it launched just prior to SXSW last year. It instantly became the app du jour among Silicon Valley insiders after appearing on Product Hunt and attracted more than 120,000 users in its first two weeks. Then, just after SXSW, it raised $12 million at a pre-money valuation of $40 million from Greylock, Gary Vaynerchuk, Ashton Kutcher and other Hollywood-based investors. Meerkat mania began to subside, however, as Twitter pushed its own live video efforts with its recently acquired Periscope. Meerkat continued to release new features throughout the remainder of 2015, but the buzz was very much gone. Finally, barely a year after its launch, Re/code reported that Meerkat was “failing” (its number of broadcasters had peaked in May of 2015) and was pivoting away from livestreaming video broadcasts. Meerkat’s founder and CEO Ben Rubin followed up that report with a highly candid Medium post that effectively cited a lack of product-market fit as the reason why the company needed to pivot:

So far, the value proposition of being live is just not clear to people who are not celebrities/media/news… for most regular people — it has been hard to figure out when or even why to go live.

All of these [1-to-many mobile live video] platforms are struggling to create repeat broadcasters at a growing rate and the viewership isn’t much higher today than we thought it was last summer.


Since launching last May, Tastebud has received about 10,000 app installs. And, overall and per user engagement metrics (recommendations, requests, song plays, comments, etc) have generally been trending up. But the most important metric, retention, is below where it needs to be. Tastebud has a “leaky bucket” problem. No matter how many new users we acquire, Tastebud will never achieve significant growth if existing users continue to churn out of the app at a high rate.

That’s not to say that Tastebud has no loyal users. It most definitely does. There are people who joined the app in its earliest days and have continued to use it, on a near-daily basis (often multiple times per day), to post recommendations or to discover what others are recommending. There just aren’t enough of these people. Over the past three months, Tastebud has averaged roughly 1,250 monthly active users, who have, on average, launched the app about 2.5 times per month (about 20% of users launch Tastebud 3+ times per month).

So, why exactly is Tastebud having difficulty retaining users? One explanation is that we’re experiencing the “chicken-and-egg” and “ghost town” problems, (more on these below) that commonly occur with network-effects products, products that become more valuable as more people use them (the telephone system being the classic example). A critical mass of users must be reached in order for the network effects to take hold and sufficient value to be realized. In the case of Tastebud, which has yet to reach critical mass, what many new users are experiencing is the following: (1) a lack of friends or other trusted/recognizable sources on the app when they first sign on; (2) inconsistent volume of new and relevant recommendations or requests to keep them engaged; and (3) minimal feedback on the content that is being created. Tastebud won’t be able to keep users until it has enough users.

To overcome the chicken-and-egg challenge of achieving critical mass, Chris Dixon recommends building tools into network-effects products that allow for a “single-player mode”. This provides the product with standalone or intrinsic value that enables it to generate initial user traction even before the network effects take hold. Dixon refers to this as a “come for the tool, stay for the network” strategy and offers up Delicious (my bookmarks) and Instagram (filters for my photos) as examples. Pinterest, which initially attracted many users as a personal scrapbooking tool before evolving into a more fully formed social networking platform, is another example.

Though we didn’t explicitly build Tastebud’s share functionality to help it overcome the chicken-and-egg problem, it certainly enables the app to function in single-player mode. Users can quickly search for items they want to recommend and then easily create a nicely packaged recommendation that can be fully shared out to Facebook, Twitter, Pinterest, and phone contacts (SMS). The ‘share message’ contains a large image of the entertainment item and a link to a web-based view of the full recommendation. This enables friends and followers to have a high quality Tastebud experience even though they aren’t on the app.

Example of recommendation made on Tastebud shared out to Twitter

The Platform Business Model

When we refer to Tastebud as a network-effects product, what we’re really saying is that it’s a platform; specifically a two-sided platform that enables the creation and consumption of entertainment recommendations.

Sangeet Paul Choudary an entrepreneur and advisor at 500Startups, is an expert on platform strategy and network economics. Explaining how platforms differ from traditional business models, Choudary says:

Traditionally, Pipes has been the dominant model of business. Firms create stuff, push them out and sell them to customers. Value is produced upstream and consumed downstream. There is a linear flow, much like water flowing through a pipe. Unlike Pipes, Platforms do not just create and push stuff out. They allow users to create and consume value. At the technology layer, external developers can extend platform functionality using APIs. At the business layer, users (producers) can create value on the platform for other users (consumers) to consume.

These are the defining characteristics of a platform:

1. A platform provides the infrastructure and tools for producers to produce and consumers to consume.
2. A platform creates the rules-of-play and conditions for interactions between producers and consumers to occur
3. Value is created and network effects are realized when there are enough producers and consumers with overlapping intent for interactions to spark off between them.
4. The goal of the platform is to scale its ability to enable more and better interactions.
5. The platform must manage quality and relevance (often through editorial, algorithmic and/or social curation) to ensure the interactions continue.

Mutual Baiting and Ghost Town Challenges

Choudary also refers to the chicken-and-egg challenge described above as the “Mutual Baiting” problem. For a two-sided business to work, both producers and consumers need to be on the platform. However, producers won’t come to the platform without consumers and vice versa. Consumers act as a bait to get the producers to come in and vice versa.

This problem is particularly challenging when producers and consumers are two entirely distinct groups (e.g. a two-sided network such as Uber). In the case of Tastebud, many users fill both roles; that of making recommendations and viewing (and providing feedback on) recommendations made by others.

The “Ghost Town” problem is a related challenge encountered by platforms, which often don’t have any standalone value. In its earliest days, users visiting the platform find nothing to consume and, therefore, no value in the platform. Producers, in turn, don’t contribute unless they see some consumer interest. A vicious cycle ensues and the ghost town remains a ghost town.

In addition to the “single-player-mode” tactic described above, there are several “seeding” strategies that a platform can use when attempting to overcome the Mutual Baiting and Ghost Town challenges.

Its important to note that attempting to acquire both producers and consumers at the same time can be extraordinarily difficult. It’s typically best to focus efforts on one side of the platform or the other. Once the first side has started to build it will act as bait to begin attracting the side side.

On a recent episode of This Week in Startups, Casey Winters, Product Growth Lead at Pinterest, describes (beginning at 6:57 mark) the strategy that helped launch GrubHub while he ran marketing at the restaurant food delivery service:

Do you go and get the restaurants first? Or, do you go and get the consumers first to attract the restaurants?

GrubHub actually went and got all the delivery menus for restaurants first. Put them all up online. And that was great for SEO. And that got people to come visit GrubHub. And then we went to restaurants and said, “Hey there’s all of these people that are using GrubHub to find which places to order from. Would you like to appear on the site higher than a lot of these restaurants and get more orders?

And then later on, GrubHub said let’s start with restaurants [presumably, focusing on consumers first didn’t work so well]. Let’s go send sales people to a market. Let’s sign up 50 restaurants with a pay-for-performance/cancel-at-any-time contract and then once we have these restaurants in, let’s say, SOMA we can go out and bid [on Google AdWords] on “SOMA Chinese delivery”, “SOMA Thai delivery” and get customers that way and start ranking organically.

Reddit, the successful link-sharing news site, has become a classic example of a business that ignited its massive growth by seeding the supply-side of the platform. Co-founders Steve Huffman and Alexis Ohanian were so embarrassed by the barren state of Reddit when it first launched, that they decided to populate the site with link submissions using hundreds of fake accounts. To new potential users, it appeared as if Reddit had a robust community of contributors that filled the site with great content to consume. The two founders kept up the fake account creation for several months until there were enough users on both sides of the platform to sustain the Reddit community on its own.

With Tastebud, we’ve been testing two different approaches to seeding the platform. The first is the “fake it till you make it” tactic, albeit on a much smaller scale, that Reddit utilized. Shortly after Tastebud launched we created several fake user accounts and have used them to post both recommendations and requests for recommendations, to follow other users, and to provide feedback on other users’ posts. This tactic has been largely ineffective as the number of fake accounts we created has been too small to generate any noticeable difference in the level of overall user activity. Even at their peak (when user activity was in a trough two months after launch), fake accounts represented no more than 5.0% of total weekly active users and currently represent less than 1.0%. In addition, we’ve used these fake accounts to create activity on both the supply and demand sides of the platform, rather than focusing on trying to attract either producers or consumers.

To be completely honest, despite the fact that it’s a fairly common growth tactic, using fake accounts is just something that I haven’t been able to get fully comfortable with; especially when Tastebud is designed to be a platform built on trust. Therefore, we’re ending the use of fake accounts and reallocating those efforts toward a different seeding strategy; that of having the platform itself create the product. For Tastebud, this is a more viable and effective strategy and we’ve approached it by launching a series of themed or topic-based “house” accounts. The initial set of accounts has been designed to work across multiple entertainment categories and have fairly broad consumer appeal. The “Feeling Nostalgic” account, for example, offers throwback recommendations of classic movies, songs, TV shows and novels primarily from the 80s and 90s. “Brain Binge” recommends items that will enlighten as well as entertain, such as documentary films, biographies and educational podcasts. “Celeb Picks” offer fun recommendations from celebrities that we’ve curated from magazines, social media accounts, etc.

Tastebud “house” accounts feature recommendations curated around specific themes

We’ll also be testing more narrowly focused themed accounts such as “ComicConned” (sci-fi, fantasy, superhero and nerd content), “Hip Hop” (music, movies and TV grounded in hip hop culture) and “Horror” (movies, books, and TV shows in the hugely popular horror genre). We’ll expand or prune themed accounts based on the level of consumer interest and activity they generate.

Reverse Network Effects and Curation

As critical mass is reached and the platform begins to scale, a new problem will likely emerge if the overall quality of the platform isn’t managed properly. Typically, the earliest adopters of a platform are sophisticated users who are adept at creating value. As new, less-sophisticated users come on board, their output may begin to dilute the platform’s value, which, in turn, lowers engagement and ultimately drives consumers away.

Content-based platforms such as Tastebud will fall victim to these so-called “reverse network effects” unless they maintain a high signal-to-noise ratio, where the amount of relevant content exposed to consumers is significantly greater than the amount of irrelevant content. Platforms can manage content relevancy through any combination of the three broad forms of curation:

1. Editorial Curation (expert generated): The personal tastes and opinion of human “experts” determines what content is presented to users. This manual form of curation is particularly effective in the early days of a platform before it scales. Editorial curation helps establish patterns that can then be automated and scaled.

2. Social Curation (user generated): Users provide feedback in the form of data via tools (voting, rating, favoriting, etc.) regarding the quality of the content. The aggregated user data is then used to sort and rank content and determine its relevance.

3. Algorithmic Curation (software generated): Software is used to automate the selection of what content should be presented to consumers based on a specific set of rules. Algorithmic curation is highly scalable but is often criticized for its inability to capture subtleties such as creativity, emotion and context in the way that human curation can. Personalization, which can be active or passive, falls within algorithmic curation. With active personalization, the user is selecting the rules that will govern which content is displayed to him/her. Passive personalization uses data gathered behind-the-scenes about the user — such as prior behavior or location — to present content that will likely align with his/her preferences and interests.

Example of filtering feed on Books

Tastebud currently offers active personalization in the form of a filtering tool that allows users to select which of the six entertainment categories will display in their feed as well as whether to show recommendations only, requests only, or both. In addition, the themed accounts described above offer users a form of editorial curation. Tastebud “editors” identify and recommend to users the most interesting entertainment items that fit within the given theme. There’s also currently a leaderboard of the app’s 50 top “tastemakers”, generated by a combination of social and algorithmic curation. This gives new users what is essentially a “suggested users” list of people they should follow when they first join Tastebud.

In the future, we plan to leverage social curation to provide Tastebud users with “Most Popular” and “Trending” lists. We’ll be able to feature, for example, the 25 most popular books on Tastebud (since launch, past year, past month) based on total recommendations, total favorites/saves, or a combination of both. Or, we might present users with the Top 10 songs that are trending over the past week based on recommendations or audio plays.

Its also likely we’ll at least test algorithmic curation to provide users with some form of passive personalization. This, however, will be a bit of a balancing act. While, we want to ensure that users of Tastebud always come away from the app with a recommendation of something they’ll love, we also believe that the element of surprise and serendipity are critical to providing a fantastic overall experience. Our hope is that every now and then, users of our app will be exposed to something unexpectedly great that expands their tastes in movies, music, TV or books, thereby popping any filter bubble that begins to take shape.


Viewing Tastebud through the lens of Choudary’s Platform framework helps us to be more thoughtful about how we approach the “chicken-and-egg” and “ghost town challenges”. Seeding the producer side of Tastebud with interesting recommendations from themed accounts is a great start, but we need to do more. We need to provide users with a hook!

According to Nir Eyal, a successful entrepreneur, educator and author of the bestselling book “Hooked: How to Build Habit-Forming Products,” a hook is an experience designed to connect the user’s problem to the company’s solution with enough frequency to form a habit. A habit is a behavior done with little or no conscious thought. Researchers at Duke University have found that habits account for about 45% of our behaviors on any given day. And If I can get people to use Tastebud out of habit, then I’ve reached the holy grail of product design.

From his years as an entrepreneur in the gaming and ad industries and his research on product psychology — the deeper reasons underlying why users do what they do — Eyal identified a design pattern companies use to build habit-forming products that he calls the Hook Model. The Model has four fundamental phases intended to help designers build more engaging products: a trigger; an action; a reward; and an investment.

Phase 1: Trigger

Every ‘hook’ starts with a trigger, which tells us what to do next. External triggers function by placing information within the user’s environment. They leverage sensory stimuli and contain the what-to-do-next information within the trigger itself, such as a ‘BUY NOW’ button, a phone ringing or a calendar app notification.

Internal triggers, on the other hand, rely upon memories or specific associations in the user’s mind (with certain people, situations or routines) to prompt action. Internal triggers manifest automatically in your bran and you can’t see, touch or hear them. The most frequent and powerful internal triggers are negative emotions, such as fear or grief, which produce a slight “pain” in the user and prompt a reflexive action to alleviate the discomfort.

Users who find a product that can help solve their pain or discomfort will form, over time, strong positive associations with that product as a source of relief. When we’re lonely or feeling detached we check Facebook or use Snapchat. When we’re unsure we use Google. Its important to note that new habits are initiated by external triggers, but it’s the associations with internal triggers that keep users hooked.

Phase 2: Action

The Hook Model’s action phase, defined as the simplest behavior done in anticipation of a reward, is where the habitual behavior occurs (e.g., a scroll on Vine, a search on Google). Minimizing the amount of effort it takes to get the reward is essential to habit-forming design.

According to Dr. BJ Fogg, Direcor of the Persuasive Technology Lab at Stanford University, in order for a given behavior to occur, the user must have sufficient motivation (how much we want to do a particular action), sufficient ability (the capacity to do a particular action — how easy or difficult is it to do), and a trigger must be present.

Fogg has identified three factors — each with two sides — that influence motivation: (1) seek pleasure/avoid pain; (2) seek hope/ avoid fear; (3) seek social acceptance/avoid rejection.

He’s also highlighted six factors that influence ability: (1) time — how long to complete an action); (2) money — cost to complete an action=; (3) physical effort — labor used to complete an action; (4) brain cycles — level of mental focus and effort required; (5) social deviance — how accepted is the behavior; and (6) non routine — how much the given action disrupts existing routine.

Eyal implores product designers to focus on maximizing ability before attempting to influence motivation when seeking a desired behavior, as increasing a product’s ease-of-use generally delivers a greater ROI than increasing someone’s desire to use the product.

Phase 3: Reward (Variable)

In this phase of the Hook Model the user is rewarded by having their problem solved — their itch is scratched — reinforcing their motivation for the action taken in the previous phase. In order to hold our attention, however, habit-forming products must exhibit a degree of novelty. The rewards they deliver need to be at least somewhat unpredictable. As Eyal says, “Variable rewards must satisfy users’ needs, while leaving them wanting to reengage with the product.”

The concept of variable rewards was born out of research conducted throughout the 1930s-1950s by behavioral psychologist B.F. Skinner, who studied the impact that different schedules of reinforcement (variable vs fixed) had on the rate at which a given behavior takes place and how long the behavior lasts. Skinner observed that lab rats responded most voraciously and sustained that behavior for the greatest length of time when rewards were given on a random schedule.

There are three types of variable rewards: rewards of the tribe, rewards of the hunt and rewards of the self.

Rewards of the Tribe (social rewards) are driven by our need to be connected with other people — to feel validated, accepted, important, attractive, and included. The likes, comments and reposts that we periodically receive on Facebook, Pinterest and Instagram are powerful acts of social affirmation that motivate us to keep browsing our feeds day after day.

Rewards of the Hunt (“physical” rewards) are driven by our pursuit of resources and information, a need that’s been hardwired into our brain dating back to when we hunted for physical objects such as food and other supplies necessary for our survival. The infamous “right rails” found on Daily Mail and HuffPo offer rewards of the hunt to users who never quite know what provocative story or photo they might find as they scroll through the variety of gossipy news items.

Rewards of the Self (intrinsic rewards of achievement) are based on a user’s internal motivation to gain proficiency and control; the need to feel a sense of personal accomplishment or achievement. Loyalty programs and video games are classic examples of products/services where user behavior is often influenced by rewards of the self. Many apps incorporate game mechanics (points, badges, leaderboards) in order to leverage the psychology behind rewards of the self. But game mechanics can’t be gratuitous; they must tie directly back to the problem that’s being solved in order to be truly effective.

Phase 4: Investment

In this phase of the Hook Model users are encouraged to contribute something to the product/service so that they begin accumulating “stored value” in the form of content (articles saved to Pocket), data (credit card info saved on Amazon), followers (others users followed on Twitter) or reputation (your seller status earned on eBay or Etsy).

Investments help initiate the next pass through the ‘hook’ cycle by “loading” the next trigger and they increase the likelihood of users returning over the long term by improving the product the more its used. Its important to keep in mind, however, that the investment phase inherently introduces friction into the product experience as the user is being asked to do some level of work in order to create the stored value. Its therefore critical for users to be made fully aware of the longer-term benefit provided by the stored value.

Although I’ve only provided a brief summary of Eyal’s insights into building habit-forming products, it’s probably easy to see why I became such a big fan of his after reading “Hooked” (and having followed him on Twitter). That’s why I jumped at the opportunity to attend Eyal’s workshop when I saw that he’d be offering one in New York this past January.

Nir Eyal’s ‘Habit-Forming Products’ Workshop in New York, January 18, 2016

The workshop covered much of the same material that’s in “Hooked”, but Eyal also went a bit deeper on certain topics and provided updated real-world examples. Most importantly, Eyal had attendees break into small groups and then served as a facilitator as we worked through mini case studies using our own startup products.

Going through this process led me to have a bit of an epiphany. All along I’d been hyper focused on the experience that Tastebud would offer and the utility it would provide to users who I assumed would very deliberately be looking for (or making) recommendations. I’d failed to consider that Tastebud could ultimately be used purely out of habit. Under this scenario, the average user would engage with Tastebud with much greater frequency than if they only used the app with specific intent.

So, does Tastebud have the potential to be a habit-forming product? We can quickly run it through the Hook Model to get our answer.

Boredom is the “internal trigger” for Tastebud

1a. Internal Trigger (critical to forming a long-term habit)
Question: What “pain” are we relieving? What negative emotion occurs most often that will serve as a cue?
Answer: Boredom. The frustrating experience of wanting but being unable to engage in satisfying activity, boredom seems to occur when we feel physically or mentally trapped (e.g., waiting on line at Starbucks, stuck in class with an uninspiring teacher, sitting in a theater before the movie starts, etc.). The fear of missing out (FOMO) could also serve as an internal trigger.

1b. External Trigger (used initially before the habit takes hold)
Question: What external triggers bring users to Tastebud?
Answer: A great entertainment experience might move the user to share on Tastebud what they watched, read or listened to with friends and others. A push notification of a like or comment could induce a user to open Tastebud to respond in kind. A user who has requested a recommendation will likely open the app when notified that someone has responded. Similarly, if notified that someone they follow needs help finding a great movie, book, TV show, etc., a user might feel compelled to offer a recommendation.

2. Action:
Question: What’s the simplest user behavior done in anticipation of reward?
Answer: Scrolling through the main feed of recommendations. Since the initial Tastebud MVP, the main feed and the steps it takes to get to it have evolved a fair amount. The biggest change we made was dropping the signup requirement to access the feed, which was part of a seven-step onboarding process. It now takes just two steps to begin browsing the feed of recommendations as an “anonymous” user (plus only an additional three steps to sign up if you want to be able to fully engage with the app and the Tastebud community).

Image on right show’s updated design (as of March 22) with image size increased by 75%

3. Variable Reward
Question: What are the variable rewards that alleviate the user’s pain but leave them wanting more?
Answer: As Tastebud users scroll through the main feed, the act of discovering something great to watch, to read, or listen to provides them with rewards of the hunt. They return to the stream of recommendations hoping to learn about the next cool song or binge-worthy show and stay in the know. With this variable reward in mind, we we increased the the size of the entertainment image on recommendations by 75% in a recent app update. Rewards of the tribe are exhibited in multiple ways on Tastebud. Likes, comments and responses to a request for a recommendation provide recognition and/or validation from the Tastebud community and motivate users to continue posting. Seeing that someone else shares your taste in music or books enables you to have an affinity with them and feel a sense of belonging. People offer suggestions when they see another user asking for a recommendation because they enjoy the feeling of helping their fellow Tastebud users and earning the respect of other entertainment enthusiasts. Finally, users are awarded points based on how much they’ve contributed to (e.g., recommendations that are liked) and engaged with (e.g., sharing of others’ recommendations outside of the app) the Tastebud community. A leaderboard displays the top 100 Tastebud users and doubles as a “suggested users” to follow list. This light layer of game mechanics helps highlight users’ accomplishments and status, thereby providing rewards of the self. In the future, top “tastemakers” may earn tangible benefits, such as special curation privileges or even gifts from advertising partners (free digital downloads, access to movie premieres, advance copies of books, etc).

4. Investment
Question: What ‘bit of work’ do users invest in Tastebud? Does it load the next trigger, improve Tastebud with use and increase the likelihood of return visits?
Answer: Users can ‘invest’ in Tastebud and accumulate stored value in multiple ways. Following other users creates a more personalized feed of recommendations, improving the relevance of content. Posting high quality recommendations helps users earn points, gain status and attract more followers so that future recommendations (and requests) can reach a larger and more attentive audience. Posting a request for a specific recommendation is a great example of an investment that “loads the next trigger”, as responses to the request initiate push notifications that will draw the user back to the app. Lastly, the latest app update to Tastebud included a ‘save for later’ functionality that enables users to save under their profile recommendations of movies, TV shows, books, music, apps and podcasts they’d like to consume at a later date. This stored content helps lock in users who don’t want to lose this entertainment “wish list”.

Shows act of saving a song recommendation to profile

While the lack of quick and easy access to quality entertainment recommendations is a very real, very large and tangible problem that Tastebud can solve, it will be the app’s ability to relieve the psychological ‘pain’ of boredom that will drive its audience growth and enable it to become a truly successful product.

We must therefore continue to improve the experience across all four stages of the ‘hook’ to ensure that new users will be compelled to make multiple return visits to the app and, with each successive use, increasingly come to rely on Tastebud for its ability to relieve their boredom.


Hopefully this post did its job of providing you with either a solid overview of or more clarity on what I believe are some of the most important and useful product development concepts and frameworks.

Speaking of “doing its job”, one of the more valuable product concepts I didn’t cover (simply because I didn’t really apply it to my thinking on Tastebud) is the Job-to-be-Done framework. You should definitely familiarize yourself with Job-to-be-Done, the core principle of which is that people “hire” products to do a very specific job. Clay Christensen popularized the framework and often references this quote by legendary Harvard Business School marketing professor Theodore Levitt: “People don’t want to buy a quarter-inch drill. They want a quarter-inch hole!”

Would love your feedback! What did I miss? What did you find useful?

If you enjoyed this post, please tap or click the “♥︎” to help promote it to others.



Craig Ettinger

Work History: VP, Finance & Biz Dev, Condé Nast | CDO, Parade Media | founder, Josar Media, creator of @thetastebudapp | GM, TIME dotcom