Why Lumosity’s onboarding flow is brilliant

A step-by-step walkthrough with screenshots.

Author’s note, November 2019: I originally wrote this article in January 2015. Two months later, I joined Lumosity’s growth team as a product manager and worked there for four years. I no longer work there and if I were to analyze Lumosity’s onboarding flows now, I’d write a different article than what follows. Nonetheless, I think there’s still valuable stuff in here so I leave it up for posterity.


At first glance, Lumosity’s onboarding flow seems pretty typical. There’s a landing page, a personalization wizard, an assessment, a summary, and then a payment screen. But on closer inspection, you’ll find that the whole thing is executed uncommonly well, and that key parts are startlingly original. In this post I’ll break down the entire flow and analyze it step-by-step.

A note: I don’t work at Lumosity, and all of this analysis is my own opinion drawn from publicly available information.

The flow

All of Lumosity’s acquisition activities are designed to funnel users into two places: the website home page, and the app store download pages. I’ll pick up the story at the website home page.

Home page

Today’s Lumosity home page looks like this.

Lumosity home page, January 27 2015

If you’ve been watching over the last few months, you’ll know that this is a relatively recent update. In Q4 2014 they were A/B testing a major redesign of the home page. What you see to the left was the experimental variation, and it was the winner.

There are two key things to notice about this page.

The science. Of the five content areas on this page, four are focused on science. Lumosity is a company that tests principles rather than tweaks, and in this variation they tested the principle that foregrounding the science leads to subscriptions. Apparently it does. Compare this to the old version, which did emphasize science (this isn’t a new idea for Lumosity), but also focused on social proof (see the extraordinarily cool Lumosity members at the bottom — Sandy is literally shooting a gun while riding a horse).

Aggressive funneling. You’ll need to visit the page to see this, but nearly every single element on the page links to one place: the start of their onboarding flow. You want to learn more about Mike Scanlon, co-founder? Onboarding. Curious about that “Prestigious research network” or the “40+ scientific games”? Onboarding. Want to “Get started”? Oh good: onboarding. Lumosity is singular in its intention: they want to move users into the onboarding flow.

There must be something good going on in there. Let’s check it out.

Personalization wizard

The first phase of the onboarding flow is what I’ll call the ‘personalization wizard’. Here Lumosity collects data for the purpose of creating a customized training program.

Upon clicking “Get started” (or just about anything else on the home page), the user taken to this screen:

Let’s take a look at what’s going on here.

At the top, a headline: “Welcome! Let’s build your Personalized Training Program”. Sounds sensible — if I were starting with a personal trainer at a gym, he’d need to know what I want to accomplish.

The page then instructs: “Select all aspects of your memory that you want to challenge”. Let’s look at the options. I can apparently get better at:

  • Remembering patterns and locations
  • Associating names with faces
  • Keeping track of multiple pieces of information in my head
  • Recalling sequences of objects and movements

Hmm. Those all sound good. Can I select them all? Yep. Okay, I’d better prioritize. Which ones do I want most? How about remembering names — that’s something I often struggle with. And keeping track of multiple pieces of information in my head? Like upgrading my RAM? Sure, that sounds awesome.

Okay. Next.

On the second screen, I’m asked to select all aspects of my attention that I want to challenge. Again, all of these look pretty good. Who doesn’t want to be better at ignoring distractions?

As I continue through the next three screens, I’ll be asked about my desire to improve my cognitive speed, flexibility, and problem solving.

By the time I make it to the end of all five screens, I’ve now considered what my life would be like with each of 20 mental power-ups. All the while, the implication framing the experience was that these abilities would be attainable if I were to use Lumosity. Consider the effect that this experience has on my state of mind: I’ve been educated, and I’ve been motivated.

Let’s unpack it.

The purpose of an onboarding flow is to get the user to onboard — that is, to subscribe and pay. I haven’t subscribed. I haven’t even been given the option to subscribe. Why would Lumosity put extra work between me and the thing they really want? Do they really need to collect the data to personalize the training program I haven’t bought yet right now? Why not wait until after I’ve subscribed, or at least given my email?

Lumosity must figure that they’re going to get more of what they want (subscription and retention) if they guide me through this experience right now.

In other words, this experience is here because it improves subscription and/or retention. And if that’s the case, it means that the primary purpose of the personalization wizard is to modify the state of the user’s mind, rather than the state of a database on Lumosity’s servers. Basically, it has the same purpose as a standard product sales page.

And there’s the brilliance. The user experiences this as a survey, rather than a sales page. What does the user do when she’s taking a survey? She carefully reads the questions and considers the answers. In the experience Lumosity has created, the user applies her full, unguarded attention to imagining what her life would be like if she possessed each of 20 improved cognitive abilities, all while associating those visions with the instrumental path to realizing them — using Lumosity.

Let me just say it again.

While collecting data that will be useful for personalizing the program later on, Lumosity has made massive gains on two primary onboarding goals: education and motivation.

Education. For many people, the concept of brain training is unfamiliar. One of the boxes Lumosity needs to check off is that the user understands the benefits she can expect to gain by using Lumosity. How to do that? As soon as you get didactic about your product, you’re going to lose people in droves. This personalization wizard lets Lumosity educate users without boring them.

Motivation. As the user envisions a better version of her life (one in which she remembers faces, solves problems, and ignores distractions with ease), her motivation to take steps toward that better life increases. At the upcoming conversion phase where she’ll be asked to pay for the product, her motivation level is one of the key factors determining whether she will convert.

Stanford professor BJ Fogg’s behavior change model

This dynamic is well illustrated by BJ Fogg’s behavior change model. According to that model, whether or not a person will take a given action depends on the person’s motivation level and her ability (the perceived difficulty of the task) at the moment of a trigger.

In Lumosity’s case, the behavior in question is the user subscribing to the product. The trigger will come when Lumosity presents the payment options at the end of the flow. Whether or not the user subscribes will depend on her motivation level and the perceived difficulty of the task (which comprises factors such as the cost and her expectation of whether she’ll be able to stick with the training program) at the moment of the ask.

The personalization wizard experience increases the user’s motivation by getting her to visualize the reward she stands to earn. As Lumosity product design director Sushmita Subramanian explains,

we know from our customer research and also from a body of neuroscience research that letting people reflect and introspect and then share and disclose information about themselves actually activates parts of the brain associated with reward — similar to those that you find from food, sex, and money. So we thought that having some parts of those in the product would help us out as well.

And sure enough, in an A/B test, the team found that directing users through the survey rather than sending them straight to signup increased subscription rate by almost 10%. This was despite the fact that the survey variation performed worse on conversion to the signup page. In other words, without the survey, more users made it to the signup page, but fewer users actually signed up. (There’s a lesson in this: make sure you’re optimizing for the right metric. Josh Elman calls it the only metric that matters. Avinash Kaushik’s Digital Marketing and Measurement Model is a great tool for getting your thoughts clear.)

Smart. Now that the user is all jazzed up on visions of her own imminent superintelligence, it’s a great time for…

Lead capture

Lumosity knows that many of its users aren’t going to pull the trigger and buy on the first visit. They’ll need to be nurtured. Everyone cites different stats, but this guide to lead nurturing by Marketo claims that

up to 95 percent of qualified prospects on your Web site are there to research and are not yet ready to talk with a sales rep, but as many as 70 percent of them will eventually buy a product from you — or your competitors.

If the user leaves without providing any contact info, Lumosity will have to rely on ad retargeting and luck to get her attention again. But if the user shares her name and email, Lumosity can send her targeted, personalized, optimized messages at any time. Think of these messages in Fogg terms: each message is a chance to increase the user’s motivation and sense of ability, and each message is a new trigger that might land at a fortuitous moment.

So it’s important that Lumosity capture the user’s contact information. After she completes the last step of the personalization wizard, here’s where she’ll be directed:

A form like this — asking a stranger on the internet to hand over her name and contact information — is one of the leakiest joints of any funnel. But the Lumosity user has just spent five minutes envisioning herself with mental superpowers. And she sees Lumosity as something scientific — ie, trustworthy. I’ll bet the conversion rate on this form would make any growth hacker envious.

As we continue walking through the flow, keep in mind that if the user drops out at any point after completing this form, she’s in Lumosity’s database. She’ll be getting regular emails until she unsubscribes or pays.

Now let’s fill in this form and go on to the next step in the flow, where Lumosity asks for…

Additional data

Think back to Robert Cialdini’s famous principles of influence. Number 2 on his list is “commitment and consistency”. Cialdini says that humans are moved by a deep desire to be consistent. When we’ve committed to something, that drive towards consistency will make us more inclined to follow through with it.

At this point in the onboarding flow, the user has just just shared her name, email address, and birthdate. We could say that she’s made a commitment to seeing what this is all about. How likely is she to bail out at the next step? Not very.

That makes this a good opportunity to ask for some more information.

The text at the top tells the user that the reason Lumosity needs this information is to provide an additional level of personalization. I could write an entire post analyzing Lumosity’s use of Cialdini principles in this flow. To quote Cialdini: “A well-known principle of human behavior says that when we ask someone to do us a favor we will be more successful if we provide a reason. People simply like to have reasons for what they do.” Look back and you’ll see that Lumosity always provides a reason when asking for information.

Whether or not Lumosity needs this data in order to personalize the product, it’s valuable information for the marketing and product teams. Since the team found that adding additional complexity at this stage doesn’t hurt the onboarding metrics, how about one more page of survey questions before moving on:

After these survey questions, it’s on to the next step:

Reflecting user choices

Remember how we said that the personalization wizard isn’t primarily about personalization? So what is Lumosity going to do with the information it collected in that stage of the flow?

Here’s one thing. It will reflect that information back to the user, reinforcing the idea that this is personalized. Here’s what the user sees next:

First she gets a nice animation of a pie chart reflecting the categories she said she wants to prioritize.

This animation goes on to tell her that her priorities will be factored into her Personalized Training Program.

And finally, she’s prompted to get started on the next major phase of the onboarding flow, the Fit Test.

The purpose of the reflection phase is to satisfy the principle that personalization — or at least the expectation of personalization — will improve the user’s expectation of value and thus improve conversion and retention. In addition, it’s simply a matter of respect: the user has just answered a series of personal questions, and shows her that she was heard.

Next the user clicks “Start Your Fit Test”, and it’s on to…

Assessment

Now we’re getting to something close to the actual product. Here Lumosity is going to have the user play three games “to calibrate [her] starting point”. Again, it sounds reasonable. If I were beginning with a personal trainer at the gym, he’d need to know my starting fitness level, right?

Each game takes a few minutes to complete. They’re fairly challenging, but are also adaptive — if the user screws up, it gets easier. Aside from being good game design, this ensures that everyone, including the sharpest test-taker, gets to see that there’s room for improvement.

Here’s why that matters. Lumosity just had the user imagine all the cognitive power-ups she stands to gain. With her aspirations set high, she’s now confronted with objective evidence of her current shortcomings. This assessment will make the difference between where she is and where she wants to be painfully obvious.

Assessment 1: Speed. As quickly as possible, decide if the card that’s flashed is the same as the last one shown.

Following the first assessment, the user gets a little bit of encouragement and education.

Assessment 2: Attention. Control railroad switches to direct train cars to their like-colored homes.

Following the attention test, some more reinforcement of the science messaging.

Assessment 3: Memory. Glimpse a pattern of colored tiles, and then recreate the pattern.

Following the three assessments, the user gets another overlay telling her that the system is setting up her personalized training program.

At the second screen in the overlay, the ‘Next’ button will take the user into the penultimate section of the onboarding flow, which we can call the ‘walled garden’.

Before going on, let’s pause for a moment and take stock. The user has been in this flow for 10 or 20 minutes by this point, and she hasn’t once been asked to pay. Come to think of it, she hasn’t even seen a price tag.

So what’s been accomplished?

  1. The user is educated about the product and the concept of brain training.
  2. The user is motivated to achieve improved cognitive abilities.
  3. The user is aware of the objectively-measured distance between where she is and where she wants to be.
  4. The user has committed 10–20 minutes of her time and attention, which creates momentum for her to continue in the flow in order to be consistent.
  5. Lumosity has collected the user’s name and email address, which means that even if the user doesn’t buy today, Lumosity can communicate directly with her in the future.

Not bad at all. Closing time!

Walled garden

Now it’s time to close the deal. The user is deposited into a small walled garden of content — 6 or so pages that all flow downhill into the payment screen. In the name of saving space, I’ll show only the first page. If you want to see the others, here’s an Imgur gallery.

At the top, the user now sees her scores on the three assessments. Key point: these are not presented as contextless raw numbers. They’re presented as percentiles— how do I compare with others? This leverages the user’s natural competitiveness and curiosity about social status.

Below that, the user is again reminded of the science.

Here, the “daily workouts” line reinforces the gym-membership-like positioning and sets long-term training expectations that will serve retention.

More science.

And at the bottom, one last time, the page leverages our universal need to know where we stand compared to others.

Here’s how this walled garden works. Once the user has made it this far, any time she comes back to Lumosity on the same account, she’s limited to accessing these pages and a daily training session, which consists of three short games. The walls stay up until she either A) logs out, in which case she’s back to the beginning, or B) subscribes.

Lumosity would love for the user to proceed to the payment page and subscribe right now, but if she doesn’t, hopefully a seed will germinate in her mind. The emails she’ll be receiving almost daily should help to nurture that seed.

Finally, let’s go to the last step in the flow: payment.

Payment

When the user clicks Unlock, here’s where she’s taken.

There’s a lot of smart going on here. I want to focus on three key things.

A special offer is pre-loaded. First, notice that the page comes pre-loaded with a special offer to save 20% “today only”. (It’s there every day for new users.) This accomplishes two things. The first is that it adds a sense of urgency. I feel like I have to buy now to get that savings. The second is that it allows Lumosity to pre-fill the Promotion Code field below. That’s a fantastic idea, because these are a notorious conversion killer — users see the promotion code field and go off Googling for coupons rather than completing the purchase.

Multiple subscription length options. Offering four options for subscription length (monthly, yearly, two-year, and lifetime) accomplishes a couple of things.

First, it takes advantage of the contrast effect, whereby the subjective value of one object can be increased by positioning it alongside more expensive options. The classic example comes from a 1992 marketing research paper (sorry, no free version). Williams-Sonoma was selling a breadmaker for $275 in their print catalog. Sales of the machine were weak. Later the company introduced another breadmaker and began selling it for $429 on the same page. Sales of the original machine nearly doubled.

By offering a product for $239.96, Lumosity makes $11.95 seem cheap. In Fogg terms, this decreases the user’s perceived difficulty (of purchasing) without actually decreasing the amount of money Lumosity receives.

Second, offering multiple subscription length options allows Lumosity to capture the full amount of money a customer is willing to spend right now, rather than leaving some of it on the table for later. You can be sure that the lifetime subscription price, $239.96, is greater than or equal to the time-discounted expected lifetime value for a monthly subscriber. If a customer is motivated to spend that much right now, Lumosity will put that money in the bank.

Prices are displayed as monthly and total. Lumosity wants to make sure the user knows that she’ll save a lot by buying a long-term subscription. But Lumosity doesn’t want the user to balk because she’s confused about the total cost. So they display both, foregrounding the number that will contribute to a decreased sense of purchase difficulty.

When the user selects a plan, she’ll be taken to the checkout page.

This payment form is beautifully simple. It asks for name, credit card number, and expiration date; nothing more. At this point in the funnel, the rule is simple: minimize friction, maximize conversion.

Conclusions and takeaways

So that’s it — an onboarding flow that’s uncommonly well-executed and has at least one moment of brilliance.

What can we take away?

  1. Simpler ≠ better. Sometimes adding more complexity to the flow results in a net win — particularly if you need to establish unfamiliar background. Just make sure you understand your user’s psychological state in that moment, and minimize the burden relative to her level of commitment.
  2. Optimize on principle. Think about the deeper psychological and motivational forces at work during the user’s journey through your flow. Hypothesize principle-driven improvements to that flow, and test whether your principle was right. This results in powerful learning that will carry over into other parts of your product.
  3. Don’t just climb hills. Or risk getting stranded at a local maximum. Include in your testing program an appetite for radical, principle-driven experiments.

That’s it. Any questions or comments, get me on Twitter @mgmobrien or email me at mgmobrien@gmail.com. Happy onboarding.

Practical and epistemic rationality

In this post, I want to look at the notion of rationality and gain an intuitive understanding of what we might mean when we say that something is more or less rational.

Broadly, there are two common senses in with the term is used.

One is in the sense of what’s called epistemic or theoretical rationality. Intuitively, this is the kind of rationality that aims for the truth. The practice of epistemic rationality involves making inferences from the things one believes to the things that are logically entailed by those beliefs.

The second sense of rationality, to be distinguished from epistemic rationality, is practical rationality, which is about deciding upon actions that one could take which cohere with one’s goals. From the perspective of practical rationality, believing something is just another action, and whether or not it’s rational to do so depends on the degree to which one can expect that action to take one closer to one’s goals.

For an example, let’s take Pascal’s Wager. The idea is that we can’t be certain whether God does or does not exist. But we can cultivate a belief one way or the other by selectively exposing ourselves to the right experiences. The epistemic rationalist would say that the thing to do is to collect evidence and reason carefully about it, and believe whatever is most strongly supported by the evidence. Pascal, in a famous example of practical rationality, looks at the cost/benefit analysis. Suppose God does exist. In that case, if I believe, then God will reward me with an eternity of bliss. But if I don’t believe, he will punish me with an eternity of hell. On the other hand, if God really doesn’t exist, then it only matters a little bit whether I believe or disbelieve, because there’s no eternity in heaven or hell to worry about. Given this situation, Pascal thinks, it’s in my best interest to believe in God, regardless of what the evidence says.

These are intuitive notions. To use them in an unambiguous way, we’ll have to analyze them further and try to define, as precisely as we can, the criteria of evaluation that will be used to measure some act for its practical or epistemic rationality.

Why it matters that we understand people’s true epistemic aim

It’s difficult to get a man to understand something when his salary depends upon his not understanding it.

Upton Sinclair

If we want to get a man to understand something, but he gets paid to not understand it, then what should we do? I think that the first step is to understand that this man is not trying to believe the truth. Only then can we start to develop decent strategies.

Smart people have observed that it’s harmful to a society to have members who hold false beliefs and employ enabling epistemologies. This is because all of our success, as individuals and as groups, depends on the quality of our reasoning. False beliefs and faulty epistemology lead to poor reasoning and bad decisions.

Religious institutions can be seen as propagators of false beliefs and faulty epistemology. The strategy that the so-called ‘new atheists’ typically employ is to use careful reasoning to show that the claims that these religions make about the world are false.

However, this strategy assumes that religious people are fundamentally aiming to believe what’s true, and are simply lacking information. I claim that in actuality, people are — and should be — aiming to believe what’s good for them. In that case, then the ‘new athieist’ strategy is missing the point and is unlikely to work.

Religious belief can provide significant benefits — things like existential comfort, community, and a psychological toolkit effective at dealing with life’s difficulties. It’s difficult to get a man to understand something when his social and existential security depends upon his not understanding it.

If my answer to the Peter Thiel question is right, then the way to reduce the prevalence and limit the propagation of faulty epistemology is to reduce the switching costs and provide greater switching benefits.

Humans can cultivate false beliefs

This is a followup to a post in which I give an answer to Peter Thiel’s favorite interview question: “Tell me something that’s true, that almost nobody agrees with you on.” My answer is that if we want to be rational, we shouldn’t aim to believe what’s true; rather, we should aim to believe what’s good for us to believe.

An objector might say: “You can’t control your beliefs. Changing a belief is not like changing a shirt. Humans are forced into their beliefs by the evidence and experiences that they’re confronted with.”

In this post I’ll respond to that objection.

We can voluntarily influence our own beliefs

Let’s start by making a distinction. Let’s say that humans can have two kinds of voluntary control: we can have direct voluntary control, or we can have indirect voluntary control.

If we have direct voluntary control, we can simply choose to perform an act, and the act will occur. For example, I can simply choose to lift my hand, and my hand will rise.

Indirect voluntary control refers to situations in which we can influence or control an outcome, but only by indirect means. A good example is falling asleep. I cannot simply choose to fall asleep and immediately do so in the way that I can choose to lift my hand. But I can indirectly cause myself to fall asleep by laying in bed while I’m tired and filling my mind with relaxing thoughts.

The original objection holds that it doesn’t make sense to talk about where to aim our belief because we don’t have a choice in the matter: we lack voluntary control.

When it comes to direct voluntary control, I grant the point. I can’t choose to believe that polar bears are green.

But we do have indirect voluntary control over our beliefs. Every experience we have influences our beliefs. And we can choose, to some extent, what experiences we’ll have. We can choose whether to research a question or not; we can choose whether to reflect on an issue or not; we can choose whether to engage in a conversation or not.

Further, we often know which way our beliefs are likely to change if we take a certain action. Recall the example of the omnicidal aliens who will destroy Earth unless you swallow a pill that will cause you to believe that polar bears are green.

In practice, the process of influencing our beliefs often happens with less conscious awareness. A person might feel that a certain thing is bad or dangerous, and so avoid it. Or he might feel that something else is good or pleasurable, and so seek it. Consider the culture within many religions that censors and stigmatizes ideas which might rot the faith of the believers.


In any case, it’s clear that we do have voluntary control over our beliefs. Since the path aimed at true belief sometimes diverges from the path aimed at beneficial belief, we can’t escape the obligation to choose an aim. In the next post, I’ll discuss why it matters that we get it right.

My answer to the Peter Thiel question

Peter Thiel says that there is one question he likes to ask of interviewees:

“Tell me something that’s true, that almost nobody agrees with you on.”

Here’s my answer:

If we want to be rational, we shouldn’t aim to believe what’s true. We should aim to believe what’s good for us to believe.

Most people either think that this is false or moot. Those who think it’s false say  that it’s always rational to aim for the truth, even if the truth hurts. Those who think it’s moot say that the two aims amount to the same thing because true beliefs and good beliefs are equivalent.

But both views are incorrect. In this post I’ll address these traditional views, and show that they can’t be right.

Traditional view 1: always aim for the truth, even if it hurts

Let’s start with the commonly-held idea that we should always aim for the truth, even if it hurts.

This can be analyzed into two separate claims: we should aim to believe all truths (pursue truths), and we should aim to believe only truths (avoid falsehoods).1Following Paul Horwich Value of Truth 2006 Neither is acceptable as a universal maxim.

Regarding the ‘pursue truth’ aim: very many truths are too trivial or too costly to be worth pursuing. Take, for example, the billionth digit of pi. Most people haven’t pursued this nugget of knowledge, even though it’s easy to do (it’ s 9). You’ll be forgiven for not pursuing a true belief on this matter (and many others like it).

Regarding the ‘avoid falsehoods’ aim: it’s not hard to construct scenarios in which any rational human would choose to aim for a false belief over a true one. For example, imagine you’re abducted by aliens who convincingly tell you that they will destroy Earth unless you come to believe that polar bears are green. They then offer you a pill and credibly inform you that taking the pill will cause you to believe that polar bears are green. Would you not be rational to aim for a false belief by taking the pill?

So the maxim that we’d be rational to pursue all and only truths fails. What we actually believe is that we’re rational to pursue some truths, and accept some falsehoods.

Traditional view 2: the true and the good are the same

The second commonly-held view  is  that aiming for the truth and aiming for what’s good amount to the same thing, because the truth is always good. The alien abduction scenario from the last section gives a counterexample. We might give a more earthly example to bolster the point. Imagine you’re being interrogated in a windowless room by a government agent. On the table between you is a folder. The agent credibly informs you that the folder contains information which would correct a false belief that you currently hold, if you were to read it. You are free to read it, he tells you, but if you do, he’ll kill you. What should you do? If you’re aiming for the truth, you’ll read the folder. If you’re aiming for what’s good for you, you’ll walk away. I think that this is sufficient to show that the truth is not always the good and the good is not always the true.


I believe these examples are sufficient to discredit the idea that we should always aim to believe the truth. In the next post I’ll address the objection that claims we don’t have a choice in the matter of what we believe.