In 2002, Danny Kahneman was awarded the Nobel Prize in economics for his groundbreaking work on the psychology of judgement and decision making. Along with Amos Tversky, his long-time collaborator, Kahneman uncovered systematic shortcuts in how the human mind works that often turn out to be wrong. They labeled these systematic biases “heuristics”, and described three in depth: Availability, Representativeness, and Anchoring (later researchers discovered many more). One common thread between the various heuristics is how horrible humans are at thinking probabilistically.
There are entire industries predicated on exploiting these cognitive biases to get people to do things that are not in their best interest, such as consumer finance. Predatory lenders trap unwary consumers in lifelong debt, and charge usurious 20-30% interest rates. On his podcast “Against the Rules”, Michael Lewis interviews Jason Brown, the creator of an app called Tally which helps consumers get out of debt. Brown describes the way credit card companies use the Anchoring heuristic to trick people into remaining in debt: the big, bold number on a credit card statement is typically the minimum payment due. By anchoring the customer’s mind to a smaller number, they pay a smaller amount than they otherwise would, keeping their debt and interest payments higher. These types of financial sleight of hand disproportionately impact the poor and the elderly.
Tally helps consumers make better financial decisions and get out of credit card debt. Users connect multiple credit cards to the app, and Tally intelligently pays off the right cards at the right time to minimize overall payments. Tally is one example of an emerging software trend dubbed Coaching Networks: systems that use data and an understanding of human psychology to “nudge” users towards making more intelligent choices and choosing healthier behavior. While traditionally professional coaching was only available to elite athletes and business executives, recent technological advances are bringing the cost down and making coaching available to a wider audience.
Coaching Market Landscape
The US Self Help industry is estimated to generate $11 billion annually, with a little over $1 billion going to personal life coaches. The market for coaching business executives is much larger, with estimated revenue of $15 billion in 2020, and growing on average 5% per year from 2015-2020. According to a PriceWaterHouseCoopers study commissioned by the International Coaching Federation, coaching is the second-fastest growing industry in the world, after only Information Technology. The 2016 survey concluded that there were 53,000 professional coach practitioners worldwide, with coaches in the US averaging income of $62,000 and serving 11 clients at a time. The majority of respondents listed “managers” and “executives” as clients.
The growing demand for coaching has encouraged a slew of startups to enter the space. BrightEye Ventures produced this market map of the Professional B2B coaching and mentoring landscape, with level of automation on the Y-axis and hard vs soft skills on the X-axis. They estimate $640M has been invested in this space over the last 8 years, with BetterUp($145M), Chorus ($55M) and Guru ($40M) leading the way.
Most of the activity in the space to date has been in Quadrant I, with human coaches teaching soft skills. Most of these offerings are essentially an agency model, with a roster of vetted coaches available for deployment to specific clients. Leaders BetterUp and Torch go one step further and offer an integrated assessment and analytics software suite to help coaches create individually tailored programs. For independent coaches, Maslo offers an AI-powered journaling app to help coaches track client progress.
Coaching Networks
In the above market map, Quadrant III is automated coaching for hard skills. Companies in this area are re-imagining the definition of coaching, replacing the human coach with a Coaching Network, defined as software that uses machine learning to guide workers toward doing their jobs more effectively while they’re doing it. Emergence Capital describes Coaching Network design in 4 steps:
1 – Gather Behaviors: Sensors passively gather user activity: written, verbal, physical
2 – Combine With Peers: Data is used to train a Machine Learning model, emphasizing contributions from the best performers
3 – Predictive Advice: Users are coached with individually tailored, real time suggestions and feedback
4 – Human Creativity: Brilliant outliers improve the overall performance of the network
Guru provides an instructive example. Guru offers both a Chrome plugin and Slack integration that allows workers to capture “cards” of information as they type or search. Cards are then tagged and verified by experts within the organization, before being added to the corpus of institutional knowledge. Guru then proactively suggests information cards with relevant content as users compose messages and emails.
Google Docs has integrated a similar feature in the “Explore” side panel. As I type this article in the Google Docs editor, I have quick access to a number of related documents “suggested based on the document content”.
Chorus is a Coaching Network that uses AI to analyze sales calls. Calls are recorded and transcribed, and the Chorus software then uncovers patterns that describe the calls of the best performing salespeople. These patterns are used to make real-time suggestions to other salespeople. Another startup called Textio bills itself as an “augmented writing platform” that helps recruiters write better job postings. Textio measures hits from thousands of job postings to give recruiters feedback on language that is most inclusive and best represents the values of their company culture.
By automating the process of “nudging” users into better behavior, Coaching Networks make the benefits of coaching scalable and much more accessible. Companies that pay executive coaches $500/hr can now deploy coaching programs for a few dollars per month per employee.
Consumer Niche Coaching Network
The coaching networks mentioned above are focused on Enterprise customers. What are the areas where consumers could most benefit from ubiquitous coaching?
Finance
We have already seen the example of Tally above in helping consumers pay off credit card debt. “Robo-advisor” companies such as Betterment, Ellevest, and Wealthfront help consumers by automating the investment process.
Insurance
Insurance is an area where consumers are expected to make complex, life altering decisions based on very little data. And they are up against massive, well-funded, multinational corporations who are experts in the field. Imagine an insurance app that coaches users through the various terms and options when purchasing life or automotive insurance. The cost would justify itself when compared to the savings over the lifetime of a policy.
Career
Who doesn’t wish they had received better career advice early on in life? Many of us make decisions at the tender age of 18 with input only from our parents or a few meetings with a high school or college guidance counselor. Although our parents and school counselors mean well, much of their experience is out of date and can’t possibly reflect the dynamic, rapidly changing nature of the modern economy.
As automation disrupts all sectors of the economy, constant learning and job retraining will become the norm. There is a huge opportunity here for a Coaching Network that provides up to date, individually tailored guidance to help people make smarter career decisions. Again, the cost of a few hundred dollars on software is far outweighed by the potential tens of thousands of dollars earned from a better job placement.
Communication
Executive communication coaches routinely earn $400-500 per hour coaching senior managers. What if the same calibre of communication coaching were available to everyone? Textio has already demonstrated how better writing can benefit recruiters, and it is not hard to imagine how the same technique could be applied to help job seekers write better cover letters and resumes. A consumer-facing communication coaching app would take inputs from emails and text messages to give users feedback on sentence structure, word choice, and tone. Phone calls could be added to give feedback on vocalized pauses and filler phrases, rate and pace of speech, and clarity of expression.
Problems With Coaching Networks
For the most part, the machine learning algorithms that underpin Coaching Networks are open source and freely available. Emergence Capital advises aspiring startups to niche down and build a data set in a particular market to differentiate themselves and create a competitive advantage.
However, Andreessen Horowitz argues that these “data moats” are a weak defense to competition. The problem with data networks is that the cost of adding additional unique data to a corpus goes up while the usefulness of that data goes down. Take the example of a customer service chatbot. Training responses to the most basic customer queries like “Is my router connected?” are relatively easy and extremely valuable because these queries are so common, so training the first 40-60% of responses is a synch. But in order to interact with live customers, the chatbot needs to be able make reasonable responses to the incredibly specific and nonsensical questions humans are likely to ask such as “If my cat walks across the router, does it make the WiFi signal fuzzy?” or “Why doesn’t Alexa respond when I say Hey Google?”. Training the final 20-40% of “long tail” human queries turns into an infinitely difficult (and expensive) problem to solve.
Coaching as Design
Richard Thaler, Nobel Laureate and longtime collaborator of Kahneman and Tversky, says that changing people’s behavior is hard unless you make it very easy for them to choose the right behavior over the wrong. He often tells an anecdote about a bowl of cashew nuts: at a dinner party he hosted in the 1970’s, Thaler served a bowl of cashews with cocktails, and observed his guests devouring the nuts and ruining their appetites before dinner. So Thaler decided to hide the nuts in the kitchen, and his guests thanked him for removing the temptation. The lesson was simple: he made it easier for them to make the right choice. Thaler’s insight led him to establish the field of Behavioral Economics.
Thaler suggests that you can set up “choice architecture” to help guide people to make better decisions for themselves. Choice architecture can be built-in to the design of anything humans interact with: from the flow of pedestrian traffic in a public space, to the ergonomics of a household appliance to the UX of a well designed website. The irresistible stairs at Seattle’s Bullitt Center are an excellent example of (literal) choice architecture. The architects designed the carbon neutral building so that when visitors enter they are greeted with a light-filled staircase with stunning views of Puget Sound. The steps are made from Douglas-fir and large landings at each level encourage conversation and chance encounters. By making the stairs so inviting, and concealing the main elevator bank behind doors at the far end of the reception hall, the designers made it both easy and enjoyable for people to get more exercise without having to think about it. And research showed that ⅔ of people who work on the top floor opt to take the stairs on their way to work.
Just like good design, good coaching is under appreciated because it is invisible. People only notice bad design because it creates friction and obstacles to completing tasks. Well-designed products function so smoothly the user forgets they are there, while their mind is on whatever task they want to complete next. The question is: how can we incorporate coaching and positive “nudges” into products and services so seamlessly that users don’t even realize they are being coached?
The Labyrinth 