0s and 1s of Product Strategy

Biologically, humans and animals have been evolving at the same rate. Humans learning a unique language made them devise strategies to conquer giant, fast, and deadly animals (1). When humans applied math, they solved many creative problems. One example is the ability to predict the funds needed for Scottish widows' pensions in 1765. The analysis of the past obituary details helped estimate the reserve amount needed by the government. The reliance on machines made humans outsource physical labor. In the second machine age (2), combining mechanical devices and math has outsourced humans' daily tasks at many levels. Machines with artificial intelligence are currently ubiquitous. But how might we train machines, presumed to execute repetitive tasks, to create product strategy?

Product strategy is a major part of a product manager's role. A product manager makes difficult choices for balancing desirability, viability, and feasibility in many organizations. Mastering the art of product strategy requires training, exposure, and being wrong a few times. In this digital era, it is not outrageous to question what if product strategy was a bunch of 0s and 1s.

Let's start with defining a product. A product is a vehicle that profitably delivers unique value to known customers and stakeholders. One of the first binaries in product strategy is identifying what the product is and is not. A very relatable example can be Southwest Airlines. It has achieved 47 years of consecutive profitability. No wonder it has made some tough yeas and nays in the process.

The same yes's and no's also apply to defining the customers. On one side, a product has its early adopters, who like to gather information and personal experience with technology before recommending it to others. Conversely, there are non-customers for whom this product would not be an ideal fit. Understanding the characteristics of these two extremes can be very assertive when designing the product further.

For a product to be competitive, it must make a series of yes and no choices. Richard P. Rumelt, In Good Strategy Bad Strategy, refers to these choices as guiding policies. He has committed a whole chapter to defining focus. For SaaS applications, one of the choices is classifying the game being played. For now, there are three games of engagements as per the desired value exchange.

Picking one category out of three is not binary. WAIT A MINUTE        

When I led this workshop with humans, I gave examples of different products and their respective games.

We then did some thought exercises to see which game of attention the participants would associate some of the familiar products with and why. Most participants classified DocuSign and Chewy as games of transactions. During one workshop, one person asked, "Aren't all products ultimately playing the game of transactions?" It was true. After all, the company should monetize a product. So, we started with a hypothesis. "We believe Netflix is a game of transaction. The more units of content their customers watch, the more value the customer would exchange." This experimentation would require that the customer would pay per movie or show. The experience would be similar to Amazon. When customers would purchase one movie, Netflix offers movie bundles. From Netflix's standpoint, they already conducted these experiments in the 2000s. They concluded that unlimited content access with one monthly fee aligned better with the customers' expected value, "attention." The more time their customers spent watching the content, the more likely they were to renew their membership.

Another participant asked why we should classify a product into only one game. What if, for some customers, the primary value is entertainment, while for others, it is efficiency? Then we went back to the customer identity framework. I quoted one famous saying,

"Designing the product for everyone is designing it for no one."

If they were to design a product for the primary persona, who would they pick and why. We time-boxed that discussion and moved on.

In future training, a participant can ask about the limitation to only three games. What if a fourth game is risk-reduction, similar to the security products running behind the scenes? That would be valid input to consider returning to the board and evolving the existing categories based on the value pyramid.

Overall, the participants adapted to the classification exercise through examples and discussions. When some participants didn't agree, they remained disagreed.

What if the same decisions were to be made by a program? How would someone go about designing the software that would categorize the product? The first question I usually start with is if it requires convergent thinking or divergent, another binary :)

Picking one game of engagement out of three possible ones requires convergent thinking. The program needs to figure out which possibility is the best match. Similar to humans, the learning mindset of the program loves examples. It loves a series of good and bad samples. It needs inputs and desired outputs. The inputs are customers suggesting the value that they perceive from the product and output can be the game of engagement. There are multiple classification algorithm and while working with the engineers, a suitable approach can be identified. I usually break these decision making processes down in multiple binaries.

Simplifying complex decisions into binary ones can reduce the cognitive load for humans and computers. H&R Block's tax preparation product is an excellent example of a product designed for homo technologicus. It automates the decisions when needed. When an option needs to be selected by a human, it only gives two choices at a time. For example, there were three options to enter certain earnings. One could log in to the earnings portal from H&R Block to transfer the earnings, upload 1099, or enter it manually. Instead of playing the game of attention by showing all the information at once, assuming one of them would succeed, H&R Block recognized the purpose of the product. H&R Block's customers value the product for its accuracy, efficiency, and mastery. H&R Block captured those values by playing a game of productivity. It first showed two choices for entering 1099 information; "type in the institute's name to log in" and Other. When the tax filer picked "Other," it asked if they wanted to upload a 1099 PDF or enter it manually. It reduced the decisions for the tax filer to 0 and 1. I finished my tax return within two hours. When designing the product, assuming that humans would be successful if they think like robots may not be lousy precedence.

One common hurdle in creating a program is that it needs more data. Suitable sample size is required before the classification can be considered "accurate." Should that be a hurdle, though? To begin with, can inferences be made? What if we ask the program to categorize all B2B products as productivity, all e-commerce products as transactions, and all media platforms as attention? When a company, Amber Labs, was launching a new wearable product for monitoring body temperature, the company needed more data on who would buy the product. As a workaround, it estimated the market size based on a current study showing the size of the population concerned about the room temperature. Inferential statistics have helped Geico quote the insurance plan lower for married people (3).

Another common hurdle is the fear of being wrong. That's when the confidence interval comes to the rescue. Initial estimates or classifications can have a 60% confidence interval. As the model gets trained further, the confidence interval is optimized. Depending upon the application, an optimal confidence interval can be derived.

The biggest hurdle is the lack of urgency in transforming those decisions into digits. If the data is insufficient, companies can achieve 90% of customer needs in many applications by making broad-stroke assumptions. If there is a fear of being wrong, digitization helps experiment with a subset of customers and take those risks. Snooze is no longer an option. The ex-Finance Minister in India doubted the possibility of digital payments in India just five years ago.

Let's now review the advantages of instructing an algorithm compared to teaching a human brain. For classifying a product into the games of engagement based on customer value, I can train about 50 humans at a time. Then, teach 50 more and the other 50. I can write a book so each person can spend an hour comprehending it. However, training the human brain becomes optional if that algorithm gets optimized to a confidence interval based on its application. The replicability of a digital decision-making machine is infinite. A calculator that was once programmed works much faster than my brain. Google Maps does much better navigation than me going through a physical atlas.

Evolution is natural, and adaptability is a must-have. Learning to solve different problems is in a Product Manager's DNA. The digital era presents a significant challenge and an even more significant opportunity. There are many unsolved enigmas. As I write this, I need to learn how to define the product's objective digitally. It is a crucial part of a product strategy. It requires divergent thinking. I've trained humans, and I believe, I can create an algorithm. We can marvel at creating digital product strategies to solve human-centric problems by mindfully leveraging 0s and 1s. Click the like button if you agree!