Assessing Risk Versus Reward: The Product Portfolio of Initiatives

by | Mar 7, 2017 | Artificial Intelligence, Deep Learning, How To, Strategy

In the longstanding tradition of drinking coffee, beer and your choice of liquor, coffee is considered the best choice in the morning, beer can start in the early evening, then liquor later at night. If the week isn’t too stressful, imbibing alcohol happens later in the week. Also, your post-drinking experience is the most certain (minimum risk) with coffee, and least certain with whiskey.
I have found this certainty-time framework useful while creating a product roadmap. At one end of the spectrum, there are critical customer pain points and bugs. At the other, engineers want to work on the cool stuff, with very high uncertainty of resources and successful completion. Most of the interesting stuff lies in the middle; the beer, that is.

This borrows from a strategy framework developed by McKinsey called the Portfolio of Initiatives (PoI). It is used to “develop a constantly evolving strategy for a dynamic environment,” which is exactly what we product managers need to do. That is what agile means.

To illustrate, let us take the following from the functionality of our software at MarianaIQ, which uses Deep Learning (AI) and lets customers launch personalized social campaigns.

1) Show a real-time preview of Facebook ads as they are developed
2) Replace LDA (Latent Dirichlet Allocation) with Deep Learning based representation for text in people profiles
3) Automatically come up with the name of a cluster of people, AKA a “marketing persona”

The 3 dimensions used in the framework are: certainty/risk, time, and potential market capitalization. In the product context, certainty encompasses both technical difficulty and resource requirement, time represents how we decide to schedule it, and the third dimension represents the customer revenue or usage upside.

After some customer usage and market research, #1 above clearly turns out to be a low risk, low resource, high usage feature. A coffee, which should be brewed up in the “morning” — the near-term of our timeline.

#2 is a medium risk and resource, significant upside potential idea. There are two reasons for the upside — 1) it removes the limits of manual topic tagging, required by LDA; 2) it leads to a much better representation that allows for automated clustering. That gives us the ability to be industry-agnostic, and enables automated persona model creation. This was our beer, scheduled to finish in the medium term, with high priority. It took services out of the equation before the product launch, expanded our market significantly, and tremendously increased the precision (at high recall) of our results.

McKinsey says these medium-term initiatives are where you create competitive advantage, and that’s exactly how it needs to be for product development as well. A company should have multiple active initiatives in this area of the matrix at all times.

#3 (our “risky whiskey” analogy) is technically the hardest among these, as at early stages of our planning, we’re not yet sure if our machine can generate names that make sense, especially in the business context of each customer. Furthermore, it takes a few minutes of customer’s time to give a persona’s name to a cluster.

As a result, even though a very exciting feature for engineering, it’s on the longer-term list. So we’ll keep revisiting it until the risk and reward become more clear.

Everything on the roadmap can be assessed through this lens to decide the prioritization considering the limited engineering resources we have. I highly recommend you watch the videos in this link to get more understanding of the framework, which can be an outstanding tool for designing your product strategy.