Product Manager = Fund Manager

12 min readOct 11, 2018

Managing risk by creating a compounding portfolio of product investments

TL;DR

Great product managers behave like fund managers and are able to manage risk and create value over time through compounding.
One of Product Managers’ main responsibilities is prioritizing work from a virtually infinite number of options.
Many of the tools used today are overly simplistic, and do not really capture the impact of product investment over time.
Fund managers in financial institutions also face similar prioritization tasks, making decisions on where to invest and over what time horizon.
This paper introduces a model for Product Managers to make product decisions over time, adapting the J-curve model used in private equity.
Larger enterprises should balance multiple such investments over time.

Introduction

The role of Product Management has seen incredible growth in Silicon Valley over the last 20 years since the original and widely referenced Good Product Manager/Bad Product Manager memo by Ben Horowitz. While some aspects of being a PM have changed, many still remain relevant.

Arguably the main responsibility of a Product Manager is to prioritize work, balancing a number of criteria from external market conditions to internal execution capability and company strategy. However, the tools that many PMs use to support decision-making are fairly simplistic; they work well to decide between competing features for a given product scope, but are not optimal at making broader product decisions over time.

In this article, I want to explore some new ways of thinking about the role of product managers and introduce some conceptual tools. To ground this in reality, I will use an example largely based on past experience — some details and names have been changed or omitted to protect some of the people and companies involved.

For our example, consider a product manager (we’ll call her Florence) looking to introduce a brand new product at a mid-stage B2B startup. The SaaS company has already been in business for several years and has a successful offering with hundreds of customers; however, the product team hasn’t introduced new products in the last couple of years, the competitive landscape has shifted and the company is no longer on a trajectory to achieve a billion dollar valuation.

Working closely with a brilliant engineer, Florence finds they can utilize the data that the company already collects to drive an AI-based recommendation product that would re-position the company from merely a cost saver for the IT department to a top line driver that can help customers improve monetization by as much as 50%.

Let’s follow Florence on her quest to drive innovation, lean startup style. But first, some background.

Current Practice

All PMs are aware that different projects represent trade-offs between execution effort and customer or financial impact. This is often captured in an impact-effort matrix. Dividing the matrix into quadrants for low/high impact and effort, activities fall into one of four categories.

There are a few variations on this model, such as ICE and RICE, which add additional dimensions: Reach and Confidence, to provide more granular understanding around Impact and Effort, respectively. However, this sacrifices the simple visual representation of a 2x2 matrix without necessarily adding new considerations.

To make decisions, RICE is often used as a formula to calculate a score, which gives the impression of being data-driven; however, because most of the scores are subjective or broad estimates, this can easily result in manipulation to bolster preferred projects.

These models can be used to compare relative value of various features for a single planning cycle, but may not be as effective at planning for larger, longer-term efforts.

For Florence, these tools don’t even begin to tell the story. She knows she wants to follow an experimental, data-driven approach working with a handful of early adopters, so the initial stages will be low impact from a financial perspective as well as low effort since they want to keep the team small relative to the overall R&D spend until they have validated demand.
In terms of the impact-effort matrix, this would be categorized as a fill-in, however this is just an intermediate step. This would also rate low in terms of reach and confidence, given the potential technical and product-market fit risk.
Should Florence abandon this project in favor of a few quick wins tackling bugs, incremental feature requests from customers and backlog items?

Product Manager as Fund Manager

I like to think of the role of a product manager as akin to a fund manager at a financial institution, such as a hedge fund or private equity firm. Fund managers also face an almost infinite array of investment options and must prioritize investments and make decisions to allocate funds, seeking the best return.

Fund managers typically manage a portfolio of investments with different time horizons, risk profiles and funds allocation. Fund managers manage their beta, or volatility of their portfolio, by choosing assets strategically and correcting course as needed.

Likewise, product managers should create a balanced, diversified portfolio including initiatives with varying degrees of technical/market/execution risk, different duration/time horizons, and requiring different team sizes and marketing budget. Depending on their scope of responsibility, from feature owner to head of product, the number of initiatives and potential value created will vary. Regardless, each PM can look for the right mix, and Group PMs can look to load balance across the team.

Product managers should create a balanced, diversified portfolio including initiatives with varying degrees of technical/market/execution risk, different duration, and requiring different team sizes and marketing budget.

A PM who focuses exclusively on low-risk quick wins is less likely to face failure, but is also unlikely to move the needle significantly. On the other hand, betting everything on a long-term project with uncertain returns far in the future carries a lot of risk and is not the optimal investment.

Florence has no shortage of backlog items, ranging from technical debt to much-needed UX updates to requests from the sales team. These are all solid and relatively safe bets, but she knows that there’s a much bigger opportunity, one that could potentially change the company’s trajectory.
How can Florence find a solution to this quandary? Are there tools she can use to help her drive these decisions?

A (not so) new model

In finance and other fields, the J-curve refers to the shape of return on investment over time. During the early stages, returns are often negative, as early investment is required without immediate return; eventually investment and return reach an equilibrium and the curve flattens out, although cumulative returns remain negative (this is the “bottom” of the J).

From that point onwards, returns outstrip investment and eventually cumulative returns turn positive; after that it’s all net gain. However, growth often slows down and over time the curve can start to flatten out again.

Applying this model to product management is straightforward when it comes to tracking financial metrics such as product revenue (MRR, ARR, ACV) against R&D expenses and COGS/operational costs.

This model is similar to the well-known product life cycle S-curve; however, the J-curve takes into account the bottom line instead of just the revenue.

With a bit of work, it can also be used to track “return” in other, more indirect ways using the one metric that matters most. This metric may vary significantly for each product, e.g. engagement, growth, retention or some other proxy for user value. On the investment side, using person-months instead of currency will be optimal, at least for deciding amongst multiple internal execution projects. In cases requiring significant marketing revenue, this should be captured as well.

One key benefit for using the J-curve model is to add the time dimension, making it easier to understand the various phases of the project. This also allows comparing the shape of different projects; for instance, one project may have greater outlays than others at the start but then have a steeper slope after the break-even point. Which of the following two projects would you rather invest in?

Back to our example, Florence has converged on a metric: it’s time spent, which for her video streaming product translates into revenue by creating more ad placement opportunities.
She is now evaluating a few projects. Her team is responsible for the end user experience. There are a number of well-understood issues that need to be tackled and have been prioritized, including a revamped Mobile experience. She knows that in the mid-term they’ll want to surface their recommendations in the UI, so they could get started on that, but the true power for this will come from the ML results.

For the ML side, she knows they need to invest in algorithms and it takes time to build and fine-tune the models; during this time, there’s very limited customer value or impact on monetization.

Longer term, Florence’s vision is to pivot the product (and, if things work out, the company) towards a consumer play that could significantly increase the company valuation. However, this would require many internal changes, from a sales-heavy B2B to a marketing-focused B2C. She’s aligned on this with one of the company founders, but the rest of the team does not share her vision… yet.
Florence can use the J-curve to analyze and drive internal alignment.

Applying this framework can give Product Managers who are planning longer-term initiatives 3 business super powers.

Super-power #1: Risk Management

The J-curve model makes it possible to capture risk by modifying the visual presentation slightly to include a range of outcomes. A tighter area represents a lower risk/uncertainty project, while a wider area is used to capture the increased uncertainty.

In the chart above, Florence and team have a pretty clear idea of the initial investment required (the downward part of the curve would represent engineering effort) but it’s much harder for them to predict the path beyond that point. Will time spent go up even more than they anticipated? Will it be essentially flat, which in practice would mean a negative return because of the opportunity cost of not working on other projects that could have impacted that metric?

However, as time goes by and they run subsequent experiments, the progress becomes more clear and the risk, understood as the volatility or variance in the range of likely outcomes, is typically reduced, as shown by the progressively tighter areas below.

Beyond planning, Florence can use this to track actual results against projections, helping the team improve their planning effectiveness over time. Moreover, if a given initiative departs significantly from the planned trajectory, this may require adjusting the plans and possibly shutting down non-performing experiments.

With this modeling, product managers have a super-power: the ability to time travel to do more sophisticated risk management. OK, maybe not actual time travel, but the ability to fast-forward or rewind projects. However, there’s an even more powerful consequence, with much deeper implications for longer-term projects.

Super-power #2: Compounding

Effective PMs can cut down larger initiatives into smaller projects. One of the keys is making sure that each project can stand on its own while also building up towards a larger return. Continuing with the finance analogies, this is the power of compounding returns.

In this model, instead of thinking of a project as one continued effort with a bunch of stepping stones with linear returns, it is best to think of each phase as a unit of work with net positive results. Each phase would have its own J-curve, and in turn could be broken down into stages (initial investment, break even, growth, etc.)

If a phase in the larger project has net positive results (or net present value in financial terms), then it would make sense to build it. Even if it doesn’t lead anywhere beyond that initial release, the company would still be better off building it, unless they find another initiative with much higher returns.

However, since the overall plan also contemplates subsequent phases, each with their own unique return profile (J-curve), the goal is to have each phase build on the previous one. This is the effect of compounding, where the results of one phase create a higher baseline value for the next one.

Too often, product teams chase a sequence of enhancements or new features that are merely additive instead of compounding. This results in much lower overall outcomes, just like simple interest provides lower returns than compound interest.

Florence now looks at the various initiatives in her portfolio. She decides to first tackle the lower risk UX enhancements; she expects these to have positive but relatively minor impact on the time spent metric on their own. To set things up for compounding, she also includes some of the UX needed to show recommendations, and starts with a very simple rules-based engine that allows her team to create a solid baseline.
The next phase will be the ML work. Using data collected in phase 1 and enabling user feedback loops through the UI that was previously built to allow for this, the ML project can move faster towards a solid model. The team is able to test different algorithms running off of a common UX, thus isolating the effects. Conversely, they start to see significant improvements in time spent as they replace the basic rules-based recommendations.
These two projects working together are also helping set the path towards the longer term vision. Florence would like to first demonstrate that end users are getting great value from the recommendation experience. The metric she picked, time spent, works not only for advertising purposes but as a proxy metric for end user utility. With proven results here, the discussion around a phase 3 to go direct to consumers will be a much easier sell.
This would not have been possible had they not sequenced things correctly.

Super-power #3: Agility

But wait, there’s more! Using the J-curve allows an unprecedented level of agility in planning. This is accomplished by treating each phase as a conditional project: it will only be executed if the previous phase is successful. The timing of firing off the next phase will also depend on internal and external factors, which are informed through execution.

This is often the case in venture bets (including startups), where investment is done in multiple series with (usually) increasingly large investments. Instead of having to wait until a phase is completed, as shown above, once there’s enough data to reduce the variance in possible outcomes (risk) it may be used to justify increased investment.

Plotting all these phases together, it’s possible to look across phases at a compound (and slightly wiggly) J-curve for the overall project.

In Florence’s case, her plan through phases 1 and 2 was sequential, assuming a fixed investment capacity. Phase 3 would only kick in once the trends proved positive. She didn’t need to determine a priori the exact timing of that last phase. If early results were positive, they may be able to re-allocate people from other teams. Alternatively, they could wait to have overwhelming evidence and then present to the board of directors with the company founder and seek a new investment round to fuel the consumer play.

Conclusion

Product Managers can adapt a few techniques from the finance industry, including thinking about risk as a range of outcomes and managing a portfolio of products with various risk profiles, time horizons and returns on the metrics that matter most.

Perhaps the greatest value of this model and way of thinking is to integrate the time component into planning and to make explicit the return on investment, even in cases where non-financial metrics are used. The power of compounding returns through sequencing of initiatives that can stand on their own enables results that far outweigh additive efforts, and thinking about phases as being conditional on previous results can help PMs break out of analysis paralysis.