R&D-driven organizations face the constant challenge of deciding whether to continue funding existing projects and when to start new initiatives. The overwhelming majority of firms will use project prioritization to rank the opportunities as part of that exercise, with a small minority suspecting that optimization is better suited to the task of project selection.
So if is it so well-suited to the task, why isn’t optimization used and how should it be used?
Why (Almost) No One Uses Optimization
Optimization seems like something for hard-core geeks, a method that would be hard to understand and even harder to explain to management. How could we possibly explain something that throws around terms like ‘simplex’, ‘branch and bound’, and ‘simulated annealing’?
The key is to view optimization as means to an end rather than an end in itself. Frankly, it doesn’t matter if the best set of projects for the organization was selected by prioritization, optimization, the CEO’s nephew, or monkeys throwing darts at a dartboard. What matters to the organization —and what executives are truly concerned about—is pursuing a portfolio that truly supports the organization’s strategy and targets.
Compare alternative portfolios, not individual projects
This means portfolio reviews should be about portfolios, not projects, and about what the portfolio delivers, not how it was constructed. Decision-makers shouldn’t be presented with “the” portfolio, but rather a set of alternative portfolios that achieve the company’s strategic objectives in different ways.
The communication with executives should highlight the trade-offs among the portfolio scenarios, enabling them to decide which objectives should be sacrificed in favor of others. For example, are they willing to spend a bit more in order to have a greater chance of hitting the long-term revenue target? Are they willing to leave some value on the table to build up a new technology or market?
Finally, compare these alternative portfolios directly to the current, ‘base case’ portfolio. Make a strong case for why the status quo is falling short and change is imperative.
Is Project Prioritization Good Enough?
In R&D portfolio management, we want to select a set of projects that yield the biggest benefit (e.g. NPV or revenue) when resources are limited. This is precisely what optimization does.
This doesn’t sound that different from project prioritization, which ranks projects by their benefit/cost ratio, most commonly NPV/(project cost), but it differs in several fundamental ways.
- Prioritization is limited to a single constraint: the denominator you choose in the benefit/cost ratio mentioned above.
- Prioritization treats projects as independent when in fact they often are not.
- Prioritization considers each project as “in” or “out” without consideration of what a project team could do with, say 15% fewer (or 15% more!) staff.
Now let’s talk about how optimization handles each of these challenges.
Limited Resources: Not Just $$$, Not Just this Year
Optimization allows for multiple constraints, with the most important result being that multi-year budgets can be considered, so inexpensive projects selected this year won’t need to be cut in subsequent years as their spend ramps up.
Each portfolio will have different revenue, cost, staffing, and capital investment implications, which should be compared to targets and available resources
Optimization can also help balance the long-term value of early-stage projects with the short-term value of late-stage projects, either directly through the use of a minimum constraint on investment in early-stage projects, or more indirectly through a minimum constraint on the number of product launches in the out-years (when the early-stage products would be completing development).
Similar constraints can be used to ensure a minimum number of products that target a certain market, have a certain level of innovation, or advance a key technology platform. Project prioritization, in contrast, only provides a small window into which projects can or should be chosen.
No Project is An Island
Many times you can’t run one project unless some other project is also funded. For example, consider the case of a breakthrough project whose technology will be used in a series of follow-on projects.
With project prioritization, it is difficult to include rules such as “if the breakthrough project is included, the follow-ons may also be included” because the prioritization algorithm inherently looks solely at the projects ranked higher than each follow-on project when determining whether or not to fund them; what if the breakthrough project is ranked lower but is still funded?
Optimization can be used to address these dependencies through a set of constraints mandating that each follow-on project can only be funded if the breakthrough project is itself funded. More complex constraints can involve a chain of projects that are all interrelated.
Incorporating Project Scenarios
When a project has several possible development paths (e.g. buy-up/down, accelerated/delayed development), which one is the best fit for the portfolio? The most valuable option might also be the most expensive, and crowd other worthy projects out of the portfolio.
Optimization can handle this easily through a simple constraint that allows only a single scenario for each project to be selected. The optimization will select whichever scenario is best for the portfolio. It should be noted that the selected scenario may not be the “best” scenario for that project; some project value may be sacrificed in order to glean more value from additional projects or enable more aggressive scenarios for other projects to be chosen.
One Tool in the Toolkit
Nobody wants HAL 9000 telling them what to do. If executives are presented an “optimized” solution as the answer, they’re certain to push back. This isn’t to say that optimization should be swept under the rug; we’re not trying to give it Fight Club treatment. The point is that the proof is in the pudding: Rather than spend time talking about algorithms used to derive the portfolio scenarios, focus the conversation on the portfolios themselves. If a portfolio scenario looks to be the most favorable, why does it matter whether genetic, combinatorial, or no optimization methods were employed?
In truth, it’s unlikely that any algorithmic method will generate “the answer”, because it’s unlikely that all of the things the organization cares about can be perfectly captured mathematically. Methods such as optimization are intended to generate portfolios that are starting points for analysis, thought, and discussion, with the end result likely being a hybrid of a number of approaches combined with good old-fashioned human intervention. With apologies to HAL 9000, portfolio selection remains an area where artificial intelligence can help, but it’s up to the people involved to make the decisions.
Our business forecasting software, the Enrich Analytics Platform, incorporates advanced portfolio selection techniques into R&D Portfolio Management. For more information, contact us.
