Alan Agon

Product | Engineering | Investing

Product Management in R&D

Reason for existence

The so-called raison d'être for product managers is to be the advocate of the customer. To be their voice in an already crowded and loud table. To focus the energy of the company's resources into a laser targeted at a problem that is valuable to solve from both the company and customer's perspective. With those aphorisms out of the way, what do product managers actually do? The answer is: what ever is needed to make the product successful. The reality is this often includes secretarial duties, project management, negotiation with suppliers, annual planning, and a whole host of duties rarely that are mentioned in the prototypical job description of a PM.

Delivering on the promise

Now much of the above makes a lot of sense when applying it to a traditional company — that is, a company with products either already in the market or the ability to bring new products to market. However, as we continue to push the envelope with technology our appetite for new products is starting to outweigh our ability to digest them. Robotaxis that are still yet to materialize at scale, flying cars that have been in a perpetual state of 'almost ready', fusion power that is yet to deliver on its promise, humanoid robots that always disappoint when it comes to real-world applications — the list goes on. Product managers are increasingly becoming responsible for bringing technology to market that is to put it bluntly not ready. As the PM role matures and companies come to terms with how to apply PMs, one of the most interesting disciplines is PM in R&D companies and divisions. Deepmind, OpenAI, Meta's AI Platform arm, Google Wing, Amazon's Prime Air and others all have PMs working judiciously to turn the "solution looking for a problem" mindset into reverse. The challenge is PMs have developed patterns, frameworks, and tools to be applied to areas where there is not a huge gap between technological maturity and reality.

Old world thinking

Issue backlogs, product requirement documents (PRDs), and user stories all tend to fall apart as the operating model when there are no customers and likely wont be for years to come. Sure one can try to write the PRD for a self driving car, but who in their right mind believes that writing a spec for driving — essentially defining the requirements for being a human — is realistic? What is a user story for developing a voice assistant that can handle any general problem? It doesn't make sense. In days of yore, PMs would help engineering determine how to define and refine software for problems that were straightforward; build me a widget that can link customer transactions to a database that aggregates spending habits and presents it to the user in a dashboard. Now add a feature that allows the user to do X and Y because of Z. Now add metrics that show we're on the right track and continue to optimize by looking at the backlog. What is the backlog for Deepmind? There shouldn't be one, patterns such as backlogs evolutionize products, they don't revolutionize. And there lies the rub: PMs gravitate to address the former, R&D leans towards the latter. When products are still in a state of rapid technological growth, PMs will slow them down by focusing on local maximas rather than global maximas. So then is there any benefit to having a PM in an R&D organization?

Continuing to mature the role

There are benefits to having a PM and arguably the role's original purpose remains. It's just the methods in which one uses to attain that goal must change. PMs must look for customer proxies, adopt a metrics-first mindset, determine which problems can evolve into revolutionary products and be ready to axe programs that are best left to the world of academia than to value maximizing companies. The role of a PM is still nascent compared to the likes of engineering, medicine, and other areas that have university schools named in their stead. As such, we must set expectations that adopting the PM's toolkit in R&D is far more than unproductive, it's counterproductive.


Published on September 19, 2022