The Top 10 Unsolved Problems in Product Management
(A field guide for PMs who keep discovering that the real boss is entropy.)
Product management has matured enormously in the past decade-new frameworks, better tooling, and a worldwide community that will happily debate whether a roadmap should be a slide, a memo, or a mood. Yet the hard parts persist. Below are ten chronic, still‑unsolved problems in PM, each with research, practitioner quotes, and pragmatic ways to chip away at them.
1) Escaping the “feature factory” and proving outcomes (not output)
Most PMs agree we should measure success by outcomes. ProductPlan’s 2024 report shows a marked shift toward outcomes (usage, retention, revenue) over counting features shipped. But it also admits many orgs still track output because it’s easier-and because executives ask for it. In short: belief outruns behavior. (ProductPlan)
The waste is real. Pendo’s analysis of anonymized product usage found ~80% of features are rarely or never used. Their 2024 benchmarking suggests the median feature adoption rate is just 6.4%-meaning a tiny slice of your surface area drives most of the value (and support requests). (Pendo.io)
Practitioners see it, too. A plain‑spoken Redditor defines a feature factory as a team that “continually pumps out new stuff at the request of management… with little autonomy.” And a Hacker News commenter adds the cost: “More features mean a more complex system… if you’re not measuring value, parts of your product can cost far more than they’re worth.” (Reddit)
What to try next: Pair every roadmap item with a measurable behavior change (your input metric) and its link to a business outcome; kill or iterate when the behavior doesn’t move. Make “adoption reviews” as routine as launch reviews.
2) Turning strategy into something teams can actually use
Even when companies have a strategy, it often dies on impact with reality. McKinsey’s multi‑year analysis of transformations found programs are 5.8× more likely to succeed when leaders communicate a compelling change story-and 6.3× when senior messages are aligned. Product work is no different: ambiguity is the silent killer of cycles. (McKinsey & Company)
What to try next: Write the “strategy as a memo” (two pages is fine): From → To, Because, So we will, We’ll know it worked when. Then make it easy to repeat. If a designer can’t paraphrase it in standup, it’s not strategy yet.
3) Sustaining continuous discovery (without breaking delivery)
Most teams intend to talk to users; calendar physics disagrees. Teresa Torres’s benchmark is simple but hard: “At a minimum, weekly touchpoints with customers by the team building the product,” doing small research toward a desired outcome. That cadence is the difference between opinion and evidence. (Product Talk)
What to try next: Block a team‑owned, recurring 45‑minute discovery session every week (interview, usability test, or log‑review). Treat skipped sessions like an outage: capture a reason, and fix the underlying cause (recruiting, incentives, or scheduling). Tools help-but the habit matters more. (videos.producttalk.org)
4) Establishing causality: proving a release actually helped
The humbling truth from controlled experiments: at Microsoft, only about one‑third of ideas improved their target metric; a third were flat; a third made things worse. Kohavi’s work (later echoed across Amazon/Airbnb) is widely cited because it collides with our intuition. As he summarizes: our intuition is poor; 60–90% of ideas don’t improve the metric(s). (Stanford AI Lab)
You can’t A/B test everything (sample sizes, ethics, latency). But you also can’t pretend dashboards prove causality when five things changed at once.
What to try next:
Default to controlled experiments for high‑traffic decisions; use quasi‑experimental methods (switchback tests, diff‑in‑diff) where you can’t randomize.
Pre‑register success metrics and guardrails to avoid p‑hacking.
Keep a “reversal log” so you learn from ideas that hurt.
5) Prioritization under radical uncertainty (and the estimation trap)
You can RICE, WSJF, and stack‑rank all day. The problem is variance. Steve McConnell’s Cone of Uncertainty shows early estimates are inherently off by large factors-no matter how skilled the estimator-so early business cases are fragile. Using them as hard commitments turns product bets into calendar fiction. (Construx)
What to try next: Prioritize by risk retirement, not guess‑precision. Ask: “What’s the smallest, fastest test that changes our confidence the most?” Move funding from feature line items to options (bets unlocked by evidence). Teach leaders to buy information before scope.
6) Tool sprawl, data fragmentation, and the missing single source of truth
The average company now runs ~93–100+ SaaS apps; large enterprises average 231. That’s a lot of logins-and conflicting “sources of truth.” Okta’s Businesses at Work shows the number keeps creeping up, even as many firms try to consolidate. For PMs, this translates into scattered analytics, duplicate roadmaps, and painful reconciliation. (Okta)
Sprawl isn’t just annoying; it’s wasteful. Surveys peg material license waste and hidden costs from duplicate tools and shadow IT. (Tool vendors love this; your users don’t.) (TechRadar)
What to try next: Stand up lightweight product operations to own taxonomy, event instrumentation, and research repositories; federate dashboards into a few “golden” artifacts (e.g., NSM + inputs). Your goal isn’t one tool-it’s one story. (Pendo.io)
7) Shipping fast and sustainably (the debt dilemma)
Shipping is table stakes; sustainable shipping is the moat. DORA’s research ties the “four key metrics”-deployment frequency, lead time for changes, change failure rate, and time to restore-to better organizational outcomes and healthier cultures. But many orgs still treat these as “engineering stats,” not product levers. (DORA)
What to try next: Put one DORA metric on the product scorecard and discuss it in the same breath as activation/retention. Elite teams aren’t faster because they work later-they’re faster because they reduce batch size and recover quickly. (DORA)
8) Collaboration overload: alignment by meeting (and the cost of time zones)
Hybrid work made it easier to include everyone-and to schedule everyone. Microsoft’s 2025 Work Trend Index shows meetings after 8 p.m. are up 16% YoY, and ~30% of meetings now span multiple time zones. Reclaim’s tracking found professionals average 17.1 meetings/week (≈ 14.8 hours). Try strategizing with that calendar. (Microsoft)
Linked teams feel the pain; so do PMs’ families. As Lenny Rachitsky quipped on X/Twitter, “I don’t see product management work becoming faster at the same speed as engineering. I’m seeing this ratio shift.” Translation: PM “surface area” keeps expanding. (X (formerly Twitter))
What to try next: Give every recurring meeting a single decision; cancel when none exists. Push context into pre‑reads. Reserve two no‑meeting blocks per week for discovery and deep work. Treat meetings as tools, not lifestyle.
9) Designing for value without manipulation (and staying on the right side of the law)
The temptation to “optimize the funnel” can slide into dark patterns-a regulatory and reputational hazard. The U.S. FTC’s 2022 staff report catalogs common deceptive designs (hard‑to‑cancel flows, disguised ads, sneaky data sharing) and makes clear enforcement is rising. In the EU, the Digital Services Act explicitly prohibits certain dark patterns and mandates transparency reporting. PMs are now expected to build safety and fairness by design. (Federal Trade Commission)
What to try next: Add an “integrity review” (privacy, minors, accessibility, consent) to your definition of done. Keep screenshots of consent flows, cancellation steps, and pricing pages in a compliance appendix. If you’d be embarrassed to show it to a regulator-or your grandmother-don’t ship it. (FAS Project on Government Secrecy)
10) Measuring PM craft (not just product results)
There’s still no universally accepted, quantitative way to measure product managers. ProductPlan’s research shows orgs increasingly track outcomes (good!), but translating that into fair PM performance signals is thorny: results are shared, lagging, and often confounded. Cue the endless debate over KPIs vs. 360s vs. “trust me.” (ProductPlan)
The rise of product operations is one response: standardize rituals, data, and tooling so PMs can spend more time on strategy and discovery-and leaders can see the craft more clearly. (Also: fewer bespoke dashboards with eleven shades of teal.) (Pendo.io)
What to try next: Combine a team outcome score (behavior metrics + a business KPI) with a craft score (peer feedback on discovery habits, decision quality, and communication). Publish the rubric. Reward better bets, not louder slides.
Bonus problem: PM identity whiplash (and the internet’s opinions)
PMs live in the splash zone of hot takes. A popular HN comment: “The best product manager I worked with… was just an exceptionally bright person… basically a value add to the company.” A less charitable thread calls many PMs “gatekeepers.” Meanwhile on Reddit, a weary PM writes: “I learned this the hard way… my success depended on how others saw my job, not just how I did it.” None of that is data-but it is the reality of reputation in a role built on influence. (Hacker News)
A LinkedIn post, riffing on Pendo’s stat, sighed: “The saddest statistic I saw this week… 80% of features… rarely or never used.” The point isn’t doom; it’s a reminder that craft is choosing not to build most things. (LinkedIn)
Putting it all together: A battle‑tested, unsolved‑but‑improvable loop
Write a repeatable strategy. Two pages, tops. Test it in plain English. (McKinsey & Company)
Run weekly discovery. Keep it small, steady, and in the trio. (Product Talk)
Tie bets to behavior + business. Make adoption reviews routine. (Pendo.io)
Experiment where it counts. Expect to be wrong ~2/3 of the time; learn fast. (Stanford AI Lab)
Budget for uncertainty. Buy information before scope. (Construx)
Tame the stack. Fewer tools, clearer taxonomies. Product ops is your friend. (Okta)
Make flow a product goal. Put one DORA metric on your product dashboard. (DORA)
Optimize for decisions, not meetings. Use pre‑reads; cancel purposeless time. (Reclaim)
Ship with integrity. Pre‑empt dark patterns; document the choices. (Federal Trade Commission)
Measure PM craft fairly. Blend shared outcomes with behavioral evidence. (ProductPlan)
Why these remain “unsolved”
These challenges are systems problems. They involve incentives, uncertainty, and humans-none of which sit tidily in a JIRA filter. As one HN thread about “feature factories” put it, doing Scrum harder won’t save you if the org measures the wrong things. You can’t framework your way out of a misaligned system; you can only change the system-slowly, evidence first. (Hacker News)
If this all sounds a little daunting, good. PM is the craft of turning ambiguity into alignment without kidding yourself about what’s unknown. Keep the humor handy. Keep the receipts (data). And keep building teams that can argue, learn, and decide together.
“Good teams engage directly with end‑users and customers every week… Bad teams think they are the customer.” -Marty Cagan (still an aspirational norm in 2025). (Silicon Valley Product Group)