OKRs: The Framework Everyone Uses and Almost Nobody Gets Right

Advocates: Archie Abrams (VP Product, Shopify)

This is the radical position, and it comes from one of the most successful e-commerce companies in the world. Shopify's core product teams operate without metrics or KPIs. Decisions are made on taste, intuition, and a 100-year vision set by CEO Tobi Lutke. Only the growth team uses rigorous metrics.

"The core product teams don't have metrics or KPIs. They're essentially banned. And instead, decisions are made based on taste, and intuition, and building towards this long-term vision."

Lenny Rachitsky

▶ 00:01:12

Shopify does use loose quarterly OKRs for coordination across teams, but these aren't binding commitments -- they're more like "if we believe X, what should we do next?" prompts. The distinction is critical: OKRs as coordination tool versus OKRs as accountability system. When you use OKRs for coordination, they help teams stay in sync without creating perverse incentives. When you use them for accountability, you get teams gaming metrics and optimizing for the OKR instead of for the customer.

The obvious counterargument is that this requires exceptional leadership taste. Not every CEO can operate like Tobi Lutke with a coherent 100-year vision that teams can internalize. But Shopify's results suggest something important: for product teams building toward a long-term vision, metrics can actually be a distraction from doing great work. When you measure something, people optimize for the measurement. When you don't measure, people optimize for what they believe is right. If you've hired well and set clear direction, "what they believe is right" is often better than what the metrics would have told them.

The harder question is whether this can survive the departure of the visionary leader. Shopify's no-KPI approach is deeply tied to Lutke's personal judgment. Most companies can't replicate that.

Advocates: Shishir Mehrotra (CEO, Coda)

Shishir Mehrotra takes a middle path that is deceptively radical: use OKRs, but reject the piece of Google lore that has infected every implementation. Google's guidance that hitting 70% of your OKR is "success" made sense when AdWords was printing money in the basement. For everyone else, it's a license to sandbag or, worse, to feel good about missing targets.

"We throw out the piece about 70% of hitting your goal being considered success. I believe this is a luxury you have when there is a money-printing machine in your basement (like Google). Instead, we goal around getting to 100% of our commitments."

Shishir Mehrotraon Lenny's Podcast (on rejecting Google's 70% OKR target)

The 70% target was designed to encourage ambitious goal-setting: if you're hitting 100% of your OKRs, you're not being ambitious enough. In practice, it creates a muddled accountability culture. Did you miss because the goal was stretch, or because you underperformed? Nobody knows, which means nobody is held accountable.

Coda's 100% target flips this dynamic. When you commit to 100%, the commitment itself is the unit of accountability. You're not being asked to hit a moonshot -- you're being asked to make a realistic commitment and deliver on it. This changes how teams set goals: they think harder about what's actually achievable, which paradoxically leads to better outcomes than the "set stretch goals and celebrate 70%" approach.

Coda's other key insight is the distinction between input and output metrics. Many teams set output metrics like "improve activation rate by 15%" as their OKR, when they should be setting input metrics they can actually control within a quarter. Output metrics are the strategy; input metrics are the OKR. You can't directly control activation rate, but you can control the number of onboarding experiments you run or the speed of your trial-to-paid flow. This distinction alone would fix most dysfunctional OKR implementations.

Advocates: Asha Sharma (CVP of AI Platform, Microsoft)

Asha Sharma's approach at Microsoft AI splits the difference between structure and flexibility. You need a shared north star metric that gives everyone direction, combined with loose quarterly OKRs that provide enough structure for coordination without becoming rigid commitments.

"What is the north star metric is something that we do. The second thing that we do is that we have kind of loose quarterly OKR."

Asha Sharma

▶ 00:32:06

The key word is "loose." In fast-moving domains like AI, what you learn in week three may invalidate what you committed to in week one. OKRs that prevent pivoting are worse than no OKRs at all. The north star metric provides the anchor -- it tells everyone what "winning" looks like. The loose OKRs provide coordination -- they tell teams what they plan to do next quarter to move the north star. But crucially, teams are expected to change their plans as they learn.

This approach is particularly well-suited to companies navigating genuine uncertainty. If you're in a stable market with well-understood levers, rigorous OKRs make sense. If you're building AI products where the technology capabilities change monthly and customer needs are still being discovered, you need the flexibility to pivot without feeling like you "failed" your OKRs. The north star keeps everyone pointed in the same direction even as the specific plans shift.

The risk is that "loose" becomes "meaningless." Without some accountability mechanism, loose OKRs can degenerate into aspirational wish lists that nobody references after the planning meeting. Sharma's model works because the north star metric provides the accountability that the loose OKRs don't -- teams are still judged on whether they're moving the metric, even if how they move it changes quarter to quarter.