Goal-Setting: S.M.A.R.T. + Stretch + Pyramid

The framework

OKRs aren't new. Peter Drucker formalized Management by Objectives in 1954[1]. Andy Grove evolved it into iMBO at Intel in the 1970s[2]. John Doerr brought it to Google as OKRs in 1999[3]. The mechanics are settled. How you actually run them at scale isn't.

Here's the framework I run:

Rule 1: S.M.A.R.T. + Stretch — both required

Most goals fail one of two ways. Either they're S.M.A.R.T. (Specific, Measurable, Achievable, Relevant, Time-bound) but utterly safe — the team commits to what they were going to deliver anyway. Or they're audacious but unmeasurable — "make customers love us" with no key result. S.M.A.R.T. without stretch is bureaucracy. Stretch without S.M.A.R.T. is wishful thinking.

Locke and Latham's 35-year goal-setting meta-analysis arrived at the same conclusion: specific, hard goals produce higher performance than urging people to "do their best"[4]. Hard and specific. Either alone underperforms.

Rule 2: Three difficulty tiers — aim for 70% across the portfolio

When I calibrate stretch, I sort every goal into one of three tiers:

Tier 3 is what Jim Collins called BHAGs in Built to Last: Big Hairy Audacious Goals[5]. The point isn't the score. It's that if you nailed this one, the trajectory of the org would change.

Across all three tiers, the portfolio-wide target is ~70% completion. Doerr is explicit: the expected score for OKRs is 0.7. They should stretch capabilities to the limit and should sometimes result in failure[6][12].

This number changes behavior. When 100% is the implicit standard, people sandbag — they propose easy goals to guarantee the score. When 70% is the standard, ambition becomes safer than caution.

Rule 3: Pyramid alignment — everything rolls up

Goals at every level must ladder up:

If I'm an IC, my goal should ladder to my team's. The team's to the division's. The division's to the company's. The company's to its stated mission. If a link in that chain doesn't make sense, one of the goals is wrong.

The test: ask anyone in the org "why does your goal matter?" and they should walk up the pyramid in three sentences. If they can't, the goal isn't aligned. It's just work.

Rule 4: Personal development stays separate

Personal development goals — "learn ML," "improve public speaking," "get a coaching certification" — are not performance goals. They belong on a separate track, with separate review cadence, with separate weight in compensation conversations.

When you mix them, both lose. Personal dev gets evaluated on business outcomes (it shouldn't). Performance goals get inflated with development items (they shouldn't be). The cleanest orgs I've worked in keep two separate documents.

How I apply it

Quarterly cadence at the org level (still the right unit for most teams), with monthly check-ins. Each direct report has 3-5 goals max — more than that and nothing is a priority. We tier them explicitly before committing[7]. Each goal has one number that decides success — no qualitative scoring, no "we partially achieved the spirit of the goal."

Two things I do that aren't in any book:

Pre-mortems. Before committing a Tier 2 or Tier 3 goal, I write the failure post-mortem in advance. "Imagine it's end of quarter and we missed this. What's the most likely reason?" The answers usually expose either an unstated assumption or a missing dependency. Cheap insurance against expensive failure.

Carry-over budget for multi-quarter goals. Some Tier 3 goals span quarters. I formalize this: ~20% of capacity each quarter is reserved for multi-quarter goals. Without this protection, every audacious ambition gets squeezed by short-term delivery pressure. With it, the moonshots have a lane.

Where this breaks

The framework's failure modes are well-documented. Worth naming so you spot them.

Goodhart's Law. "When a measure becomes a target, it ceases to be a good measure"[13]. Pick metrics carelessly and you'll get the metric optimized at the expense of the thing the metric was measuring[14]. The fix isn't avoiding metrics — it's using leading + lagging pairs and being willing to revise mid-quarter when you see gaming.

Cascade dilution. When goals cascade through 5+ levels, strategic intent gets watered down at each step. By the time it reaches a frontline IC, the goal is unrecognizable from the company-level objective it was supposed to support. Counter-fix: skip-level alignment reviews, where every IC's goal is read alongside the L+2 goal it should ladder to.

Sandbagging. The 70% target is supposed to encourage stretch. In practice, teams burned by leadership "punishing" missed goals will quietly propose Tier 1 goals dressed as Tier 2[15]. Counter-fix: leaders publicly score themselves on the same scale, and explicitly reward Tier 3 attempts that fell short over Tier 1 successes that overdelivered.

Mixing in personal development. The most common failure I see. Companies that bake "growth" into performance reviews end up with neither real growth nor real performance pressure. Pull them apart and protect each.

Goal-setting mechanics will keep evolving — AI is already consuming parts of the workflow. The tooling will shift. The principles won't: ambition needs measurement, measurement needs stretch, stretch needs alignment, and growth deserves its own protected lane.