OKR vs KPI for analysts and PMs
Contents:
Why analysts need to care
Almost every product analyst interview at Stripe, Airbnb, Uber, or DoorDash includes some version of "What's the difference between an OKR and a KPI?" or "What KPIs would you propose for this product?" Hiring managers use it to filter dashboard-builders from people who can shape the goal system those dashboards feed.
An analyst owns the numerical layer of the company — you compute metrics, design experiments, and tell PMs whether the line moved because of the launch or because of seasonality. Without the goal-setting frame above your dashboards, you end up producing accurate but irrelevant numbers while the team drifts.
Short version, unpacked below: KPI is the speedometer, OKR is the navigator. One tells you how the engine runs now. The other tells you whether you're driving toward the right city.
What a KPI actually is
A KPI (Key Performance Indicator) is a single metric with a target value that you track on a recurring cadence to understand the health of a process, team, or product surface. It is meant to be boring on purpose — you want a flat or improving line, not drama.
A useful KPI is concrete (resolves to a number or percentage), computable from the warehouse without bespoke surveys, targeted with a number it should hit or stay above, and recurring on a fixed cadence with the same definition.
Examples of KPIs an analyst owns
| Area | KPI | Typical target |
|---|---|---|
| Product engagement | DAU | 50,000 |
| Retention | Day 7 retention | 25% |
| Marketing | CAC (blended) | < $40 |
| Monetization | ARPU | $4.50 / month |
| Support | First response time | < 2 hours |
| Reliability | API p99 latency | < 250 ms |
A KPI works exactly like a speedometer on a dashboard. DAU dipped below 50,000 on Tuesday? You don't celebrate, you don't panic — you open the segmentation view and figure out which surface broke. The KPI itself doesn't tell you what to do, but it tells you that you should do something.
Gotcha: A KPI without a target and a trendline is just a number on a slide. "DAU was 48,213" means nothing — you need the target, the prior period, and the cohort split before it becomes information.
What an OKR actually is
OKR (Objectives and Key Results) is a goal-setting framework popularized at Intel by Andy Grove and scaled at Google by John Doerr. It has two parts and they are not interchangeable.
The Objective is a qualitative, ambitious statement of direction. No numbers. It answers "Where are we going this quarter?" and it should be memorable enough that an engineer can recite it without checking Notion.
The Key Results — typically 2 to 5 per Objective — are the measurable outcomes that would prove the Objective is being achieved. They are quantitative, time-bound, and live in your warehouse. They answer "How will we know we got there?"
A worked OKR
Objective: Make onboarding feel obvious to a first-time user.
Key Results:
- Lift Day 1 retention from 30% to 45%.
- Cut median time-to-first-value from 5 minutes to 2 minutes.
- Reduce share of new users who contact support on day 1 from 8% to 3%.
A healthy team hits roughly 70% of its Key Results and calls that a win. OKRs are designed as stretch goals — the math is supposed to be uncomfortable. If your team is hitting 100% every quarter, your OKRs are too easy and you've turned them into KPIs in disguise.
Side-by-side comparison
This is the table candidates draw on the whiteboard. Memorize the shape, not the words.
| Dimension | KPI | OKR |
|---|---|---|
| What it is | A metric with a target | An ambition with measurable outcomes |
| Horizon | Continuous, ongoing | One quarter (sometimes one year) |
| Ambition level | Realistic — 100% achievable | Stretch — 70% is a good score |
| Volume | Dozens per org | 3-5 Objectives, 2-5 KRs each |
| Focus | Health of what already works | Change you're trying to drive |
| Authoring | Top-down from leadership | Mix of bottom-up and top-down |
| Bonus linkage | Frequently tied | Usually decoupled on purpose |
| Failure mode | Stale dashboards no one reads | Wishlists with no measurable KRs |
The load-bearing distinction: KPIs answer "How are we doing?", OKRs answer "Where are we going?" KPIs maintain. OKRs change.
How an analyst works with both
The two artifacts demand different analytical muscles. KPIs are mostly a reporting and monitoring job. OKRs are mostly a modeling and forecasting job.
KPI work
For KPIs, an analyst maintains the dashboards, instruments the alerting, and decomposes the metric when it moves. When ARPU drops 8% week-over-week, you don't write a Slack message saying "ARPU is down" — you split it into average order value, purchase frequency, and paying-user share, and you come back with the segment that moved. You also keep an eye on freshness, null rates, and dimension parity so the KPI doesn't lie because of a broken pipeline.
OKR work
For OKRs, an analyst is the person who turns vague ambition into measurable Key Results. "Make onboarding feel obvious" is not a KR. "Lift Day 1 retention from 30% to 45%" is. You also compute the baseline — without it there is no honest target. Then you trek the weekly check-in: where each KR sits versus its trajectory, what experiments are in flight, what the confidence interval on the projected end-of-quarter value looks like.
A combined view for e-commerce
A reasonable quarter for a mid-stage e-commerce product:
OKR: Objective — make repeat purchase the default behavior. KR1: 90-day repurchase rate from 20% to 30%. KR2: orders per buyer from 1.3 to 1.8. KR3: loyalty program with 15% adoption among repeat buyers.
KPIs running underneath: DAU above 50,000; cart-to-checkout conversion above 45%; median delivery time below 2 days; NPS above +45.
The OKR says "this is the change we're driving." The KPIs say "and nothing essential breaks while we drive."
North Star vs OKR vs KPI
A third concept always shows up in this conversation, so handle it explicitly. The North Star Metric (NSM) is the single number that captures product value to the user, and it changes once every few years, not every quarter.
A clean mental model has three layers:
- North Star — one metric, multi-year, e.g. minutes of music listened per active user per month for Spotify.
- OKR — quarterly bets designed to move the North Star, e.g. "improve discovery recommendations" with KR "+15% discovery rate."
- KPI — operational metrics that keep the lights on, e.g. DAU, free-to-premium conversion, churn rate.
NSM is strategic. OKRs are tactical. KPIs are operational. If your OKRs aren't visibly moving the NSM over time, the OKRs are the wrong bets.
Common pitfalls
The first and most common failure is dressing a KPI up as an OKR. A team will write "OKR: DAU = 100,000" and call it a quarter. That's a KPI with a target. The OKR shape would be: Objective — become the daily-driver tool for product analysts; KR1 — DAU from 50K to 100K; KR2 — share of users with 3+ sessions per week from 20% to 40%. The objective gives meaning, the KRs give measurement, and the combination forces a discussion of why the number should move.
Another reliable failure is OKR sprawl. Ten Objectives with five Key Results each is fifty metrics. That's not focus, that's a metrics dictionary. The empirically useful range is 3 Objectives per team, 2-4 KRs per Objective. Anything more and people stop being able to recite them, and unrecited OKRs are dead OKRs.
A third trap is writing outputs as Key Results. "Ship the new landing page" is an output — it either happens or doesn't. "Landing page conversion from 3% to 6%" is an outcome. Outputs hide behind Jira tickets; outcomes show up in the warehouse. This is the single most useful edit an analyst can suggest during OKR drafting.
A fourth pitfall is wiring OKRs into bonuses. The moment compensation depends on OKR attainment, the team negotiates cushy KRs in week one and the stretch-goal premise collapses. Bonuses belong with KPIs, where 100% is the honest target. OKRs live in a no-blame zone where 70% is celebrated.
The last is the context-less KPI. "DAU = 48,000" is not analysis. Without the target, the prior-period comparison, the segmentation, and a sense of normal variance, it's a number on a slide. Every published KPI should ship with at least the target and the trend.
Interview questions and model answers
The OKR-vs-KPI prompt rarely shows up alone — it usually pulls two or three follow-ups. Practicing the chain matters more than memorizing the definitions.
Q: What's the difference between KPI and OKR? A KPI is a metric with a target tracked on an ongoing cadence — it shows the health of a process. An OKR is a quarterly objective with measurable Key Results — it sets the direction of change. KPI is the speedometer, OKR is the navigator. You use both: OKRs move the product, KPIs ensure nothing breaks while it moves.
Q: What KPIs would you propose for a subscription product? The core set is MRR, churn rate, LTV, CAC, the LTV-to-CAC ratio, trial-to-paid conversion, and Net Revenue Retention. The headline KPI is usually MRR or NRR; the rest exist for diagnosis when the headline moves. I'd also add a product-side guardrail KPI like weekly active users among paid subscribers, so monetization isn't decoupled from engagement.
Q: How would you write OKRs to improve retention? Objective: make the product feel essential to the user's weekly workflow. KR1: Day 7 retention from 20% to 35%. KR2: average sessions per week per active user from 2 to 4. KR3: share of returning users brought back by push notifications below 30%, so the growth is organic rather than nag-driven. The third KR is the guardrail — without it you could "win" by spamming.
Q: What if the KPI hits but the OKR doesn't? That means operations are healthy but the strategic change you bet on isn't happening. The investigation has three branches: were the initiatives chosen the right ones for that OKR, did the OKR receive enough resourcing relative to BAU work, and is the OKR fighting against entrenched processes that need to change first. Usually the answer is option two — the team ran the quarter on autopilot KPIs while the OKR starved.
Related reading
- North Star Metric for PM
- How to choose a North Star Metric
- Product metrics: DAU, MAU, ARPU, LTV explained
- Product manager case interview guide
If you want to drill product-sense and metric-design questions like this one daily, NAILDD is launching with hundreds of product analytics, SQL, and case-study problems built around exactly this pattern.
FAQ
Can a team run OKRs and KPIs at the same time?
Not only can — should. KPIs cover ongoing health (engagement, conversion, latency, support load); OKRs cover quarterly direction. A team that runs only KPIs becomes a maintenance org; a team that runs only OKRs forgets to keep the lights on. Healthy state: a small set of KPIs as background monitoring plus 3 OKRs as the active campaign.
Who owns the OKR — the PM or the analyst?
The product manager owns the OKR — accountable for the outcome and the pitch to leadership. The analyst is co-author and quality control: computing the baseline, proposing KRs that aren't outputs in disguise, and tracking weekly trajectory. Without an analyst at the drafting stage, OKRs become wishlists with vibes-based KRs.
How often should OKRs be reviewed?
The standard cadence is weekly check-ins, mid-quarter recalibration, end-of-quarter retrospective. Weekly is for trajectory and escalation. Mid-quarter is for course correction — kill an experiment, double down, reallocate engineering. End-of-quarter is for honest scoring and lessons. Rewriting OKRs mid-quarter is reserved for real pivots, not minor friction.
Should KPIs ever roll up into OKRs?
Yes, and it's how mature companies bridge the two systems. A KPI like Day 30 retention can become the source metric for a KR like "Day 30 retention from 18% to 25%". The KPI provides the measurement plumbing; the OKR provides the temporary push to move it. Once the quarter ends, the OKR retires but the KPI stays on the dashboard at the new baseline.
What's the difference between a Key Result and a metric?
A metric is anything you can measure. A KR is a metric with a starting value, a target, a deadline, and an owner. "Conversion rate" is a metric; "conversion from 4.1% to 6.5% by end of Q3, owned by Growth" is a KR — the difference between a chart and a commitment.
How are OKRs different at startups versus large companies?
At a 10-20 person startup, OKRs are usually company-wide and existential — runway, conversion, first $1M ARR — and everyone is on the same set. At a 5,000-person company like Notion or Stripe, OKRs cascade through company, product-area, and team layers with explicit alignment. The risk at scale is that the cascade becomes copy-paste, at which point the system stops doing useful work.