Three words people muddle
Last session we established the trap: any rewarded number gets gamed, so measurement has to be designed, not just adopted. This session gives you the vocabulary to design it. An OKR is a goal you are trying to change this quarter. A KPI is a vital sign you keep healthy all the time. A metric is any number you can track - the raw material both are built from. OKRs measure change; KPIs measure health; and the health metrics you refuse to let slip become your guardrails. Get these three straight and half the confusion in a planning meeting disappears. We keep last session's rail in force throughout: an OKR score is for steering and learning, never a direct input to pay.
Objective + Key Results, scored 0 to 1 8 min live
An OKR has two parts. The Objective is qualitative and aspirational - a memorable sentence about where you want to be. The Key Results are the two-to-three measurable outcomes that prove you got there. If you cannot put a number on a Key Result, it is not a Key Result. The framework traces to Andy Grove at Intel and was carried to Google by John Doerr, whose "Measure What Matters" is the standard reference.
LiveCommitted vs stretch, and the 0-1 score3 min▶
Two kinds of goal live under the same framework, and confusing them is how OKRs turn toxic:
- Committed - you are expected to fully deliver these. A score below 1.0 is a real miss and needs a conversation. These are the promises.
- Aspirational / stretch - deliberately ambitious "moonshots". At Google, landing a stretch goal at roughly 0.7 is considered success; a consistent 1.0 means you set the bar too low. Treat ~0.7 as a Google-specific norm to adapt to your context, not a universal law.
The score exists to spark a conversation - what did we learn, where do we adjust - not to rank people. Wire the 0-1 score to a bonus and you kill the willingness to set ambitious stretch goals overnight, because now a miss costs money. That is exactly the rail we set in a1, and it holds here.
Why Google keeps scores off comp. Google re:Work is explicit that OKRs work precisely because they are decoupled from compensation: people set genuinely hard goals only when missing one is safe. The instant the score touches pay, everyone sandbags to protect themselves - and you lose the ambition the framework was built to unlock.
OKRs measure change; KPIs measure health 7 min live
Here is the cleanest way to tell an OKR from a KPI. An OKR is what you are trying to change this quarter - it has a start, a target, and an end. A KPI is a vital sign you keep healthy forever - conversion rate, gross margin, uptime. The KPIs you promise not to let slip while you chase a goal become your guardrails. And any top-line number decomposes into a tree of levers - which is how you give every team something real to own.
Self-studyWhy the tree is the fairness engine, not just a diagram2 min read▶
The metric tree does more than organise numbers - it solves an fairness problem you cannot solve any other way:
- It assigns ownership. Revenue is shared by four teams, so it is nobody's fair individual grade. But "checkout completion rate" has a clear owner in Product, and "cost per acquired customer" has one in Growth. Grade the leaf; watch the root together.
- It exposes the attribution problem early. When two teams both move conversion, the tree shows it - and forces the honest conversation about shared credit instead of pretending one team did it all. We take this on directly in a3.
- It keeps guardrails visible. Chasing "Sessions" with cheap paid traffic will quietly wreck "LTV:CAC". Because both sit on the same tree, the trade-off is impossible to hide.
Revenue = Sessions × Conversion Rate × Average Order Value is the canonical e-commerce identity; the LTV:CAC ≥ 3:1 rule of thumb is an illustrative benchmark, not a guarantee. Treat the exact ratios as starting points to calibrate for Cartwheel, not gospel.
Is it an OKR, a KPI, or a guardrail? ★ 10 min · think it through
A thought exercise. Take Cartwheel's numbers and sort each into the right tool. The test is simple: are you trying to change it this quarter, keep it healthy forever, or protect it while you chase something else?
Put four Cartwheel numbers on the table: checkout conversion, gross margin %, repeat-purchase rate, and site uptime.
Ask of each: are we deliberately moving this from a start to a target this quarter (OKR), or holding it steady as a vital sign (KPI)?
Name the guardrails: which healthy KPIs must not slip while we chase the OKR? Margin and uptime protect the conversion push.
Write the pairing out loud: "This quarter we change conversion (OKR), while we hold margin and uptime (guardrail KPIs)."
| Cartwheel number | Tool | Why |
|---|---|---|
| Checkout conversion 2.4% → 3.0% | OKR (change) | a deliberate, time-boxed lift with a target |
| Gross margin % | KPI / guardrail (health) | kept healthy always; protects the conversion push |
| Repeat-purchase rate 22% → 35% | OKR (stretch) | a bold change this quarter, scored ~0.7 as success |
| Site uptime 99.9% | KPI / guardrail (health) | a vital sign, not a quarterly goal |
Give every Cartwheel team a leaf to own ★ 8 min · think it through
Walk the metric tree from the top and hand each branch to the team that genuinely controls it. This is the move that turns one shared revenue goal into fair, owned OKRs.
Start at the root: Revenue. Acknowledge out loud that no single team owns it - four teams move it together.
Walk down one level: Sessions → Growth, Conversion → Product + Engineering, Average Order Value → Merchandising.
Give each team a Key Result on its own controllable leaf, not on Revenue: Growth owns cost-per-acquired-customer, Product owns checkout completion, Merchandising owns items-per-order.
Keep Revenue and LTV:CAC as the shared numbers everyone watches together in the monthly review - watched, not personally graded.
The attribution trap you just avoided. If Cartwheel had graded all four teams directly on Revenue, a great quarter for Growth could mask a broken checkout, and a heroic Engineering fix could be invisible under a soft-market dip. Grading the controllable leaf makes each team's contribution legible and fair; watching the shared root together keeps everyone rowing the same direction. a3 makes "grade the lever, watch the outcome" the explicit rule.
Try it yourself - this week ◐ 25-40 min total
- Write one draft Objective for your own team - a single qualitative, aspirational sentence - plus two Key Results that are genuinely measurable. Label each Key Result committed or stretch.
- Sanity-check your Key Results: could you put a number on each today? If not, it is not a Key Result yet - rewrite it.
- Take one top-line number your org cares about and decompose it one level into its metric tree. Note who owns each branch.
- Ask of your draft: if this OKR score were wired to bonuses, would you still set the stretch goal that ambitiously? Sit with the answer - it is the reason scores stay off pay.
- Optional: skim Google re:Work "Set goals with OKRs" and the KPIs-vs-OKRs note on whatmatters.com before a3.
Sources covered (cited)
Taught from the OKR and e-commerce-metrics canon. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · What is the cleanest distinction between an OKR and a KPI?
OKRs have a start, a target, and an end - they measure change. KPIs are vital signs you keep healthy over time. The KPIs you protect while chasing an OKR become your guardrails.
2 · At Google, landing a stretch OKR at a score of about 0.7 is treated as...
On stretch goals, ~0.7 signals healthy ambition; a consistent 1.0 means the bar was set too low. It is a Google norm to calibrate for your own context - and the score never sets pay.
3 · In the Cartwheel metric tree, why is Revenue a poor individual grade?
Revenue = Sessions × Conversion × AOV, and those branches belong to different teams. Grade each team on its controllable leaf; watch the shared root (and LTV:CAC) together.