Outputs, outcomes and impact

Three words get mixed up in every planning meeting. Keep them apart and most goal problems go away.

• Output: what the team ships. "Offline receiving is live in all stores."
• Outcome: a change in what people do. "Staff receive deliveries in the app at the dock, even when Wi-Fi is down."
• Impact: the business result that follows. "Fewer app orders have an item cut at picking."

Output goals are tempting because the team fully controls them and they are easy to check. That is also their weakness. You can ship on time and change nothing, and the goal still says "done". Melissa Perri calls this the build trap: a team measured on what it ships keeps shipping, whether or not anything gets better.

Impact goals have the opposite problem. Cut orders depend on the Store app, but also on DC deliveries, staff numbers, the weather and a festival. A team can't steer by a number with that many causes, and it shows up months late.

Outcomes sit in the middle. The team can move them within weeks, and they are close enough to the impact that moving them should matter. Aim your key results at outcomes, and keep one eye on the impact they are meant to cause.

What an OKR is made of

An OKR has one objective and a few key results. The objective is a short, plain sentence that says what you want to be true. A store manager should understand it on first reading. The key results are the numbers that show the objective is true. John Doerr, who brought OKRs to Google in 1999 after learning them at Intel, sums up the format as "I will (objective) as measured by (these key results)".

Each key result needs five things written next to it:

• Baseline: the value today, measured, not guessed. If you can't measure it yet, your first job is to measure it.
• Target: the value you aim for.
• Date: usually the end of the quarter.
• Source: the query, report or sample that produces the number, so nobody argues about it in week 12.
• Owner: one person who reports on it every week.

Keep the set small. Christina Wodtke's Radical Focus recommends one objective per team per quarter with three key results. Two to four is a sensible range. More than that and the team is back to doing everything.

Decide how ambitious each key result is, and say so. Google's internal OKR guidance separates committed OKRs, which the team agrees will be met and which are expected to score 1.0, from aspirational ones, which are expected to average around 0.7. Wodtke suggests setting key results where you feel about a 5 in 10 chance of hitting them. Either way, the label matters: a 0.7 on an aspirational key result is a good quarter, and a 0.7 on a committed one needs an explanation.

Run two tests on every set. The sufficiency test: if we hit every key result, would the objective really be true? The task test: is any key result a task in disguise? "Ship offline receiving" is a task. It belongs on the roadmap (pm-13), not in the OKRs.

Leading and lagging results

A lagging indicator tells you whether you succeeded, but late: stock accuracy measured by counts, cut orders, write-offs. A leading indicator moves early and predicts the lagging one: the share of deliveries received in the app at the dock this week.

A good set has both. At least one key result should move in the first few weeks, so you can tell whether the work is landing while there is still time to change course. At least one should be close to the impact, so you can't declare victory on activity alone.

A leading indicator is only useful if it really predicts the lagging one. Check before you commit. Imran asks Tomás to compare stores from last quarter: stores that received most deliveries in the app had stock accuracy about 7 points higher than stores that received mostly on paper. That is a correlation, not proof, but it is enough to bet a quarter on, and the lagging key result will confirm it or not.

Guardrails: what must not get worse

Any number you push hard enough can be hit the wrong way. Economists call this Goodhart's law, usually quoted as "when a measure becomes a target, it ceases to be a good measure". If the key result is "receiving takes under 12 minutes", staff will accept deliveries without counting the cases.

A guardrail metric is a number that must stay inside a limit while you push the key results, a line you promise not to cross. Wodtke calls these health metrics and suggests two to five, each marked green, yellow or red at the weekly check-in.

Pick guardrails by asking "how could we hit this key result and make things worse?". Common pairs:

• Speed and accuracy: faster receiving, but wrong quantities accepted.
• Adoption and satisfaction: more stores using the app because head office ordered it, while staff hate it and keep a paper copy.
• Reliability and cost: more on-time "Now" deliveries by adding riders, while the cost per order climbs. You will see this exact trade-off in the metric tree in North star, input and guardrail metrics.
• New work and old quality: features shipped fast while crash-free sessions drop.

Goals for products that earn no revenue

The Store app sells nothing. Neither do most tools for warehouses, riders, field agents and back offices. Their value still shows up in numbers, if you know where to look:

• Time: minutes per task × tasks per day × people × the hourly cost of their time.
• Errors: wrong quantities, mispriced items, lost paperwork, each with a cost.
• Money at risk: write-offs, fines, lost sales from cut orders, chargebacks.
• Workarounds removed: paper sheets, spreadsheets and chat-group requests that no longer exist.
• Cycle time: how long a process takes end to end, such as truck arrival to stock on the shelf.

Be careful with adoption. Internal users often have no choice, so "100% of stores use the app" can be met by an order from head office while staff keep a paper copy on the side. Measure the workaround that should disappear instead: "paper GRN sheets ordered from the print room fall from 400 pads a month to under 50".

Convert to money when you can, because it lets you compare an internal tool's goals with anything else the company could spend on, and you will need it for a business case (Business cases). Say how saved time will be used, too. Minutes saved are only worth money if they go somewhere useful, such as filling shelves before the evening rush.

Valuing a 3-point drop in cut orders (Mangrove estimates)
App orders per day            60,000
Cut-order rate                9% -> 6%   (3 points)
Fewer orders with a cut item  60,000 x 0.03   = 1,800 a day
Lost sale per cut item        about $3
Sales kept                    1,800 x $3      = $5,400 a day
Per year                      $5,400 x 365    = about $2.0 million
(Before any effect on customers who stop ordering after a bad experience.)