Perplz
로그인

Mission & task writing guide

A mission is the team's goal. A task is a result one person owns.

Perplz missions and tasks are built on the OKR shape (objective → key results), but teams that run on KPIs or MBO can use the same fields as they are. It's not that one is better — they fit different moments. Follow along and your goals will tell you how far you've come, even on your first try.

Which method fits?

There are several ways to set goals. They don't replace each other — each answers a different question. Teams pick by size, field and purpose, and the same team switches as it changes.

  • OKR

    when something needs to change

    Question it answers
    What will we change this period?
    Fits when
    You have a hypothesis to test or a direction to shift
    Missions are
    A state you'll reach (objective)
    Tasks are
    Numbers that show you arrived (key results)
    Check-in rhythm
    Weekly or biweekly checks, retro per quarter or project
    Scoring
    Average key-result progress — stretch goals may fall short by design
    When it struggles
    Tie scores straight to reviews and only safe goals get written
  • KPI

    when you keep what works

    Question it answers
    Are we running within the normal range?
    Fits when
    You're keeping something that already runs
    Missions are
    A metric to keep and its target level
    Tasks are
    Actions that keep or lift the metric
    Check-in rhythm
    Record weekly, check the range monthly
    Scoring
    Actual vs. target, or share of periods kept in range
    When it struggles
    Hitting the number can become the goal itself
  • MBO

    when each person commits

    Question it answers
    What has each person agreed to deliver?
    Fits when
    Roles are clear and personal commitments must be recorded
    Missions are
    A team goal agreed with the lead
    Tasks are
    An individual goal one person agreed to
    Check-in rhythm
    Agree at the start, check midway, review at the end
    Scoring
    Achievement vs. agreed standard × weight
    When it struggles
    Agreement turns into instruction, or early goals miss later changes
  • Perplz

    for AI-native teams

    Question it answers
    What do we check this month, this week and today?
    Fits when
    You work with AI and output lands daily
    Missions are
    A state at month's end (+ to-dos = a day)
    Tasks are
    A result confirmed by a number within the week
    Check-in rhythm
    One blocker daily, adjust weekly, retro monthly
    Scoring
    Average weekly-task progress + a direction retro at month's end
    When it struggles
    AI output volume gets mistaken for results

Pick with four questions

The four questions aren't a fork in the road — they stack. Add a method for each “yes”.

  1. 1

    Is there work that already repeats every week or month?

    If so, set lines to hold with KPIs (3 or fewer) · KPI

  2. 2

    Is there something to test or change this period?

    If so, set one OKR objective with key results · OKR

  3. 3

    Must personal commitments be recorded for review?

    If so, agree on individual goals with MBO · MBO

  4. 4

    Does AI output land every day while you revisit the plan once a quarter?

    If so, split it into a month, a week and a day with the Perplz method · Perplz

Four “no”s may mean you don't need a goal system yet. Writing this month's single most important thing as a mission is enough. Two or more “yes”es? Each mission can follow its own method — changes follow the OKR tab, lines to hold follow the KPI tab.

Common combinations by size, field and purpose

There's no single right answer. The same team changes its mix over time.

  • A 1–3 person side projectSet one OKR objective, or just “this month's most important thing”
  • A study group or clubHold consistency with KPIs (attendance, problems solved) and add an OKR when something needs to change
  • A contest or hackathonOKR — the period is short and the change is clear
  • A class project or capstoneOKR for the team's output, MBO for each person's contribution
  • An internshipMBO — agree on individual goals with your mentor and review together
  • A launched productOKR + KPI side by side, or a North Star metric with input metrics
  • A 5–20 person team3 or fewer KPIs as lines to hold, one OKR for what to change this quarter
  • An organization of 100+Teams run KPIs and OKRs; individual reviews often use MBO
  • A team building daily with AI agents and generatorsPerplz — monthly missions, weekly tasks, daily to-dos. Add KPIs when a metric needs holding

What many companies call “KPI reviews” is usually MBO — reviewing the goals each person committed to. Choose by the question a method answers, not by its name.

Writing by method

No method is ahead of the others. Pick by where your team is right now. Your choice is remembered in this browser, and the quick guide in project rooms opens with the same method. The Project Agent follows the goal method set in project info.

Objectives and Key Results

An OKR pairs where you want to go (the objective) with how you'll know you got there (the key results). The Perplz mission tracker, which groups tasks under each mission, was designed in this shape.

Question it answers
What will we change this period?
Fits when
  • You have a hypothesis to test or a direction to shift
  • The change to make is clear, as in contests, hackathons or new features
  • The whole team looks at one goal and each person picks what to add
Check-in rhythm
A 10-minute check every week or two; score and retro when the project or quarter ends

In Perplz terms

Example · plant watering app

  • Project introPurposeWhy are we doing this project?Help people who live alone keep their plants alive
  • 2–3 missionsObjectiveWhen we're done, who has changed and how?Plant owners never miss a watering day
  • Tasks under a missionKey resultHow do we know we got there? — number, owner, due dateAll 10 test plants get a watering alert (engineer)
  • To-dos under a taskInitiativeWhat do we do today to move a key result?Register push tokens on 3 test phones and receive the first alert

Writing missions (objectives)

A mission says, in one sentence, who will have changed and how once you're done. If you write a feature name or a to-do, you can finish everything and still not know what got better.

  1. BeforeTo-do

    Build plant alert feature

    You can ship it and still not know who is better off

  2. Try thisState the result creates

    Plant owners never miss a watering day

    Anyone can tell what's true once you're done

  3. StretchChange in people

    People who kept killing plants water on time

    It names what you really want to change

“Try this” is enough. If real people will use what you make, go for “Stretch”.

Formula

[Who] + [how they change]

The subject is a person — users, the team, judges. The sentence describes a state, not an action you take.

3-second check

  • Is it a state you'll reach, not a to-do?
  • Is a person the subject?
  • Does it make sense without a number? (Numbers go in tasks.)
  • Does it make the team want to go for it?
  1. 1

    Two or three missions, in order

    Try to do everything at once and nothing gets finished. Split the project into stops on the way: first, second, third.

  2. 2

    Numbers go in tasks, not missions

    “Reach 100 sign-ups” is a task, not a mission. The mission says why you need those 100 people.

  3. 3

    Make the first mission checkable within 2 weeks

    People given several near goals got further than those given one distant goal. An early win sets the team's pace.

    Bandura & Schunk (1981)

  4. 4

    Add one line of why

    Write why this mission matters in the project intro and the team runs in the same direction. Goals handed down without a reason performed clearly worse.

    Latham, Erez & Locke (1988)

  5. 5

    New to it? Make it a learning mission

    Instead of “Launch a Flutter app”, write “I can build screens in Flutter on my own”. Chasing result numbers on unfamiliar work gets in the way of learning.

    Seijts & Latham (2005)

Writing tasks (key results)

A task is how you check you've reached the mission. With a number in it, the due date tells you clearly: done or not done.

  1. BeforeActivity

    Work on alerts

    Nobody agrees on how much is enough

  2. Try thisOutput + number

    All 10 test plants get a watering alert

    Count how many of the 10 worked and you know

  3. StretchMetric from → to

    Testers' missed waterings: 3/week → 0

    It shows whether the outcome really changed

Formula

[What / who] + [number] + [done by the due date]

Put the step-by-step work in the Project Agent's steps or the feed. Keep only the result in the task title.

Marks of a good task

  • It has a number (count, people, %) or a from → to
  • On the due date it's clearly done or not done
  • One owner (helpers are welcome)
  • It fits in 1–2 weeks (split it if bigger)
  • It leads to the mission — “If this is done, are we closer?”
  1. 1

    Two to five tasks per mission

    More than that? Split the mission in two, or move small work into the steps inside a task.

  2. 2

    If a number can be gamed, pair it

    “Post 10 blog posts” can be filled with rushed posts. “Post 6 blog posts, each reviewed by a teammate” protects quality. Pair speed with quality, and quantity with reuse.

    Grove, High Output Management (1983)

  3. 3

    Everyone writes their own tasks

    When a leader writes everyone's tasks, gaps and overlaps get hidden. Teams where each person looked at the mission and wrote the number they'd add did better. For work you need from someone else, send a task request.

    Kleingeld, van Mierlo & Arends (2011)

Rewrite it like this

  • BeforeDesign

    Try thisPick 1 of 2 main-screen drafts by vote

  • BeforeBackend work

    Try this3 plant APIs pass all 10 tests

  • BeforeStudy

    Try thisFinish 5 React tutorial chapters, post 1 recap

  • BeforeMeeting prep

    Try thisGet 10 interview questions reviewed by 3

  • BeforeBug fixes

    Try thisAll 3 reported login bugs no longer reproduce

  • BeforePromotion

    StretchGrow survey replies 0 → 30 with 3 card posts

Breaking one mission into tasks

MissionPlant owners never miss a watering day

  • PlannerValidate 3 alert-time rules with 5 testers
  • DesignerPick 1 of 2 plant-signup drafts by vote
  • EngineerAll 10 test plants get a watering alert
  • WildcardInterview 5 plant owners, find 3 pain points

Writing to-dos (initiatives)

An initiative is what you do today to move a key result's number. The key result is the outcome; the initiative is the work that produces it — if you finish the work and the number doesn't move, change the initiative.

  1. BeforeA vague chunk of work

    Fix up the alert feature

    Tonight nobody can say whether it happened

  2. Try thisToday's action that moves a key result

    Register push tokens on 3 test phones and receive the first alert

    It finishes today and moves one step toward the key result

  3. StretchThe action + a handoff

    Find why 2 plants got no alert, note it in the feed and hand it to the planner

    A day's result becomes the next person's action

Formula

[today's action] that moves [one key result]

Checking it moves that task (key result) to “in progress”. The key result's number doesn't rise on its own — record it at the weekly check

Marks of a good to-do

  • You know which key result it belongs to
  • It is clearly done or not done by tonight
  • An action is the subject — results and numbers live in the key result
  • When it's done, the number moves or you learn why it doesn't
  1. 1

    An initiative is not a key result

    If you do the work and the key result doesn't move, swap the initiative. Don't mix what was made (output) with what changed (outcome).

    Perdoo, The Ultimate OKR Guide — Initiatives

  2. 2

    1–3 a day, re-picked at the weekly check

    Initiatives are the execution layer you re-choose every week. If the number isn't moving, change next week's initiatives and keep the key result.

    Doerr, Measure What Matters (2018)

  3. 3

    A check is not achievement

    Even with every initiative checked, mission achievement is computed from key-result numbers only. Don't count checks as results.

Rewrite it like this

  • BeforePush setup

    Try thisSend a test message from the FCM console and receive it on a phone

  • BeforeContact testers

    Try thisSend the install link to 5 testers and get their replies

  • BeforeLook at bugs

    Try thisSkim yesterday's missed-alert logs and note 2 likely causes

  • BeforeDesign meeting

    Try thisPost 2 sign-up screen drafts in team chat and open a vote

For a project like

Pick the type closest to your project and follow along. The three missions are stops on the way, in order.

Building an app or service

Example · Plant watering reminder app

Mission

  • BeforePlan and wireframe screens

    Try thisThe whole team builds from one screen flow

  • BeforeBuild the alert feature

    Try thisPlant owners never miss a watering day

  • BeforeLaunch the app

    StretchPeople who kept killing plants water on time

Task

  • Try thisCore 5-screen flow approved by all 4 members
  • Try thisAll 10 test plants get a watering alert
  • Try thisRecruit test users from 0 to 10
  • Try this7 of 10 testers log watering after 2 weeks

Contests & hackathons

Example · Public data contest

Mission

  • BeforePlan the idea

    Try thisJudges get our idea from a single page

  • BeforeBuild the MVP

    Try thisThe core features run smoothly before judges

  • BeforePrepare the pitch

    StretchAnyone on the team can answer any question

Task

  • Try thisFinalize a 1-page plan covering all 4 criteria
  • Try thisRun 3 core features end to end in a 3-min demo
  • Try thisPrepare 10 Q&A answers and rehearse twice
  • Try thisFind 3 things 5 past winners have in common

Study groups

Example · 8-week coding test study

Mission

  • BeforeStudy algorithms

    Try thisEach week we solve and learn from each other

  • BeforeSummarize concepts

    Try thisWe can explain data structures to others

  • BeforePrepare for coding tests

    StretchNobody on the team fears coding tests

Task

  • Try thisSolve 5 problems weekly for 4 weeks and post
  • Try this6 data structures, one 10-min talk each
  • Try thisSolve 2 of 3 problems in the mock test
  • Try thisRe-solve 5 missed problems and compare notes

Class projects & capstones

Example · Major course team project

Mission

  • BeforeChoose a topic

    Try thisThe professor agrees our topic matters

  • BeforePrepare the midterm talk

    Try thisThe audience gets the problem and fix at once

  • BeforeWrite the final report

    StretchGraders find nothing missing in our report

Task

  • Try thisSubmit a proposal reflecting all 3 comments
  • Try thisBuild a 10-min talk: problem, fix, 3 reasons
  • Try thisFill all 8 items on the grading rubric
  • Try thisSummarize 3 similar cases on 1 page

Research & data analysis

Example · Students' breakfast habits survey

Mission

  • BeforeRun a survey

    Try thisWe learn students' real breakfast habits

  • BeforeAnalyze the data

    Try thisWe find patterns we didn't know about

  • BeforeWrite the report

    StretchReaders rethink their own habits

Task

  • Try thisCollect survey responses from 0 to 100
  • Try thisChart 3 patterns found in the answers
  • Try thisPilot 15 survey questions with 2 people
  • Try thisHear “I learned something” from 4 of 5 readers

Events & communities

Example · Developer networking meetup

Mission

  • BeforeBook a venue

    Try thisGuests get a comfy place to come and talk

  • BeforeRecruit attendees

    Try thisThe right people sign up for our event

  • BeforeRun the event

    StretchFirst-timers look forward to the next meetup

Task

  • Try thisLock a 30-person venue 3 weeks ahead
  • Try thisGrow sign-ups from 0 to 40
  • Try thisGet 30 attendance confirmations by the eve
  • Try thisHear “I'll come again” from 80% of guests

Ongoing operations

Example · YouTube channel or club

Mission

  • BeforeRun the YouTube channel

    Try thisSubscribers look forward to each weekly video

  • BeforeGrow views

    StretchViewers watch our videos to the end

  • BeforeHold retros

    Try thisThe team finds a way to improve every week

Task

  • Try thisPost 1 video a week, 4 in 4 weeks
  • Try thisRaise average watch time from 30s to 1 min
  • Try thisCompare click rates of 2 thumbnail options
  • Try thisAct on 1 change from each of 4 retros

Task examples by role

Under the same mission, each role adds a different number. Pick your role's examples and change the numbers.

Planner

  • BeforeWrite a plan

    Try thisGet sign-off on a 1-page plan of 3 features

  • BeforeMarket research

    Try thisCompare 3 similar services, find 2 edges

  • BeforeUser research

    Try thisInterview 5 target users, list 3 pain points

  • BeforeManage schedule

    Try thisFill owner and due date for all 10 tasks

Designer

  • BeforeUI design

    Try thisPick 1 of 2 main-screen drafts by team vote

  • BeforeDesign system

    Try thisOrganize 10 components like buttons, inputs

  • BeforeLogo

    Try thisFinalize 1 of 3 logo drafts with feedback

  • BeforeDesign fixes

    Try thisApply all 5 pieces of test feedback

Engineer

  • BeforeBuild login

    Try thisAll 5 testers log in with email

  • BeforeDesign the DB

    Try thisPass the table design review by 2 teammates

  • BeforeFix bugs

    Try thisMake all 3 reported bugs non-reproducible

  • BeforeDeploy

    Try thisAll 4 teammates open the test server link

Wildcard

  • BeforePromotion

    Try thisBring 30 survey replies via 3 card posts

  • BeforeCommunity care

    Try thisSend welcome notes to all 10 newcomers

  • BeforeTesting

    Try thisFind 5 pain points from 5 friend testers

  • BeforeOrganize notes

    Try thisPost a 3-line decision summary same day

Scoring

With OKR, a mission's achievement is the average progress of its tasks (key results)

  • Mission achievement = average task progress
  • Task progress = (now − start) ÷ (target − start); a checked-off task counts as 100%

Example · Plant owners never miss a watering day

  • 8 of 10 test plants got alerts80%
  • Signup screen draft picked (done)100%
  • 3 of 5 friend interviews60%

Average 80% → Partial

  1. 1

    Committed vs. stretch objectives

    Commit to 100% on must-have objectives. For stretch objectives, agree up front on what counts as success, like 70%. If you went past 130%, talk in the retro about whether the target was too low.

  2. 2

    Scores don't go straight into pay formulas

    Tie OKR scores directly to reviews and everyone writes easy goals. Use scores to decide what help comes next.

    Google re:Work, Set goals with OKRs

When it struggles

  • The mission is a feature nameRewrite it as “when we're done, who changes and how”
  • Numbers crept into the mission, so it's a taskMove the number into a task and keep the mission a state
  • Only easy key results get written to lift scoresCheck whether scores feed straight into reviews or pay, and separate them
  • The kickoff doc is never looked at againPut check-in dates on the calendar at the start

Sources · Doerr, Measure What Matters (2018) · Google re:Work, Set goals with OKRs · Grove, High Output Management (1983)

Key Performance Indicators

KPIs watch the health of work that already runs, in numbers. Instead of shifting direction, you set lines to hold and levels to raise, and check them steadily. They shine where consistency is the result.

Question it answers
Are we running within the normal range?
Fits when
  • Work repeats every week or month (channels, clubs, communities)
  • You must keep today's quality, as with a launched service
  • Results only show once they accumulate
Check-in rhythm
Record values weekly and check the range monthly. Once a quarter, ask whether the metrics themselves still fit

In Perplz terms

Example · running a dev club

  • Project introWhat to keep and growWhat do we keep steady and grow?New members keep coming all semester
  • 2–3 missionsKey metric + target levelWhich number do we hold, and at what level?Keep weekly meetup attendance at 80%+
  • Tasks under a missionActions that hold the metricWho does what to hold this number?Send the meetup reminder all 4 times in 4 weeks (organizers)
  • To-dos under a taskActivityWhat action or record moves the metric today?Post tomorrow's meetup reminder in the group chat and note how many read it

Writing missions (key metrics)

With KPIs, a mission is one number to keep and its target level. One line should show what you count and what counts as normal.

  1. BeforeTo-do

    Run the club

    There's no way to tell what normal looks like

  2. Try thisMetric + target level

    Keep weekly meetup attendance at 80%+

    Count the same way each week and you see if it's normal

  3. StretchTarget level + alert line

    Attendance 80%+; below 70%, find the cause

    You also decide what to do when it drops

“Try this” is enough. If the number swings often, add an alert line too.

Formula

[Metric] + [target level or normal range] + [period]

A metric must be countable the same way every time. An alert line that tells you it dropped makes it even better.

3-second check

  • Can you count it the same way every time?
  • Is it a number your team can move?
  • Are the normal range and alert line set?
  • Is there a way to hit the number but miss the point?
  1. 1

    Three key metrics or fewer

    Too many metrics and nobody looks. Put only the three most important health signals on the dashboard.

  2. 2

    Pair a slow metric with a fast one

    Numbers like revenue or churn move only after results settle. Watch one that moves first, like the reminder response rate, alongside.

  3. 3

    Set the normal range and alert line together

    Deciding up front — “80%+ is normal, below 70% means find the cause” — spares you a debate every time you look.

  4. 4

    Agree on how to count first

    If who counts, when and where differs by person, the number wobbles. In week one, write the counting rule in one line in the feed.

  5. 5

    Pair metrics so the metric doesn't become the goal

    When a measure becomes a target, it stops being a good measure. Measuring response speed? Watch answer quality too. Counting uploads? Watch watch time too.

    Strathern (1997) · Goodhart's law

Writing tasks (actions that hold the metric)

With KPIs, a task is an action that keeps or lifts the metric. Bundle repeating work into one task with a period.

  1. BeforeActivity

    Attendance management

    Nobody knows how much is enough

  2. Try thisAction + frequency + period

    Send the meetup reminder all 4 times in 4 weeks

    Count whether all 4 went out and you know

  3. StretchLeading metric from → to

    Raise RSVP replies from 50% to 90%

    It lifts a number that moves before attendance

Formula

[Action] + [frequency or standard] + [period]

Give repeating actions a period, like “weekly for 4 weeks”. Write who counts the metric and when in the task's steps.

Marks of a good task

  • It has a number or frequency
  • At the end of the period it's clearly done or not done
  • One owner (helpers are welcome)
  • If it repeats, it has a period (usually 2–4 weeks)
  • It's clear which metric it protects
  1. 1

    Crossed the alert line? Add a task to find the cause

    In a week the metric drops, add a task like “Ask the 3 people who missed why”. Repeating only the upkeep actions hides why it fell.

  2. 2

    One person owns one metric

    With one owner per metric, nobody wonders who brings the number each week. Helpers are welcome.

  3. 3

    If a number can be gamed, pair it

    “Post 1 video a week” can be filled with rushed videos. “1 video a week, average watch 1 min+” protects quality.

    Grove, High Output Management (1983)

Rewrite it like this

  • BeforeAttendance management

    Try thisSend the meetup reminder all 4 times in 4 weeks

  • BeforeCS response

    Try thisAnswer 90% of inquiries within 24h for 4 weeks

  • BeforeContent uploads

    Try thisPost 1 video a week, 4 in 4 weeks

  • BeforeServer management

    Try thisKeep downtime under 1 hour this month

  • BeforeSocial media

    Try thisReply to every comment within 24h for 4 weeks

  • BeforePromotion

    StretchGrow weekly sign-ups from 10 to 20

Breaking one mission into tasks

MissionKeep weekly meetup attendance at 80%+

  • PlannerShare the agenda the day before, 4 of 4 weeks
  • DesignerMake 4 meetup cards and post one each week
  • EngineerBuild a check-in form that tallies 4 weeks
  • WildcardMessage every member who missed 2+ meetups

Writing to-dos (activities)

An activity is what you do today to hold the metric, plus its record. Outcome metrics like attendance move slowly, so today you write the leading activity (reminders, outreach) that produces the number.

  1. BeforeA metric's name

    Attendance management

    Nothing shows what you did today

  2. Try thisLeading activity + record

    Post tomorrow's meetup reminder in the group chat and note how many read it

    Action and record stay together and roll into Friday's number

  3. StretchAct before it drops

    Check in on 2 members who missed last week and record whether they replied

    You move before the metric falls

Formula

[today's action that moves the metric] + [the record to leave]

One line in the feed is enough as a record. Friday's weekly numbers gather here

Marks of a good to-do

  • You know which metric it moves
  • It finishes today and leaves a record
  • For a recurring activity, it is today's share only
  • It is an action, not a numeric target — the target level lives in the mission
  1. 1

    Write the leading activity

    Outcome metrics (attendance, churn) move late. Put the activities that drive them (reminders, outreach, guidance) in your to-dos and you can act before Friday.

    Kaplan & Norton (1992), The Balanced Scorecard — Measures That Drive Performance, HBR

  2. 2

    An activity comes with a record

    Objectives → measures → targets → initiatives: the activity is the last cell. Without a measure, the activity never comes back as a metric — write who counts and when.

    Kaplan & Norton (1996), The Balanced Scorecard

  3. 3

    If activities pile up and the metric doesn't move, change the activity

    Swap the activity instead of lowering the target. If 4 weeks of the same activity leave the number unchanged, it wasn't a leading indicator.

Rewrite it like this

  • BeforeAnnouncement

    Try thisPost this week's meetup place and time in the notice channel

  • BeforeNewcomer care

    Try thisSet up a coffee chat with 1 new member

  • BeforeRecord

    Try thisEnter today's attendance in the sheet and leave one line in the feed

  • BeforeContent

    Try thisPick 1 activity photo and post it to the feed

For a project like

KPIs shine where consistency is the result. Pick a similar situation and change the numbers.

Running a content channel

Example · YouTube or blog channel

Mission

  • BeforeRun the channel

    Try thisAt least 1 upload every week this month

  • BeforeManage views

    Try thisKeep average watch time at 1 min or more

  • BeforeManage subscribers

    StretchUnsubscribes ≤ 2%/month; above 3%, find why

Task

  • Try thisPost 1 video by each Friday, 4 in 4 weeks
  • Try thisAdd a 30-second summary to all 4 videos
  • Try thisReply to every comment within 24h for 4 weeks
  • Try thisLog 3 metrics in the feed every Monday

Running a club or community

Example · Dev club

Mission

  • BeforeRun the club

    Try thisKeep weekly meetup attendance at 80%+

  • BeforeManage newcomers

    Try thisLose no more than 2 new members a semester

  • BeforeTrack activities

    Stretch1+ activity log a week; make up missed weeks

Task

  • Try thisSend the meetup reminder all 4 times in 4 weeks
  • Try thisMessage every member who missed 2+ meetups
  • Try thisHave 1:1 coffee chats with 5 new members
  • Try thisPost 1 activity photo in the feed each week

Operating a launched service

Example · Plant watering app (after launch)

Mission

  • BeforeOperate the service

    Try thisKeep notification delivery at 99%+

  • BeforeHandle customer inquiries

    Try thisAnswer 90% of inquiries within 24h

  • BeforeManage users

    StretchChurn ≤ 5%/month; above 7%, interview leavers

Task

  • Try thisReview failed-alert logs once a week for 4 weeks
  • Try thisWrite 10 answer templates and share them
  • Try thisAsk 5 users who left why they left
  • Try thisPost the weekly metric report 4 times

A steady study group

Example · Algorithm study

Mission

  • BeforeStudy algorithms

    Try thisEveryone solves 5 problems every week

  • BeforeRun presentations

    Try this1 talk every week, 0 weeks skipped

  • BeforeManage the study group

    StretchAttendance 90%+; adjust after 2 missed weeks

Task

  • Try thisSolve 5 problems weekly for 4 weeks and post
  • Try thisShare the talk order every Monday for 4 weeks
  • Try thisPick 3 mock-test problems every other week
  • Try thisAsk members who missed 2 weeks about timing

Customer support on a team

Example · CS part of a team workspace

Mission

  • BeforeHandle support tickets

    Try this90% first replies within 1 hour

  • BeforeManage satisfaction

    Try thisKeep support rating at 4.5 or higher

  • BeforeHandle repeat questions

    StretchKeep repeat-inquiry rate at 20% or lower

Task

  • Try thisAnswer all overnight inquiries by 10 am, 4 weeks
  • Try thisPublish 10 FAQs as help articles
  • Try thisCollect 30 support-rating responses in 4 weeks
  • Try thisSort 5 repeat inquiries by cause each week

Task examples by role

Under the same metric, each role owns different actions.

Planner

  • BeforeMetric tracking

    Try thisShare 3 weekly metrics every Monday

  • BeforeMeeting upkeep

    Try thisFinish all 4 weekly checks within 10 minutes

  • BeforeGoal setting

    Try thisAgree on range and alert line for 3 metrics

  • BeforeReport writing

    Try thisPost a 1-page metric summary on the 1st

Designer

  • BeforeDesign upkeep

    Try thisFix all 3 reported layout bugs within 1 week

  • BeforeBanner making

    Try thisPost 4 weekly notice banners in 4 weeks

  • BeforeUI improvement

    Try thisCut signup drop-off from 40% to 30%

  • BeforeGuide upkeep

    Try thisKeep 10 design-guide components up to date

Engineer

  • BeforeServer monitoring

    Try thisResolve every error alert within 24h, 4 weeks

  • BeforeRelease management

    Try thisKeep a weekly release for 4 weeks

  • BeforePerformance

    Try thisKeep first-screen load under 2 seconds

  • BeforeBug handling

    Try thisFix 90% of reported bugs within 48 hours

Wildcard

  • BeforeCommunity care

    Try thisReply to every new post within 24h, 4 weeks

  • BeforeSocial media

    Try thisPost 3 times a week for 4 weeks

  • BeforeSurveys

    Try thisCollect 30 monthly satisfaction responses

  • BeforeInquiries

    Try this90% of first replies within 1 hour

Scoring

With KPIs, mission achievement is actual vs. target. The math depends on the metric's shape

  • Higher is better = actual ÷ target (e.g. 3 of a 4-a-week target → 75%)
  • Lower is better = target ÷ actual (e.g. churn target 2%, actual 2.5% → 80%)
  • Keep-in-range = weeks kept in range ÷ total weeks

Example · Keep weekly meetup attendance at 80%+

  • Weeks at 80%+ out of 580%

80% → Partial

  1. 1

    Holding the range is 100%

    KPIs are about keeping, so 100% means “it ran as planned”. If you keep overshooting, talk about raising the target level.

  2. 2

    Numbers start the conversation

    In a month below the alert line, find the cause before scoring. Judge people by the number alone and hitting the number becomes the goal.

    Strathern (1997)

When it struggles

  • Numbers look normal every month, yet nothing's moved in a yearSomething needs to change. Add one OKR objective (see the OKR tab)
  • Chasing the number misses the point of the workPair metrics — quality with speed, reuse with quantity
  • So many metrics that nobody looksCut to 3 or fewer and demote the rest to reference
  • The number wobbles because people count differentlyWrite who counts, when and where in one line in the feed

Sources · Kaplan & Norton, The Balanced Scorecard (1992) · Strathern (1997) · Goodhart's law · Grove, High Output Management (1983)

Management by Objectives

With MBO, a lead (mentor, professor) and each member agree on goals through conversation, then review those commitments together at the end. It shines when roles are clear and what each person delivered must be recorded.

Question it answers
What has each person agreed to deliver?
Fits when
  • You agree on goals with a mentor and get reviewed, as in internships
  • Individual contribution is graded, as in classes and capstones
  • A team with stable roles sets half-year or yearly commitments
Check-in rhythm
Agree at the start, check once midway, review together at the end. For goals longer than a month, check in briefly monthly

In Perplz terms

Example · capstone helping local shops

  • Project introOrganizational goalWhat must this team deliver this period?Help 5 local shops promote themselves online this semester
  • 2–3 missionsTeam goal (agreed result)What did the team agree to deliver by the end?By semester end, all 5 shops open promo pages
  • Tasks under a missionIndividual goal (agreed commitment)What did I agree to deliver, to what standard, by when?Finish promo pages for 2 shops by end of Oct (Minji)
  • To-dos under a taskAction planWhat is today's step toward the agreed goal?Collect 10 menu photos from shop A's owner and file them

Writing missions (team goals)

With MBO, a mission is a result the lead and team agreed on. One line holds what, to what standard, and by when.

  1. BeforeTo-do

    Support promotion

    Nobody knows what counts as keeping the promise

  2. Try thisResult + deadline

    By semester end, all 5 shops open promo pages

    At the end, you review against the same standard

  3. StretchResult + standard + deadline

    5 pages live by term end, owners rate 4+

    Agreeing on quality too keeps the review steady

“Try this” is enough. If the goal will be used in reviews, agree on a quality standard too.

Formula

[What] + [to what standard] + [by when]

It should be a sentence you set with your lead (mentor, professor). If you wrote it alone, align on it at the kickoff 1:1.

3-second check

  • Did you set it with your lead?
  • At the end, will you judge it by the same standard?
  • Can you deliver it within your role and authority?
  • Is there a midpoint check date?
  1. 1

    Agreement, not instruction

    Goals handed down from above don't become commitments. Members propose first, then align with the lead in conversation.

    Drucker, The Practice of Management (1954)

  2. 2

    Look at results and ways of working together

    Agree on results like “submit the report” along with ways of working like “share a draft 2 days early”, and the review has more to talk about.

  3. 3

    Put a midpoint check on the calendar

    Goals set at the start go stale when conditions change. At least once midway, check together whether the goal still fits.

  4. 4

    Give each goal a weight

    With 3–5 individual goals, set weights that add up to 100%. Everyone then shares the same picture of which commitment matters most.

  5. 5

    Agree on adjustment rules up front

    If the goals feed reviews, decide at the start how to adjust when difficulty or outside conditions change. The more the leads paid attention and reviewed together, the stronger the effect.

    Rodgers & Hunter (1991)

Writing tasks (individual goals)

With MBO, a task is an individual goal one person agreed to. The owner is the one who committed, with a deadline and a standard attached.

  1. BeforeActivity

    Make pages

    Nobody knows what level by when keeps the promise

  2. Try thisResult + deadline

    Finish promo pages for 2 shops by end of Oct

    Check whether the result exists by the deadline

  3. StretchResult + standard + deadline

    Open 2 shop pages by Oct, with owner approval

    An agreed quality bar makes the review clear

Formula

[What] + [how much or to what standard] + [by when]

If the deadline is over a month away, add a midpoint check to the task's steps. Note the weight in the task description or feed.

Marks of a good task

  • The result or standard shows as a number or condition
  • It has a deadline (midpoint check if over a month)
  • One owner — the person who committed
  • It was agreed with the lead
  • It's clear which team goal it adds to
  1. 1

    Check that all individual goals cover the team goal

    Line up everyone's commitments and check together whether any part of the team goal has no owner.

  2. 2

    Don't review by achievement rate alone

    Look only at rates and everyone picks easy goals. Review difficulty, outside conditions and ways of working together.

  3. 3

    Reflect changed conditions by revising the goal

    If conditions change mid-period, revise the goal and post why in the feed. The record is what makes the end-of-period review fair.

Rewrite it like this

  • BeforeMake pages

    Try thisFinish promo pages for 2 shops by end of Oct

  • BeforeReport

    Try thisWrite 3 report pages to the rubric by Nov 15

  • BeforeSales

    Try thisBook 10 new calls, send 5 proposals this quarter

  • BeforeTraining

    Try thisComplete 3 internal courses by end of Sept

  • BeforeClient care

    Try thisCheck in once a quarter with all 20 clients

  • BeforeProcess improvement

    StretchCut weekly reporting from 2 hours to 1 by Nov

Breaking one mission into tasks

MissionBy semester end, all 5 shops open promo pages

  • PlannerSummarize interviews for all 5 shops by Oct
  • DesignerGet 1 page template approved by Nov 15
  • EngineerOpen 3 shop pages by end of Nov, owner-checked
  • WildcardFinish promo pages for 2 shops by end of Oct

Writing to-dos (action plan)

An action plan is the ordered steps toward your agreed individual goal. You agree on the goal with your mentor or professor, but you set the route yourself — today's to-do is the next step on it.

  1. BeforeA lump of work

    Work on the promo page

    You can't see how far you got today

  2. Try thisThe next single step

    Collect 10 menu photos from shop A's owner and file them

    It finishes today and the next step is visible

  3. StretchSomething to show before the check-in

    Show shop A's draft page to the owner and note 3 things to fix

    You have something to show at the midpoint check

Formula

[the next step] of [the agreed goal] + [who to ask when stuck]

Cut steps so they finish before the next check-in (1:1). Checking one moves the task to “in progress”

Marks of a good to-do

  • You know which individual goal and which step number it is
  • It finishes today
  • It says who to ask about what if you get stuck
  • It gives you something to show at the next midpoint check
  1. 1

    Agree on goals, control the plan yourself

    Once the goal is agreed, each person builds and checks their own action plan. When the mentor dictates the steps too, the commitment becomes an instruction.

    Drucker, The Practice of Management (1954) — management by objectives and self-control

  2. 2

    Cut steps to the check-in dates

    If there is a 1:1 or midpoint presentation, split into steps that finish before it. A week with nothing to show means the steps are too big.

    Odiorne, Management by Objectives (1965)

  3. 3

    Change the plan; don't change the goal without agreeing

    If the route is blocked, fix the steps. To lower the goal or move the deadline, agree again and leave a record.

Rewrite it like this

  • BeforeInterview

    Try thisSend 5 questions to shop B's owner and set a visit time

  • BeforeReport

    Try thisDraft the background section of the 2-page report

  • BeforePresentation prep

    Try thisOutline the midpoint slides and share with the team

  • BeforeProfessor meeting

    Try thisFill in the 1-page progress sheet and email it before the meeting

For a project like

MBO shines when each person's commitment must be on record. Pick a similar situation and swap in your deadlines and standards.

Internships

Example · Startup marketing intern (8 weeks)

Mission

  • BeforeDo intern work

    Try thisIn 8 weeks, run 1 new channel on my own

  • BeforeCreate content

    Try thisPublish 8 brand pieces to the guide in 8 weeks

  • BeforeReport results

    StretchMeet all 3 agreed goals by the final 1:1

Task

  • Try thisGet a 1-page channel guide approved by week 2
  • Try thisPost 1 piece of content weekly for 8 weeks
  • Try thisApply 3 pieces of week-4 midpoint feedback
  • Try thisReview a 1-page results summary in week 8

Class projects & capstones

Example · Capstone helping local shops

Mission

  • BeforeSupport promotion

    Try thisBy semester end, all 5 shops open promo pages

  • BeforePrepare the midterm

    Try thisShow 3 shop cases at the Oct 20 midterm

  • BeforeWrite the final report

    StretchMeet all 8 rubric items in the report by Dec 10

Task

  • Try thisFinish promo pages for 2 shops by end of Oct
  • Try thisWrite 1 interview summary per shop, 5 by Oct 15
  • Try thisDraft 2 report pages by end of Nov for review
  • Try thisFinish 10 presentation slides by Dec 5

A company team (half-year)

Example · B2B sales team

Mission

  • BeforeGrow sales

    Try thisSign 10 new client companies in the first half

  • BeforeManage clients

    Try thisKeep 90% renewal in the first half

  • BeforeImprove efficiency

    StretchHalve proposal-writing time by end of June

Task

  • Try thisBook 15 new sales calls by end of March
  • Try thisMeet each of 20 clients once a quarter
  • Try thisBuild 3 proposal templates by end of April
  • Try thisOffer renewals to all 12 due clients by June

Club officers (term)

Example · Dev club organizers, 1-year term

Mission

  • BeforeRun the club

    Try thisHold 20 regular meetups over the 1-year term

  • BeforeRecruit newcomers

    Try thisRecruit 15 new members by end of March

  • BeforePlan events

    StretchHost 2 guest talks rated 4+ by December

Task

  • Try thisPost 3 recruiting posters, get 30 applicants
  • Try thisAnnounce 10 meetup dates in each term's week 1
  • Try thisContact 5 speaker candidates by end of June
  • Try thisLeave 1 handover document at term end

Bootcamp team assignments

Example · 4-week web dev bootcamp project

Mission

  • BeforeDo the team assignment

    Try thisAll 10 requirements built by week 4 Friday

  • BeforeReview code

    Try thisEvery member gets 1+ review on each PR

  • BeforePrepare the demo

    StretchShow 3 core features smoothly in a 5-min demo

Task

  • Try thisBuild 2 features, login and list, by week 2
  • Try thisOpen 1+ PR a day and get it reviewed
  • Try thisWrite 20 test cases by week 3
  • Try thisRehearse the demo twice by week 4 Wednesday

Task examples by role

Under the same team goal, each role commits to different individual goals.

Planner

  • BeforePlanning work

    Try thisGet 1 service plan approved by end of Sept

  • BeforeMarket research

    Try thisDeliver a 5-competitor analysis by Oct 15

  • BeforeSchedule management

    Try thisClose all 3 project deadlines on time this quarter

  • BeforeMeeting upkeep

    Try thisShare meeting notes same day for 12 weeks

Designer

  • BeforeDesign work

    Try thisGet 5 main screens approved by end of Oct

  • BeforeDesign system

    Try thisDocument 20 components by November

  • BeforeBranding

    Try thisFinalize 1 logo and color guide by end of Sept

  • BeforeUsability work

    Try thisRun 2 usability tests, apply 5 fixes this quarter

Engineer

  • BeforeDev work

    Try thisShip payments by end of Oct with 0 errors

  • BeforeCode quality

    Try thisRaise test coverage from 60% to 80% this quarter

  • BeforeTech study

    Try thisEarn 1 cloud certification by November

  • BeforeIncident handling

    Try thisFile every incident report within 48 hours

Wildcard

  • BeforeMarketing

    Try thisGrow newsletter subscribers from 500 to 1,000

  • BeforeSales

    Try thisBook 10 new calls and close 3 deals this quarter

  • BeforeOps work

    Try thisSubmit 1 ops report by the 5th of each month

  • BeforeTraining

    Try thisMake 1 onboarding pack and train 2 new hires

Scoring

With MBO, rate each individual goal against its agreed standard, multiply by weight and add up. The team goal (mission) rate is set with the lead based on this total

  • Individual goal achievement = how much was delivered vs. the agreed standard
  • Total = sum of achievement × weight (weights add up to 100%)

Example · Minji's individual goals this semester

  • Promo pages for 2 shops (weight 50%)100%
  • 5 interview summaries (weight 30%)80%
  • 2 report pages drafted (weight 20%)50%

Total 80% → Partial

  1. 1

    Using it in reviews? Agree on adjustments first

    Reviews are what MBO is for. So agree at the start how to adjust when difficulty or outside conditions turn out different.

  2. 2

    Review ways of working in conversation

    Ways of working like “share a draft 2 days early” are reviewed together in the end-of-period 1:1, not as a number.

When it struggles

  • Agreement turns into instructionMembers propose first, then align with the lead in conversation
  • Early goals don't keep up with changesRevise at the midpoint check and record why
  • Reviewed by rate alone, everyone writes easy goalsLook at difficulty, outside conditions and ways of working together
  • All individual goals together still leave gaps in the team goalWhen agreeing, put individual goals next to the team goal and look for gaps

Sources · Drucker, The Practice of Management (1954) · Rodgers & Hunter (1991), Journal of Applied Psychology

The Perplz method — Perfect Plans: month · week · day

Perplz comes from Perfect Plans — a plan is only perfect once it is carried out, so planning and doing live in one place. The Perplz method links the two through three layers: the mission (a month) is the plan, the to-do (a day) is the doing, and the task (a week) is the number that connects them. Working with AI, many teams make things faster than ever but still revisit their goals once a quarter. So plan by the month, check by the week and act by the day. Perplz's home to-dos, weekly calendar, morning briefing and mission achievement are exactly these three layers.

Question it answers
What do we check this month, this week and today?
Fits when
  • AI agents and generators turn out work daily, so a quarterly plan goes stale fast
  • A team of 2–10 writes its to-dos in Perplz every day and re-aligns tasks every week
  • A solo builder working with AI wants to review a month by what changed, not by how much was made
Check-in rhythm
Declare one blocker every morning, adjust for 10 minutes every Friday, pause for a 30-minute retro in the last week of the month

In Perplz terms

Example · AI draft replies for customer support (3 people, with coding agents)

  • Project introDirectionWhy this project — reread once a quarterA 2-person support team buried in 100 tickets a day stops drafting replies by hand
  • 2–3 missionsA month (plan)Who will be different, and how, when the month ends?Reps finish a reply just by editing the AI draft
  • Tasks under a missionA week (check)What is confirmed by a number this Friday?40 of 50 tickets start from an AI draft (engineer)
  • To-dos under a taskA day (do)What is today's one blocker, and what can you do in 15 minutes?Find 3 ticket types where the draft comes back empty

Writing missions (a month)

A mission says, in one sentence, who will be different and how when this month ends. A month is the longest stretch in which an AI-paced team can notice a wrong direction and turn back. The mission is the planning layer — at month's end, pause and ask whether the to-dos (the doing) completed it and whether what the AI made is the direction you wanted, then rewrite next month's plan.

  1. BeforeA to-do or feature name

    Build an AI draft-reply feature

    Even when it ships, nobody can say what changed for the reps

  2. Try thisA state at month's end

    Reps finish a reply just by editing the AI draft

    Anyone on the team can say whether it happened a month later

  3. StretchA change for people + a line to hold

    Reps trust drafts, customers get no wrong answers

    It speaks of the change for people and customers, not the volume of AI output

“Try this” is enough. If AI output reaches customers, add a line to hold, as in “Stretch”

Formula

[who] + [how they are different at month's end]

The subject is a person — users, reps, the team. If “the AI makes…” becomes the subject, output volume becomes the goal

3-second check

  • Can it be confirmed within a month? (If bigger, split it in two)
  • Is a person the subject — not the AI or a feature?
  • Does it describe what changed, not how much the AI made?
  • Is the end-of-month retro on the calendar?
  1. 1

    A mission is a month — 2 or 3 of them

    Output lands daily, but a quarterly check lets a wrong direction run for three months. Cutting to a month lets you notice and turn back.

  2. 2

    Pause at month's end — a strategic pause

    Look past the numbers and ask: does what the AI made match our direction and values? Could it harm a customer? The faster you move, the more deliberately you must schedule time to look back.

  3. 3

    Volume is not results — the productivity illusion

    AI feels faster, but once you add the time spent verifying and fixing, it often isn't. Unverified output is not a result; it is review work.

    Becker, Rush, Barnes & Rein (2025), METR

  4. 4

    Frequently discussed, transparent — FAST

    A mission is read aloud every week, not filed away. When everyone sees each other's tasks and to-dos, adjustments come faster. The room tracker and feed are that place.

    Sull & Sull (2018), MIT Sloan Management Review

  5. 5

    Keep missions apart from reviews

    Tie achievement straight to reviews and only easy missions get written. Agree on reviews separately, as in the MBO tab, and keep missions as the place to stretch.

Writing tasks (a week)

A task is a result you confirm with a number this Friday. A week is long enough to verify what the AI produced and correct course, so re-align every week to adjust — not to report.

  1. BeforeAn activity

    Work on the draft-generation prompt

    On Friday nobody can say whether it's done

  2. Try thisThis week's result + a number

    40 of 50 tickets start from an AI draft

    Judged by a number on Friday

  3. StretchA result that includes verification

    2 reps review 40 drafts and send 60% unchanged

    It counts results people checked and used, not what the AI generated

Formula

[what or who] + [a number] + [done within the week]

Put verification (a share or count people checked) in the number, and you count usable results instead of AI output

Marks of a good task

  • It has a number (count, people, %) or a “now → target”
  • It finishes within the week (if bigger, lift it to a mission or split it)
  • One owner (even if an agent makes it, one person owns verification)
  • It says who verifies the AI output and how
  • It leads to the mission — “does this bring us closer to this month's mission?”
  1. 1

    Re-align every Friday — adjust, don't report

    Say what will block you next week before what you did last week. Fix wrong numbers on the spot; for a task you're pushing back, move the deadline with “Postpone” and leave the reason in the feed.

  2. 2

    Stop work from expanding

    Fill the time AI freed up and the week is full again, often with work outside your role. Keep it to 1–2 tasks per person per week and 2–5 per mission.

    Ye & Ranganathan (2026), UC Berkeley Haas

  3. 3

    Put verification time into the task

    “Generate 100 drafts” takes the AI ten minutes. Write “2 reps review 40 drafts and send 60% unchanged” so the week measures something real.

Rewrite it like this

  • BeforeImprove the prompt

    Try thisCut the draft rewrite rate from 50% to 30%

  • BeforeWrite tests with the agent

    Try thisReview 30 agent-written tests and keep 20

  • BeforeAnalyze logs

    Try thisFind 3 empty-draft ticket types in 200 log entries

  • BeforeOnboarding doc

    Try this2 new reps send a first draft using only the doc

  • BeforeReply quality control

    Try thisCut wrong-answer reports from 5 to 1 a week

  • BeforeExpand automation

    StretchAll 3 reps clear the day's tickets before leaving

Breaking one mission into tasks

MissionReps finish a reply just by editing the AI draft

  • PlannerWatch 3 reps, find 5 reasons they discard drafts
  • DesignerEdit screen: all 3 reps send in under 1 minute
  • Engineer40 of 50 tickets start from an AI draft
  • WildcardAsk 10 customers after a reply; 8 say “good”

Writing to-dos (a day)

The to-do is the doing layer — this is where the plan gets completed. It is the one blocker you pick today. In the morning, write the one thing blocking your task and turn it into an action you can start in 15 minutes. With AI alerts and threads open everywhere, one to-do gives you a place to come back to.

  1. BeforeA chunk you can't finish today

    Raise draft quality

    You can't check it off tonight

  2. Try thisToday's one blocker → a 15-minute action

    Find 3 ticket types where the draft comes back empty

    Tonight it's clear whether you found them or not

  3. StretchThe blocker + a handoff

    Find the 3 empty-draft types and hand them to the engineer in the feed

    A day's result turns into the team's next move

Formula

[today's one blocker] → [an action you can start in 15 minutes]

Write it in Perplz home to-dos; checking it moves the task to “in progress”. Leave finished to-dos as one line in the feed

Marks of a good to-do

  • You can check it off today
  • It is today's single biggest blocker (a second one goes to tomorrow)
  • You can start the first action within 15 minutes
  • You know which task it belongs to (from home, a task is optional)
  1. 1

    Declare one blocker

    Read the morning briefing and pick the one thing that, once cleared, moves the task. A long list turns a day into a week.

  2. 2

    Timebox to cut the AI threads

    Agents and chats send alerts from several windows at once. Block to-do time in 15-minute boxes and silence alerts while you're in one. The to-do is where you return.

  3. 3

    Don't take the AI's answer as is

    Under time pressure you wave things through unchecked. Write a “review” action into the to-do. Hand over all your thinking and tomorrow's judgment gets slower.

Rewrite it like this

  • BeforeTweak the prompt

    Try thisRead 10 empty replies, change 1 prompt line and rerun

  • BeforeCheck the agent's output

    Try thisFlag the files without tests among the 5 the agent changed

  • BeforeTalk to a rep

    Try thisCall 1 rep for 15 minutes and hear why they discarded a draft

  • BeforePrep the retro

    Try thisNote 3 things the AI got wrong from this week's done to-dos

For a project like

Even when everyone works with AI, teams work in different shapes. Pick the one closest to yours and follow it by month, week and day.

Each person running agents, working in parallel

Example · 3 people · 2 coding agents, 10 PRs a day

Mission

  • BeforeIncrease agent use

    Try thisThe team merges reviewed agent code within a day

  • BeforeImprove code quality

    Try thisNo agent change ships without a human review

  • BeforeDevelop faster

    StretchThe day ends with no more than two threads open

Task

  • Try thisMerge all 20 agent PRs through human review
  • Try thisKeep stale PRs at 0 for 3 of 4 weeks
  • Try thisReview 30 agent-written tests and keep 20
  • Try thisCut open threads at once from 5 to 2

A small team where each person wears several hats

Example · 2 people · planning, design and code split with AI

Mission

  • BeforeGrow the scope of work

    Try thisBoth clear today's to-dos before logging off

  • BeforeOrganize the role split

    Try thisEveryone knows who verifies which AI output

  • BeforeMake work more efficient

    StretchIf one is out, the other takes that week's tasks

Task

  • Try thisCut tasks per person per week from 3 to 2
  • Try thisHave 15 days in 4 weeks with every to-do checked
  • Try thisAssign 1 verifier to each of 6 AI output types
  • Try thisWeekly check done in under 20 min, all 4 weeks

An async team that works in writing, not meetings

Example · 4 people across time zones · the feed instead of calls

Mission

  • BeforeImprove communication

    Try thisTeammates know today's blockers without a meeting

  • BeforeOrganize the records

    Try thisThe AI summary alone covers last week's decisions

  • BeforeImprove collaboration

    StretchWhoever is stuck gets an answer within 24 hours

Task

  • Try thisEveryone posts 1 feed a day for 4 weeks
  • Try thisPost the AI weekly summary each Friday, 4 weeks
  • Try thisBlockers answered within 24 h: 60% → 90%
  • Try thisCut meetings from 3 a week to 1

A team where reviewing AI output is the bottleneck

Example · 3-person content team · 30 AI drafts a week

Mission

  • BeforeIncrease content output

    Try thisEditors write more than they review drafts

  • BeforeImprove quality control

    Try thisNo factual error in an AI draft reaches a reader

  • BeforeMake review automatic

    StretchReaders can tell our writing from other AI writing

Task

  • Try thisCut editing time per piece from 60 to 30 minutes
  • Try thisCatch 100% of factual errors pre-publish, 4 weeks
  • Try thisRaise drafts used as is from 20% to 50%
  • Try thisRaise read-to-the-end rate from 40% to 55%

Building alone with AI

Example · Weekend side project · vibe coding

Mission

  • BeforeDevelop the MVP

    Try thisA friend finishes the core flow with no help

  • BeforeAdd features

    Try thisI can read and explain the code the agent wrote

  • BeforeLaunch the service

    StretchStrangers find it, use it and come back

Task

  • Try thisAll 5 testers finish the 3-step core flow alone
  • Try thisAgent-written files I can't explain: 20 → 0
  • Try thisCheck the weekly to-do all 4 weeks
  • Try this5 of the first 20 visitors come back a week later

Task examples by role

Even with AI, people verify and decide. Pick the number your role confirms each week.

Planner

  • BeforeGather requirements

    Try thisInterview 3 reps, find 5 reasons they drop drafts

  • BeforeAI adoption plan

    Try thisPrioritize 5 of 10 ticket types to hand to AI

  • BeforeRun the weekly check

    Try thisHold all 4 Friday checks over 4 weeks

  • BeforeRetro notes

    Try thisLog 3 “direction right?” answers at the retro

Designer

  • BeforeDraft-screen design

    Try thisPick 1 of 2 edit-screen drafts by rep vote

  • BeforeUse AI images

    Try thisPick 3 on-brand drafts out of 10 AI images

  • BeforeUsability check

    Try thisAll 3 reps edit and send a draft in under 1 minute

  • BeforeDesign QA

    Try thisCheck all 5 agent-changed screens vs. the design

Engineer

  • BeforePrompt tuning

    Try thisCut the draft rewrite rate from 50% to 30%

  • BeforeAgent code review

    Try thisMerge all 20 agent PRs through human review

  • BeforeStabilization

    Try thisCut draft-generation failures from 5% to under 1%

  • BeforeMonitoring

    Try thisError alerts hit the channel in 10 min, 4 weeks

Wildcard

  • BeforeCustomer feedback

    Try thisHear “good” from 8 of 10 customers after a reply

  • BeforeData cleanup

    Try thisTag all 200 ticket log entries by type

  • BeforeDocumentation

    Try this2 new reps send a first draft using only the doc

  • BeforeVerification test

    Try thisRead 30 AI drafts; confirm 3 or fewer are wrong

Scoring

In the Perplz method, a mission's (month) achievement is the average progress of its weekly tasks. At the month-end retro, write “was the direction right?” next to the number

  • Mission achievement = average progress of the weekly tasks
  • Task progress = (now − start) ÷ (target − start); a task checked done counts 100%

Example · Reps finish a reply just by editing the AI draft

  • Week 1 · tickets starting from an AI draft, 0 → 40 of 50 (target 50)80%
  • Week 2 · draft review time 20 → 12 min (target 10)80%
  • Week 3 · sent unchanged 0% → 36% (target 60%)60%
  • Week 4 · 10 wrong-answer rules written (done)100%

Average 80% → Partial

  1. 1

    Keep achievement apart from reviews

    Achievement tells you what help comes next month. Tie it straight to reviews and only easy missions get written.

  2. 2

    Ask about direction before the number

    Even at 100%, if what the AI made isn't the direction you wanted, change next month's mission. It's the first question of the retro.

When it struggles

  • AI output keeps growing and looks like resultsChange task numbers from “how much was made” to “results people verified and used”
  • Tasks multiply with the saved time and the week is full againCut to 1–2 tasks per person per week and push the rest to next week
  • The weekly check has become a “done / not done” reportStart with what will block you next week and change tasks on the spot
  • Next month's missions get written without the retroPut 30 minutes on the calendar first and answer “is the direction right?” before anything else

Sources · Sull & Sull (2018), With Goals, FAST Beats SMART, MIT Sloan Management Review · Becker, Rush, Barnes & Rein (2025), Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, METR · Ye & Ranganathan (2026), UC Berkeley Haas — research on generative AI and work intensity

Whatever the method — check with SMART

SMART isn't a method; it's a way to check sentences. Whether you write OKR, KPI or MBO, hold your finished missions and tasks up to it once.

  • S

    Specific · Specific

    Instead of broad words like “run” or “improve”, can you see the target?

  • M

    Measurable · Measurable

    On the due date, is it clearly done or not done?

  • A

    Assignable · Someone can own it

    Can one person own it? It's often remembered as “Achievable”, but the original A is about who owns it

  • R

    Realistic · Realistic

    A bit hard, but within reach of your resources and time?

  • T

    Time-related · Has a deadline

    Is there a set end date?

Doran wrote that not every goal needs all five. Use it as a check, not a checklist.

Doran (1981), There's a S.M.A.R.T. way to write management's goals and objectives

If it's your first time — learning goals

Whatever the method, make unfamiliar work about what you learned and found, not a result number. Chasing numbers before you know the method makes you anxious and blocks learning.

  • MissionI can build screens in Flutter on my own

    • Try thisBuild 1 Flutter tutorial app step by step
    • Try thisBuild 3 of my own screens without a tutorial
  • MissionWe understand what keeps people watching

    • Try thisFind 3 things 5 fast-growing channels share
    • Try thisPost 2 videos applying those 3 things
  • MissionWe can explain contest criteria in our words

    • Try thisRead 10 past winners, make a 1-page checklist
    • Try thisCheck our plan with it and fix 3 gaps

Seijts & Latham (2005)

Along the way — a 10-minute weekly check

Goals stay alive when you look at them more often than you set them. Whatever the method, meet for 10 minutes on the same day each week.

  1. 2 min

    Numbers

    Say each task's current value, no explanations. In team projects, the leader updates mission achievement here

  2. 3 min

    What we learned

    One line each on something new from the past week — whether the numbers look good or bad

  3. 3 min

    What's stuck

    Say who could help with what, and pick the helper right there

  4. 2 min

    Change anything?

    If something below applies, change the mission or task and note why

OK to change when

  • Try thisYou have evidence the direction was wrong
  • Try thisConditions like timeline or team size changed
  • Try thisYou learned a task's number doesn't really measure the mission

Don't change when

  • BeforeThe numbers aren't coming and an easier goal looks tempting
  • BeforeSomething more fun came along
  • BeforeIt's about one person's situation (change the owner instead)

When you change something, post one line in the feed about why. A goal that changes without a reason wasn't changed — it was abandoned.

Scoring achievement

In team project rooms, the leader can record each mission's achievement (0–200%, in 10% steps) at any time. When the project ends, those values pre-fill the final review. The math follows your team's method.

By method

  • OKRAverage progress of the tasks (key results)
  • KPIActual vs. target, or share of periods kept in range
  • MBOSet with the lead, based on achievement × weight across individual goals
  • PerplzAverage weekly-task progress — with a direction retro at month's end

The “Reference: n/m tasks done” in the popover is the simplest starting point. If your tasks carry numbers, the method's own math is more accurate.

Bands

0–69%
Below
70–99%
Partial
100–129%
Met
130%+
Exceeded
  1. 1

    Low numbers mid-project are normal

    30% in week 3 may be right on plan even with a “Below” badge. Read the bands as of the end.

  2. 2

    Decide first what the rate is for

    OKR scores don't go straight into pay formulas; MBO feeds reviews, with adjustment rules agreed up front. Whatever the method, don't use it to rank people against each other.

Personal projects have a done check instead of a percentage. When the tasks are done, check the mission off, and post the current numbers in the feed at each weekly check.

Common mistakes

These show up whatever the method. How each method struggles is in its own tab.

  • The mission is a feature name or a to-doRewrite it as what will be different, or the number you'll keep
  • A task is one word, like “Design” or “Study”Add a number for what and how much
  • A task takes a month with no midpoint checkSplit it, or add a midpoint check to its steps
  • The leader writes everyone's tasksLook at the mission together and each write your own
  • Nobody looks at it againPick a weekly 10-minute check day
  • A method that doesn't fit the team was copied as isPick again with the three questions in “Which method fits?”

Writing with AI

  • The Project Agent reads your mission sentence and suggests tasks and steps by the rules of the method (OKR, KPI, MBO or Perplz) set in project info.
  • Let AI organize and check; let the team decide what the goal is. The moment AI proposes a more polished goal, it stops being one the team chose.

A check prompt you can copy

Check the missions and tasks below against [our method: OKR, KPI or MBO]. 1. Is each mission written, per our method, as a state to reach, a metric to keep or an agreed result — not a to-do? 2. Does each task have a number or standard, so its due date shows done or not done? 3. Does each task have one owner and a deadline? 4. Does each mission have 2–5 tasks? 5. Is it clear which mission each task leads to? 6. Is anything missing from SMART (specific, measurable, assignable, realistic, time-related)? Mark each criterion pass or fail, and suggest one rewritten sentence for each fail. Rewrites are only suggestions — don't change what the goal is. [Paste your missions and tasks here]

FAQ

Ready to write your first mission?

Pick a method, and two or three missions with a few tasks are enough to start

Create a project

Looking for how to use the other features? See the guide

Mission & Task Writing Guide: OKR, KPI, MBO | Perplz