How to write a micro task that gets good submissions

By Tasklify TeamUpdated 9 min read

Knowing how to write a micro task is the difference between a batch of useful results and a queue of submissions you have to reject. Workers read your task once, on a phone, between other tasks. This guide shows the anatomy of a task that survives that reading, how to design proof, how to price per minute, and five templates you can copy.

Anatomy of a micro task

A micro task is a self-contained unit of work that one person can finish in a few minutes without talking to you. Every field in the task form exists to remove a question the worker would otherwise have to ask. On Tasklify a task has these parts:

PartWhat it doesGood example
TitleLets workers decide in two seconds whether the task fits them“Sign up and rate our onboarding (Android, 5 min)”
Category and linkSets context and gives a single place to startCategory “test”, link to the staging build or landing page
InstructionsExplains the goal and anything unusual, in plain words“We want to know where new users get stuck.”
Steps (up to 20)The exact sequence of actions“3. Tap Continue on the plan screen.”
Proof types and proof noteWhat the worker must submit and what it must show“Screenshot of the Welcome screen showing your username.”
Estimated timeSets expectations; drives how workers judge the reward5 min
Time limitHow long a reservation lasts before the slot is released1 hour for a 5-minute task
TargetingWho can see the task: countries (or Tier 1 only), device, min approval rate, min completed tasks, Plus-onlyiOS, Tier 1, 90%+ approval
Reward, workers, review periodBudget, sample size and how fast you commit to review (24h to 7 days, optional auto-approve)$0.35 × 50 workers, 48h review

The golden rule: one task, one outcome. “Install the app, try three features, and also fill in our survey” produces half-finished work, because each worker stops at a different point. Split it into a test task and a survey task and you get two clean datasets.

Writing clear instructions

Lead with the goal, then the steps

One sentence of context helps workers make sensible choices when something unexpected happens. “We are checking whether the sign-up form works on older phones” tells them that an error message is a valid result to report, not a reason to give up.

One action per step

  1. Open the link in the task (use your phone’s default browser).
  2. Tap Get started and create an account with your own email.
  3. Skip the tutorial by tapping Later.
  4. Take a screenshot of the Home screen.
  5. In the text answer, describe the one step that was least clear, in 1–2 sentences.

Quote button labels exactly as they appear in your product. Say what to do if a step fails (“If you see an error, screenshot it and describe what you tapped just before”). Mention anything workers should not do, such as paying real money or contacting your support team.

Write for a phone screen and a second language

Many workers read English as a second language. Short sentences, common words, no idioms, no internal jargon (“the PLG funnel”), and numbers instead of words (“3 screens”, not “a few screens”). If a word could mean two things in your product, pick one and use it consistently.

Stay within the rules. Tasks must describe genuine work. The Acceptable Use Policy prohibits fake reviews presented as genuine, rating manipulation, accounts created with false identities, artificial ad traffic and similar deceptive engagement. Design tasks around real installs, honest opinions and actual usage.

Proof requirements

Proof is how you tell a completed task from a skipped one without redoing it yourself. Good proof is specific, cheap for an honest worker to produce, and hard to fake. Tasklify supports six proof types; you can combine them.

Proof typeBest forMake it specific
ScreenshotReaching a screen, completing a flowName the screen and the detail that must be visible (username, date, order number)
Text answerOpinions, data entry, answers to questionsGive a format (“3 bullet points”, “price in USD, no symbol”)
Link / URLSomething the worker found or created publiclySay which page type is valid (“product page, not a search results page”)
Username or emailMatching a sign-up in your own databaseTell workers which identifier you will look up
Screen recordingUsability tests, bug reproductionAsk them to think aloud or narrate, and to start recording before step 1
File uploadDocuments, spreadsheets, audioSpecify format and naming (“CSV, one row per item”)

Add a unique element that only someone who did the work can know: the confirmation number on the last screen, the third word of your welcome email, or a code you display at the end of a survey. Tasklify also flags duplicate files and duplicate text answers across submissions, which catches copy-pasted proof. For a deeper look, see our guide to crowdsourcing quality control.

Pricing a micro task per minute

Workers compare tasks by what they pay per minute, so you should too. Time the task yourself on the target device, then add 30–50% because a first-time user is slower than you. Tasklify’s current suggested rewards work out to about $0.07 per minute:

Estimated timeSuggested rewardTypical tasks
2 min$0.15One-screen check, single question, short data lookup
5 min$0.35Sign-up flow with feedback, short survey
10 min$0.70Feature test with written notes, 15–20 question survey
15 min$1Usability task with screen recording
30 min$2Exploratory test with a bug report

The total you pay is reward × workers + a 12% service fee, plus an optional Boost of 20% that pins the task higher in the feed. Rewards are held until you approve each submission.

Worked example. A 5-minute onboarding test at $0.35:

  • Pilot with 10 workers: $3.50 in rewards + $0.42 fee = $3.92.
  • Full run with 100 workers: $35 + $4.20 fee = $39.20; with Boost, add $7.

The minimum top-up is $10 and your first top-up gets a 10% bonus (up to $5) you can spend on tasks, so a pilot like this fits in the first deposit. Full fee details are on the pricing page.

Underpricing is a false economy: low-paying tasks fill slowly, attract the least careful workers, and cost you more time in review than you saved. If a task takes longer than your estimate in the pilot, raise the reward before scaling.

Templates for five task types

Copy these, replace the bracketed parts, and cut every step that does not apply.

1. App or website test

Title: Test [feature] on our [Android/iOS] app ([N] min)

Steps: 1) Install [app] from the link. 2) Create an account with your own email. 3) [Do the key action]. 4) Screenshot the [final screen]. 5) Describe anything that did not work or confused you.

Proof: Screenshot + text answer (“at least 2 sentences; ‘everything fine’ is not enough”).

2. Survey or market research

Title: [N]-question survey about [topic] ([N] min)

Steps: 1) Open the survey link. 2) Answer every question honestly; there are no right answers. 3) Copy the completion code shown on the last page.

Proof: Text answer with the completion code. See how to get survey respondents for screening and attention checks.

3. Data entry or lookup

Title: Find [field] for [N] [items] ([N] min)

Steps: 1) Open the item list in the link. 2) For each item, find [field] on [source]. 3) Enter the values in the format below.

Proof: Text answer, one line per item: item ID; value; source URL.

4. Sign-up flow

Title: Sign up for [product] and tell us what stopped you ([N] min)

Steps: 1) Open the link. 2) Sign up with your real email (no test or temporary addresses). 3) Complete [first action]. 4) Screenshot the confirmation.

Proof: Username or email you used + screenshot. Genuine sign-ups only: no false identities or purchased accounts.

5. Feedback on a page, design or message

Title: Give feedback on our [landing page/pricing/email] ([N] min)

Steps: 1) Look at the page for 30 seconds. 2) Answer: what does this product do? 3) Who is it for? 4) What would stop you from trying it?

Proof: Text answer with each question numbered. Feedback stays private to you; it is not a public review.

Common mistakes that cause bad submissions

MistakeWhat happensFix
Vague proof (“send proof”)Every worker sends something differentName the screen and the visible detail
Several goals in one taskPartial work, hard to judgeOne outcome per task
Time estimate too lowRushed work, few takersTime it yourself on the target device and add a margin
Open-ended feedback questions“Good app” one-linersAsk specific questions with a minimum length
Hidden requirementsRejections that end in disputesEvery rule you will enforce must be in the task
Wrong targetingiPhone owners testing an Android-only flowSet device and countries before publishing
Staging link that expires or needs a VPNWorkers cannot startTest the link logged out, on mobile data

If a submission is close but not quite right, use request changes (available once per submission) with a short note. It is faster than rejecting and reposting, and it is fairer to a worker who misunderstood.

Pre-publish checklist

  • The title starts with a verb and states the platform and time.
  • The task has one outcome, and the steps lead to it in order.
  • Every rule you will use to reject is written in the task.
  • The proof note says exactly what must be visible or written.
  • You timed the task yourself; the reward matches the per-minute rate.
  • Targeting matches the product (device, countries, experience level).
  • The link works logged out, on mobile, without a VPN.
  • You will run a pilot of 5–10 workers before scaling.

Frequently asked questions

How long should micro task instructions be?

As short as possible while still being unambiguous. Most good micro tasks have three to eight numbered steps and a short proof note. If you need more than about ten steps, the task is probably two tasks.

What reward should I set for a micro task?

Time the task yourself, add a margin for someone who has never seen your product, and price it per minute. Tasklify suggests rewards for 2, 5, 10, 15 and 30-minute tasks when you create a task; generous rewards fill faster and attract more careful workers.

Which proof type is best?

The one that shows the outcome with the least effort. Screenshots suit “reached this screen” tasks, text answers suit opinions and data, screen recordings suit usability tests, and links or usernames suit tasks that produce something public. Ask for two proof types only when one cannot show the result.

Can I reject a submission that does not follow the instructions?

Yes, with a written reason tied to the published instructions. For fixable problems you can request changes once instead. Workers can dispute a rejection within 7 days and Tasklify reviews the instructions and proof to decide.

Can I ask workers to post reviews or ratings of my app?

No. Tasklify’s Acceptable Use Policy prohibits fake reviews presented as genuine and rating manipulation on platforms that forbid it. Ask for private, honest feedback in a text answer instead.