Coverage table
Every test case against every device — pass, fail, blocked or not run. Fails link straight to the finding.
Fixed price3 business daysPhones · laptops · browsers
Your AI wrote the test cases.
Now run them on real phones and browsers.
Pay online, upload your build and test cases, and get a verified pass/fail report from real testers on the phones, laptops and browsers your users actually own.
From $0. No account, no sales call. Testers are on it within the hour.
Job 99T-2417 · StandardRunning
Sessions accepted
6of 10
Findings so far
32 verified
Cases covered
47of 58 parsed
Report due
Thu 18 Sep
Redmi 12Android 13 · MIUI 14 · Jio 4GIn session
Galaxy A15Android 14 · One UI 6 · Vi 4GSubmitted
Realme C55Android 13 · realme UI 4Offered
iPhone 11iOS 17.6 · Airtel 4GIn session
Chrome 128 · Windows 11Laptop · 1366×768 · Airtel broadbandSubmitted
Verification 2 accepted · 1 pendingStarted 25 min ago
The AI test case is not the test
Generated cases describe intent. Only a person on a device can tell you whether the app does it.
Emulators are not real devices
MIUI kills background tasks. Jio drops mid-payment. Safari blocks the popup. A 2019 Windows laptop on Chrome scrolls differently. Device clouds don't show you that.
Every finding is verified
Incomplete reports are rejected, duplicates merged, severity graded against a published rubric before you see them.
Unfilled sessions are credited
If we can't match a device in your turnaround, the sessions that didn't run come back as credit on your email, applied to your next job.
The gap
Teams generate tests faster than they can run them.
AI test generation has been adopted across the industry. Real-device execution has not caught up.
The cases exist. What doesn't exist is the afternoon someone spends on a ₹9,000 Android with a flaky connection, following them one by one and writing down what actually happened. Most releases ship without that afternoon. The ones that had it were surprised by what it found.
Automation answers "does the script still pass". A person on a phone — or on Safari on a MacBook, or Chrome on an office Windows laptop — answers "does this work for the customer in Hubballi". Those are different questions, and only one of them is in your release checklist.
Before you ship
Seven questions your release manager gets asked.
If any of these takes more than a minute to answer, the gap is in your evidence, not your test cases.
01Which of the generated test cases actually ran on a real device, and which were never executed?Coverage
A test case that was generated but never run is a hypothesis. The report shows every case against every device with pass, fail, blocked or not run — and says why for the last two.
02Does checkout survive an app switch to a UPI app on a budget Android with aggressive battery management?Real devices
MIUI and ColorOS kill backgrounded apps by default. This is the single most common S1 our testers find, and it does not reproduce on an emulator.
03For each failure, can a developer reproduce it from the report alone — without asking who found it?Reproducibility
Every accepted finding has steps, expected, actual, device, OS version, network and a screen recording. Findings without them are rejected before they reach you.
04How many of the bugs in your tracker are the same bug reported three times?Verification
Duplicates are merged during verification, and the report tells you how many were merged so the count you're looking at is the count you have.
05Who decided that bug was an S2 and not an S1, and against what rule?Severity
Severity is graded against a published rubric with a confidence score. When the verifier changes a tester's grade, the report says so and why.
06What happens on Safari on a MacBook, or Chrome on the five-year-old Windows laptop your customer's office still runs?Browser matrix
You set targets by browser and OS as well as by phone. Sessions are matched to testers who own those exact machines, the session page records the real browser version and viewport, and the matrix in the report proves it.
07When the fix ships, will the same cases be re-run on the same devices?Re-run
Every report has a one-click re-run on the same plan with a new build. Same devices, same cases, new results — so you can show the fix rather than assert it.
We leave the "evidence" column of your release checklist to the report, not to a meeting.
What you receive
Six things in every report. One fixed price.
Delivered as a hosted page and a PDF, addressed to whoever paid, forwardable to whoever needs to act.
Reproducible findings
Steps, expected, actual, device, OS, network and a screen recording on every serious bug.
Verification record
How many findings were accepted, merged or rejected, the rubric version, and a confidence score on each.
Device and browser matrix
Which phone or laptop, which OS build, which browser version and viewport, which network, which session — so "tested on Android" or "works on Chrome" means something.
Session log
Start and end times, tester band, cases assigned and cases completed for every session that ran.
Coverage note
Anything that didn't run, and why. Plus a one-click re-run on the same plan when your fix is ready.
Three business days
From payment to report.
Minute 0
Pay
Choose a package, pay by card. Invoice issued at checkout.
Minute 2
Submit
Build link or web URL, test cases in any format, device and browser targets, deadline.
Minute 5
Plan
Cases parsed into sessions by device. You see exactly what will run.
Hour 1
Dispatch
Offers go to testers who own your target phones and laptops. First to accept takes the session.
Day 1–2
Sessions
Timed 45- or 90-minute sessions. Findings recorded live with media.
Day 3
Verify and deliver
Findings checked, merged, graded. Report emailed as page and PDF.
What we need from you
A build link — APK, Play internal testing, TestFlight — or a web URL, and your test cases. Test credentials if the app needs them. That's all.
What we don't need
A call, a scope document, a device list beyond "budget Android, iPhone 11+, Chrome on Windows, Safari on Mac", or your time after the two-minute submission.
Pricing
Fixed. Nothing further is billed.
No hourly overage, no expenses, no change requests. A slot is a phone, a laptop or a browser — mix them in one job. Add sessions in blocks of five if you need more.
Unfilled sessions credited
If a device target can't be matched within your turnaround, the sessions that didn't run are credited back automatically and applied to your next job or re-run.
Paid before work starts
Because the job is automated end to end, there is no invoice cycle. Indian customers are billed in INR with GST; everyone else in USD.
Below most approval thresholds
Starter and Standard sit under the card limit of most engineering managers. That is deliberate: no procurement.
Who runs it
20,000 testers. The phones and laptops your users own.
99tests has run crowdsourced testing for enterprise clients for over a decade. The pool is real people across Indian cities on the phones, laptops, browsers and networks that matter.
Every session is timeboxed and paid whether or not bugs are found, so testers report what they see, not what earns a bonus. Testers who don't show up stop receiving offers. Testers who find real, reproducible bugs earn severity bonuses on top of the session fee.
Verification is separate from testing: the person who found a bug is never the person who grades it.
P&GFlipkartInfosysWipro
Redmi 12MIUI 14Android 13
Galaxy A15One UI 6Android 14
Realme C55realme UI 4Android 13
Vivo Y28Funtouch 14Android 14
iPhone 11iOS 17
iPhone 13iOS 18
Windows 11Chrome · Edge1366×768 laptop
macOS 15Safari · ChromeMacBook Air
20,000testers registered across India
45 / 90minute sessions, fixed fee
3 dayspayment to report, Starter and Standard
In scope
- Execution of your test cases on real, tester-owned phones, laptops and desktops
- Exploratory testing around those cases within the session charter
- Findings with steps, media and verified severity
- Android and iOS builds distributed by link; web apps by URL on Chrome, Safari, Edge and Firefox
- Re-runs of the same plan on a new build
Not in scope — stated plainly
- Not test automation, and no substitute for it
- Not a security or penetration test
- Not a certification or an audit under any standard
- Not performance or load testing
- Not desktop native apps (Windows or macOS installers) yet
- Writing your test cases — bring your own, AI-generated or otherwise
Questions
Before you pay.
What formats do you accept for test cases?
CSV, Excel, Google Sheets export, Word, Markdown, plain text, and TestRail or Jira exports. AI-generated cases from any tool are fine. We parse them into a plan and show you the coverage before testers start.
How do I share the build?
An APK download link, a Google Play internal testing link, a TestFlight public link, or for web apps a staging or production URL. If it needs test credentials or basic auth, include them in the submission form; they are shown only to assigned testers.
Which browsers and laptops do you cover?
Chrome, Safari, Edge and Firefox on Windows, macOS and ChromeOS, on testers' own laptops and desktops. You choose the browser and OS as targets the same way you choose a phone. The session records the actual browser version and screen size, and the report shows them.
Who chooses the devices?
You set targets — brand and minimum OS for phones, browser and OS for laptops. We match testers who own those devices. If a match can't be found within your turnaround we widen once, and if it still can't be filled the unfilled sessions are credited back to your email and applied to your next job.
Is my app confidential?
Testers accept confidentiality terms before they can take sessions. Builds are shared by link only to assigned testers and are not stored by 99tests beyond the job. Recordings are stored privately and linked only from your report.
Do you invoice with GST?
Yes, for Indian customers. Enter your GSTIN at checkout and Stripe issues a compliant invoice immediately. Customers outside India are billed in US dollars and receive a standard invoice.
What if there are more test cases than the package covers?
After upload you'll see exactly which cases are uncovered and can add five sessions at a time. Or leave them; the report lists them as not run.
Can I get a custom scope?
Not yet. The three packages cover most release checks, and blocks of five sessions cover the rest.
Ship this release knowing it works on a Redmi — and on Safari.
Pay, upload, and testers are on it within the hour. Report in three business days.