Curated interview practice questions with answers. These are practice scenarios, not a verified list of past interview questions.
About Affine hiring for Software Testing roles
Affine is a Bengaluru-based analytics and AI consulting company that helps businesses in gaming, media, retail, consumer goods and technology use data science for decisions. Its teams build predictive models, marketing analytics, pricing solutions and data engineering pipelines. It hires business analysts, data scientists, data engineers and analytics consultants who combine technical skills with business understanding.
For Software Testing roles, this Affine practice set focuses most on Agile, SQL and API Testing (8), Performance and Security Basics (6), Testing Types (5), Bug Life Cycle and Jira (5). It has 6 easy, 28 medium, 6 hard questions, so you can start easy and work up to the harder ones.
Company type
Analytics
Headquarters
Bengaluru, India
Main India offices
Bengaluru
Affine Software Testing interview process
Online assessment (aptitude + SQL + statistics)
Technical interview
Case study round
HR round
Typical pattern based on public candidate experiences. Rounds vary by role, team and year.
Solved for AffineStart practicing
3 QUESTIONS FREESign up to unlock the full question bank and track your progress.Unlock every question and answer, then continue practising company-wise and package-wise.
✓ 3 questions free • Unlock the rest in seconds
01
How do you test session management in a web application?
Check that a new session ID is created after login, and that the session expires after the defined idle time. Verify logout invalidates the session on the server, so pressing back or reusing the old cookie does not work. Check that cookies use Secure, HttpOnly and SameSite flags. Test concurrent logins if they are limited, and that changing the password ends other sessions. Session IDs must not appear in URLs.
Response time is the total time from sending a request to receiving the complete response, as the user experiences it. Latency is the delay before the first byte of the response arrives, mostly network and waiting time. Throughput is how many requests or transactions the system handles per unit of time, such as 500 requests per second. Good performance usually means low response time at the required throughput.
Open developer tools and use the network tab with cache disabled to see each request's size and time, and the total load time. Look for very large images or scripts, slow API calls and requests that fail or repeat. Use throttling to simulate slow 3G or 4G. Run the built-in performance audit tool to get scores for load speed and suggestions. Record the numbers so you can compare with later builds.
05
What is idempotency in APIs and how do you test it?
An operation is idempotent if repeating the same request gives the same result as doing it once. GET, PUT and DELETE should be idempotent; POST usually is not. To test, send the same request several times and check the server state does not change after the first call. For payments, check that retrying with the same idempotency key does not create a second charge, which protects users during network retries.
Model answer: Situation: I logged a bug where the invoice PDF showed the wrong GST number, and the developer marked it minor because the screen value was correct. Task: get the right priority. Action: I explained that customers use the PDF for tax filing and showed the requirement. Result: the product owner agreed it was major, it was fixed before release, and we added PDF checks to regression.
Read the requirement and identify inputs, outputs, business rules, user roles and states. Turn each rule into a condition with an expected result, then add negative and boundary conditions. For example, 'users can withdraw up to 10,000 per day' becomes: withdraw below the limit, exactly at the limit, above the limit, multiple withdrawals adding up to the limit, and the limit resetting the next day. Each condition then becomes one or more test cases.
10
A manager says the app has 50,000 daily users, so the load test must run 50,000 virtual users at once. How do you respond?
Explain that daily users are not concurrent users. Of 50,000 users a day, only a fraction are active at the same moment, and even they pause between actions because of think time. Use analytics to find the real peak concurrency and request rate, for example 1,500 concurrent users at peak hour, and model that with a safety margin. Running 50,000 users clicking at once tests a load that never happens and wastes effort.
Internationalisation tests that the product can support many languages and formats; localisation tests that a specific locale is correct for text, currency, date and layout.
Evaluating software to find defects and confirm it meets requirements. It reduces risk, protects users and saves the cost of fixing bugs after release.
First query the main users table to see whether the update was saved. If it was, check where the report reads from: a view, a reporting copy, a cache or a nightly ETL table. The report may simply be stale until the next refresh, which may be by design. If the main table still has the old value, check the API request, the update query and any errors in logs, then log a bug with both query results.
Check that each requirement is clear, complete, consistent and testable. Look for vague words like fast or user-friendly, missing error handling, undefined limits, conflicting rules and missing roles or states. For each requirement ask how you would test it and what the expected result is. Record questions in a review log and get answers from the business analyst before writing test cases, so ambiguity does not turn into defects later.
19
A security reviewer asks you to confirm that passwords and personal data are stored safely. What do you check in the database?
Query the users table after signup and confirm the password is not stored in plain text or reversible encoding, but as a salted hash from a slow algorithm such as bcrypt or Argon2. Two users with the same password should have different hashes. Check that card numbers are not stored at all or are tokenised, that personal data like Aadhaar numbers is masked or encrypted, and that logs and audit tables do not copy the raw values.
List conditions and actions in a table and cover every combination of conditions (rules) — ideal for business rules like loan approval: salary > X AND credit score > Y → approve.
Mutation testing measures the quality of a test suite. A tool makes small changes, called mutants, to the code, such as changing greater than to greater than or equal, and runs the tests. If a test fails, the mutant is killed; if all tests still pass, the mutant survived, which shows a weak spot in the tests. The mutation score is killed mutants divided by total mutants. It is strong but slow, so teams apply it to critical modules.
23
What is a regression bug and how do you track it in Jira?
A regression bug is a feature that worked before but broke after a new change, such as login failing after a profile page update. It is usually treated with higher priority because it affects existing users. In Jira, add a label such as regression, set the affected version and link it to the story or change that caused it. Tracking these labels over time shows where regression coverage is weak.
24
What is a screen reader and which ones are commonly used for testing?
A screen reader is software that reads screen content aloud or sends it to a braille display, used mainly by blind and low-vision people. Common ones are NVDA, a free Windows reader, JAWS, a paid Windows reader common in enterprises, VoiceOver, built into macOS and iOS, and TalkBack, built into Android. Testers usually cover NVDA with a desktop browser plus VoiceOver or TalkBack on mobile.
25
Product is running an A/B test on the checkout button. What does QA need to verify?
QA does not decide which version wins, but makes sure the experiment is trustworthy. Verify both variants work end to end, users are split in the agreed ratio, each user sees the same variant consistently across sessions, and analytics events fire correctly for each group. Also check that turning the experiment off returns everyone to the default. A broken tracking event can make the whole experiment's result wrong.
Log it as a high-severity accessibility bug because it fully blocks a group of users from signing up. An image-only CAPTCHA fails WCAG unless an alternative is offered. Recommend options such as invisible or risk-based bot checks, an email or OTP verification step, an audio alternative plus another modality, and rate limiting on the server. Retest the full signup with NVDA and TalkBack after the change.
28
Write a simple Playwright test in TypeScript that logs in and checks the dashboard heading.
Turn off the network and perform supported actions, such as creating, editing and deleting records, and check they are saved locally with a pending status. Reconnect and verify everything syncs once, in order, with no duplicates. Test conflicts, where the same record is changed on two devices, and confirm the agreed rule applies. Test sync after app restart, on a slow network and when the server rejects a change.
Vulnerability scanning uses automated tools to look for known weaknesses, such as outdated libraries, missing security headers or open ports. It is fast, repeatable and can run in the pipeline, but gives false positives. Penetration testing is a deeper, mostly manual exercise by security experts who try to exploit weaknesses and chain them like a real attacker. Teams scan regularly and run penetration tests before major releases or yearly.
33
The sprint ends tomorrow and two stories are still in testing. What do you do?
Tell the Scrum Master and product owner today, not at the review. Focus testing on the acceptance criteria and the riskiest paths of the story with the most value, and ask a developer to help with test data or simple checks. If a story cannot be finished with quality, it should be carried over, not marked done with untested parts. Raise the cause in the retrospective, such as late handover.
34
How do you verify database schema changes in a new release?
Get the migration script or change list. Compare the schema before and after: new tables and columns, data types, lengths, default values, NULL rules, indexes and constraints. Check that existing data was migrated correctly, for example a new status column filled for old rows. Run the main app flows to confirm queries still work. Finally, confirm the rollback script works on a copy, so the release can be reversed safely.
Give a clear title, such as 'Checkout: Place Order button not reachable by keyboard'. Include the page URL, browser, operating system and assistive technology with versions. Write exact steps, what was announced or happened, and what was expected. Add the WCAG success criterion, such as 2.1.1 Keyboard, and the user impact. Attach a screenshot or audio recording, and suggest a fix when obvious. Severity should reflect whether users are fully blocked.
36
Tell me about a time you found an important bug that others missed.
Model answer: Situation: before a release of an insurance app, all regression tests had passed. Task: I was doing a final exploratory round. Action: I renewed a policy on the last day of the month and saw that the premium was calculated for 31 days in a 30-day month. I logged it with exact data and the formula difference. Result: the date logic was fixed before release, and I added month-end cases to the suite.
37
What is a test scenario for the login functionality at a high level?
Model answer: Situation: two days before a release, we found 15 new bugs in checkout. Task: help the team decide what to fix. Action: I stayed calm, retested each one quickly, ranked them by user impact, and shared a clear list in triage with must-fix and can-wait items. Result: we fixed five critical bugs, released on time with documented known issues, and no critical issue reached users.
40
What is the difference between DELETE and TRUNCATE?
DELETE removes selected rows and can be rolled back; TRUNCATE quickly removes all rows and resets the table.
Tips to crack the Affine Software Testing interview
Practise SQL and Python data manipulation, and be ready to solve a small analysis problem using a sample dataset.
Revise statistics concepts like distributions, sampling and A/B testing, and explain how they guide real business decisions.
Prepare for case questions on topics like customer churn or pricing, structuring your approach before jumping into models.
Keep a one-minute story ready for every project on your CV: problem, your part, tools, result.
Practise the hard questions on this page aloud; Affine interviewers usually ask follow-ups on your first answer.
Frequently asked questions
How many rounds are there in the Affine Software Testing interview?
Candidates usually go through 4 stages: Online assessment (aptitude + SQL + statistics); Technical interview; Case study round; HR round. The exact rounds change with the role, team, location and your experience, so confirm the process with the recruiter.
Which topics should I prepare for the Affine Software Testing interview?
This practice set leans on Agile, SQL and API Testing (8), Performance and Security Basics (6), Testing Types (5), Bug Life Cycle and Jira (5). Revise those first, then practise explaining one or two of your own projects end to end, because most Affine interviewers ask you to walk through real work.
Are these the exact questions asked at Affine?
No. These are curated practice questions that match the skills and difficulty usually tested in analytics interviews like Affine's. They are not a leaked or verified list of past questions.
Where does Affine hire Software Testing professionals in India?
Affine has large teams in Bengaluru. Openings and locations change often, so check the Affine careers page and filter by city and experience.