Preparing content…

Two colleagues check a shopping-cart coupon feature against a test checklist
职业发展

Software Testing or Software Development? Try Both on the Same Feature

Try a coupon function, acceptance cases, and a regression script to see how software development, testing, and test automation differ.

By: FermatMind Editorial

Published: Nov 16, 2026

Updated: Nov 16, 2026

17 min read

Content review completed · Nov 16, 2026

Subscribe to the English blog (Atom) · Add this link to an RSS reader that supports Atom for newly published articles and substantive updates.

Start with software development if you want to turn requirements into maintainable features. Start with software testing if you want to turn vague promises into verifiable conditions and find where a product falls short. Both roles may write code, review requirements, and troubleshoot production issues. If you're unsure, implement and test the same requirement, then see which work you'd rather keep refining. There's also a repeatable regression script for a taste of test development. Check career prospects and compensation separately by location, team, and level.

Chinese campus recruitment provides a concrete example. One StarTimes posting for 2027 graduates lists feature development, requirements analysis, and design for the back-end role. It lists product verification, test tool development, and defect tracking for software testing. An industrial measurement job posting from Wuxi Yuanmiao lists test software, platforms, and frameworks as deliverables in their own right. The first posting compares two sets of responsibilities. The second shows what test development can look like in a specialized field. Hiring criteria and team responsibilities can vary. StarTimes graduate posting for software testing and back-end development · Wuxi Yuanmiao software test development posting

One user requirementWhat a software developer deliversWhat a software tester delivers
Requirements reviewExplain when the rule runs, where each field comes from, and how failures are returnedUse real scenarios to find missing rules. Ask what happens one cent below the threshold, at midnight, or after a cart change
Work you ownRunnable logic, the necessary tests, and handoff notesAcceptance cases others can run, defect records, and retest results
How to tell it's doneOutputs and reasons match the agreed rules. Someone else can run and change the codeExpected results, actual results, and version are recorded. Minimal steps reproduce the difference and allow retesting

One Coupon, Two Edge Cases: One Cent Short or One Second Late

Fictional requirement: A coupon takes 20 yuan off when eligible items total at least 100 yuan before the discount. Ineligible items and shipping do not count toward the minimum. The coupon is valid if the server accepts the order before July 1 at 00:00:00 Beijing time. It expires at that time. Previewing the discount does not redeem the coupon. At submission, the server recalculates using the current eligible subtotal and the time it accepts the order. Assume all other conditions are met. You can test orders separately, using an independent coupon in each case. “20 yuan off” refers to the coupon discount only.

First confirm who owns the business rule, when to recheck it, and what the customer will see. A coupon shown as available in the interface reflects a particular preview. The customer might change the cart or reach the expiry time before submitting. Show the coupon's final status based on the rule when the order is accepted. That's why testers still need to check what happens between a quantity change and submission after the developer's pure-function tests pass. ISTQB CTFL learning objectives for test activities and defect communication · Microsoft Learn on isolated, repeatable unit tests

Development Exercise: Write a Discount Function Someone Else Can Run

Write a local function that takes the eligible subtotal in cents and the time of the check. Return the discount in cents and a reason. Passing the time as an input lets you test without waiting for midnight. Loading the cart, storing the order, and redeeming the coupon can come later. For now, define the interface and identify where the final decision is made.

In this Python example, the check and expiry times are timezone-aware. Save the file as coupon_rule.py, then run python3 coupon_rule.py. No output means all three assertions passed.

from datetime import datetime

def coupon(eligible_cents, checked_at, expires_at):
    if checked_at >= expires_at:
        return 0, "Expired"
    if eligible_cents < 10000:
        return 0, "Below minimum"
    return 2000, "Available"

if __name__ == "__main__":
    before = datetime.fromisoformat("2026-06-30T23:59:59+08:00")
    end = datetime.fromisoformat("2026-07-01T00:00:00+08:00")
    assert coupon(9999, before, end) == (0, "Below minimum")
    assert coupon(10000, before, end) == (2000, "Available")
    assert coupon(10000, end, end) == (0, "Expired")

For this exercise, the expiry check takes priority over the amount check. On a real product, agree on the messages with product and testing first. Show both call sites: the preview displays the discount and reason. Submission calculates the current cart subtotal again and calls the function with the time the server accepts the order. If that check fails, this exercise sends the customer back to the cart to see the reason and confirm before trying again. The function itself does not redeem the coupon.

Development acceptance check: Submit the function, at least two assertions that expose incorrect implementations, and brief startup instructions. Ask a classmate to change the condition to require more than 100 yuan and see whether the assertion at exactly 100 yuan fails. If everything stays green, the tests are missing the boundary. Then ask your classmate to set the check time to midnight on July 1 and see whether the coupon shows as expired. Record both failure cases and the results after restoring the correct behavior.

Testing Exercise: Write the Expected Results Before Checking the Code

Write down the expected results from the requirements before looking at the implementation. You can run the cases manually. The two submission cases need an integrated flow. If you only have the local function, record “Page and redemption status unverified” under actual results. A paper test plan is fine at this stage. Leave those cases unmarked as passes.

ScenarioInputExpected result
One cent belowEligible items total 99.99 yuan, ineligible items 5 yuan, shipping 10 yuan. Coupon unexpired0 yuan off, with a “Below minimum” message
Exactly at the thresholdEligible items total 100.00 yuan. Coupon unexpired20 yuan off
One cent aboveEligible items total 100.01 yuan, shipping 0. Coupon unexpired20 yuan off. Amount due for the items is 80.01 yuan
Before and at expiryTwo independent orders each have a subtotal of 100 yuan. The server accepts them at June 30 23:59:59 and July 1 00:00:00 Beijing time respectively20 yuan off the first order. The second gets 0 yuan off and an “Expired” message
Cart changes after previewPreview succeeds at 100 yuan. Remove 1 yuan of eligible items and submit while the coupon is still unexpiredRecalculate at 99 yuan, say the condition is unmet, and leave the coupon unredeemed
Comparison case after previewEligible subtotal stays at 100 yuan. After preview, remove only a 1-yuan ineligible item or drop 1 yuan in shipping, then submit before expiryEligible subtotal stays at 100 yuan, so the discount should be 20 yuan. Verify submission and redemption together

Suppose your first run returns 0 yuan off at the exact threshold. Record the defect like this: “Local demo v1. Set the check time to June 30, 12:00:00 Beijing time. Use an eligible subtotal of 10,000 cents and request a preview. Expected: 2,000 cents off. Actual: 0 cents off and ‘Below minimum.’ Reproduced twice.” A developer can use this record to find the amount comparison branch. After the fix, rerun 99.99, 100.00, and 100.01 yuan. Add the version, expected results, and observed results to the acceptance table.

Testing acceptance check: Give a classmate the requirements and defect record and see whether they can reproduce the issue without your help. If they have to ask whether that 1 yuan was shipping or an eligible item, the inputs are incomplete. If the team later raises the minimum for a 20-yuan discount to 120 yuan, update and run the cases for 119.99, 120.00, and 120.01 yuan. Keep testing the expiry boundary. Update the case where you remove a 1-yuan eligible item so the subtotal falls from 120 to 119 yuan, and rerun the comparison that removes only an ineligible item. This shows whether your tests cover the logic as it changes.

Test Development Exercise: Give Your Colleagues a One-Command Regression Check

A developer has written the coupon function and a tester has documented the expected results. In test development, you can turn those checks into a regression script that runs on another computer. Name each case and print its expected result, actual result, and pass or fail status. If any case fails, the script returns a nonzero exit status. Save the following as check_coupon.py in the same directory as coupon_rule.py, then run python3 check_coupon.py.

from datetime import datetime

from coupon_rule import coupon

before = datetime.fromisoformat("2026-06-30T23:59:59+08:00")
end = datetime.fromisoformat("2026-07-01T00:00:00+08:00")
cases = [
    ("One cent below", 9999, before, (0, "Below minimum")),
    ("Exactly at threshold", 10000, before, (2000, "Available")),
    ("One cent above", 10001, before, (2000, "Available")),
    ("Expires at midnight", 10000, end, (0, "Expired")),
]

failed = False
for name, cents, when, expected in cases:
    actual = coupon(cents, when, end)
    passed = actual == expected
    failed = failed or not passed
    print(name, "PASS" if passed else "FAIL", "expected", expected, "actual", actual)
raise SystemExit(int(failed))

Ask a classmate to run it on another computer. Then deliberately change the condition to eligible_cents <= 10000. “Exactly at threshold” should fail, the remaining cases should still run, and the script should exit with status 1. In a team's code submission workflow, the exit status can indicate whether the regression run passed. The script passes fixed timestamps, so there is no need to change the computer's clock. It checks the function. Concurrency, the interface, cart changes, submission, and redemption still need separate checks. Submit the run command, an example failure, and the outstanding coverage gaps.

Some test development roles mainly deliver automation, test data, or test environment platforms. Others focus on verifying business features and using code to cut repetitive work. Ask a mentor or interviewer three questions:

  1. What does the team change most often? Product features, or scripts, environments, and platforms that let the team verify features repeatedly?
  2. Who reviews and accepts the work? Users' representatives, the product team, or developers and testers who use the platform and regression data?
  3. What was the first piece of work someone from last year's intake delivered independently? Ask how much of it involved code, test cases, defect records, or retesting. The answer tells you more about the job than debating which title sounds more senior.

After trying all three exercises, note which feedback you most wanted to act on before applying. To sort through the kinds of work you enjoy, try the FermatMind Career Test (94 questions). It explores work preferences. The actual job requirements determine what you submit and retest. Use the job description evidence checklist for applicants from unrelated majors to organize your work, or see how two roles can show their contributions in this portfolio example built around the same product.

Frequently Asked Questions

Do Software Testers Just Click Through Pages?

The role can include reviewing requirements, checking APIs or data, and developing automation tools. Find out what your target role owns over time, then prepare evidence of the corresponding skills.

Do We Still Need End-to-End Acceptance Tests if Developers Write Unit Tests?

Unit tests can check local logic such as amounts and times. Once the preview, cart updates, order submission, and coupon status are working, test how they interact.

How Do I Put an Undeployed Graduate Project on My Resume?

Describe it as a local simulation. Show a rule, an implementation, a reproducible discrepancy, and evidence from before and after the fix. State what you did not connect or test, such as real orders or coupon redemption. Label a paper check as a paper check.

Related tags

职业发展

Go to topic hubs

Continue from the article into a more structured topic entry surface.

Take the test

If you want to turn reading into self-measurement, continue into an assessment.

About FermatMind

FermatMind helps young people everywhere understand themselves and explore careers. We started in the Chinese-speaking world. Explore your preferences with MBTI and your interests with the Holland assessment, then use more than 1,000 occupation profiles to compare tasks, skills, and entry paths and find a next step worth trying.

Software Testing vs. Development for New Graduates | FermatMind