Kollabe
Retrospectives
Planning Poker
Standups
IceBreakers
Pricing
Dashboard

Posts

Why Testing Is the #1 Retro Complaint (and It's Not Flaky Tests)

Stylized flat illustration of a giant magnifying glass hovering over a wall of colorful sticky notes, with a few notes glowing and small characters inspecting them
Matt Lewandowski

Matt Lewandowski

Last updated 19/07/20267 min read

When we analyzed a million retro cards earlier this month, one result surprised everyone: testing and QA is the single most common complaint on software teams, showing up more than five times as often as communication. This post zooms in on that finding. What, specifically, are teams complaining about when they complain about testing? Same method as last time: aggregate analysis of retrospectives run on Kollabe, columns classified as positive or negative by their names, themes bucketed by keyword matching. No individual card is quoted or identifiable, and since keyword matching is English-centric and conservative, every number below is a floor. That gives us 29,015 testing-related complaint cards out of 269,372 complaints, written between October 2023 and July 2026.
29,015

testing & QA complaint cards

1 in 5

retros contain a testing complaint

0.9%

of them mention flaky tests

What's actually inside a testing complaint

Nearly one in five retrospectives (18.6%) contains at least one testing complaint. Here's how the 29,015 cards break down by sub-theme. Buckets overlap, since a card can mention both QA and automation:

QA as a function (capacity, handoffs)

23.9%

Automation and coverage gaps

8.5%

Test environments and test data

8.3%

Bugs escaping to production

4.9%

Regression testing

4.3%

Slow tests and CI

3.0%

No time to test

1.6%

Manual testing burden

1.4%

Flaky tests

0.9%

The headline: the internet's favorite testing complaint barely exists in retro rooms. Engineering blogs and conference talks are full of flaky-test war stories, yet only 257 of 29,015 testing complaints (0.9%) mention flakiness, instability, or intermittent failures. What teams actually write about is people and process. The largest bucket by far is QA as a function: nearly a quarter of all testing complaints mention QA directly. And when we look at what appears alongside the word QA, the pattern is about flow, not competence: handoff timing ("work reaching QA too late in the sprint") edges out capacity ("waiting on QA", "QA is stretched"), followed by unclear requirements and acceptance criteria. Teams aren't complaining about their testers. They're complaining about how and when work reaches them.

Rare complaints get the votes

Volume tells you what's common. Votes tell you what hurts. Across all complaint cards, the average is 0.65 votes per card, and the big testing buckets sit below it: QA-function cards average 0.59 votes, automation and coverage 0.58. Now look at the rare ones:
  • Flaky tests: 1.18 votes per card. The rarest sub-theme is the most voted, at nearly double the average. When someone finally writes the flaky-test card, the whole room piles on.
  • Manual testing burden: 0.85. Few cards, strong agreement.
  • Bugs escaping to production: 0.71. Above average, for obvious reasons.
The pattern mirrors what we found in the original analysis with workload and tech debt: the complaints teams write most and the complaints teams rally behind are different lists. A facilitator who only reads volume will spend the retro on QA handoffs. The votes say the room's real energy is wherever the flaky-test and manual-regression cards are.

Teams praise testing almost as much as they complain about it

Testing isn't a hated activity; it's a watched one. 23,056 positive cards mention testing too — for every five testing complaints, there are four testing compliments. Teams celebrate coverage improvements, successful release testing, and QA catching things before customers did. No other delivery-pipeline theme gets tracked this closely in both directions.

The AI era hasn't moved the needle (yet)

We expected testing complaints to climb as AI coding assistants pushed more code through the same human verification funnel. The data says no, or at least not yet: testing's share of all complaint cards has been flat between 10% and 12% every quarter since early 2024, the entire period in which AI-assisted coding went mainstream. Interestingly, the same is not true of code review, the other human checkpoint in the pipeline: review complaints have grown about 30% over the same window. We dug into that separately in the code review bottleneck.

What to do with this on Tuesday

Flat illustration of a team moving sticky notes from a chaotic pile into three tidy labeled lanes on a board, one character holding a small flag If testing keeps appearing on your retro board, the sub-theme data suggests where to look first:
  1. Fix when work reaches QA, not QA itself. Handoff timing is the most common thing written next to QA. Pulling testers into refinement and starting test design alongside development attacks the actual complaint. Our guide to writing acceptance criteria covers the clarity half of that problem.
  2. Treat the flaky-test card as an alarm. It's rare, and when it appears the room votes for it. That's the profile of a problem that's been quietly tolerated too long. Give it an owner and a date the same day; action items with an owner complete 2.7x more often.
  3. Don't let coverage complaints die of familiarity. Automation gaps are written constantly and voted on rarely, the classic profile of learned helplessness. If the same coverage card has appeared three retros running, it stopped being feedback and became a project.
If you want to see how your own board compares, the full aggregate numbers live on our retrospective statistics page, refreshed quarterly.

QA capacity and handoffs. Nearly 24% of the 29,015 testing complaints we analyzed mention QA as a function, and the words around it are about work arriving late or waiting in the queue — not about the testers themselves.

Much rarer than the discourse suggests: 0.9% of testing complaints. But they attract 1.18 votes per card, nearly double the average, making them the most-voted testing sub-theme in our data.

Not so far. Testing's share of retro complaints has stayed between 10% and 12% every quarter since early 2024. Code review complaints, by contrast, rose about 30% in the same period.

Aggregate, anonymized analysis of 1,081,076 cards from 87,719 sprint retrospectives run on Kollabe between October 2023 and July 2026. Columns were classified by name, themes by conservative keyword matching, and no individual card is quoted or identifiable. Queried July 19, 2026; the original analysis used a July 14 snapshot, which is why its totals are slightly smaller.