The Jira Gantt app data-safety checklist

Gantt apps get write access to your Jira. Most trials test the features; almost nobody tests what the app does to the data. I read 500+ Marketplace reviews across this category — issues deleted with no recycle bin, dates mass-reset on install, weeks of planning lost to a missing save button. Every test below recreates one of those incidents. Run them in the first hour of any trial — this page works for every Gantt app, including mine.

Before you start: create a throwaway test project — 20 or so issues, a few subtasks, some with dates, some without, a couple of "Blocks" links. Never point a Gantt trial at a live project first. Note the "Updated" timestamps of a few issues; you'll need them in test 1.
0/10 tests passed · progress is saved in your browser

Why: reviewers report an app updating the last-modified date on thousands of issues before first use, wrecking reports and audit trails (Dec 2024, May 2024).

Do: install the app, open its Gantt view once, click nothing else. Go back to Jira and check the "Updated" timestamps you noted.

Pass: every timestamp is exactly as you left it.

Why: "removing" a project from one app's view actually deleted the Jira issues, with no trash bin (Jun 2026).

Do: use whatever "remove", "exclude", or "delete from view" option the app offers on a test issue. Then search for that issue key in Jira.

Pass: the issue still exists. Bonus: the app's wording made clear it was only leaving the view.

Why: some apps write to Jira on every drag with no confirmation; reviewers report dates overwritten they never meant to change (Mar 2024, Feb 2022).

Do: drag one bar a few days. Without saving or confirming anything, open that issue in Jira.

Pass: Jira is unchanged until you explicitly confirm — and you can see exactly which fields will be written before they are.

Why: "reload the page" was one vendor's official advice for errors — losing all unsaved work every time (Sep 2025, Sep 2020).

Do: make several unsaved changes, close the tab without saving, reopen the app.

Pass: your in-progress changes are still there. There is no save button to forget.

Why: after an unwanted mass change, one reviewer "had to reset the timelines for each of the activity manually" (Feb 2022).

Do: apply one date change to Jira through the app. Now put it back using only the app.

Pass: one-click undo restores the exact previous values — including "no date" if the field was empty before.

Why: sync is two-way trouble; reviewers report apps resetting all start/due dates "without any explicit consent" (Feb 2022, May 2020).

Do: change a due date in Jira itself, then reopen the Gantt app and wait for it to refresh.

Pass: the app shows the new Jira value. It never "corrects" Jira to match its own stored copy.

Why: reviewers report structure and view settings gone after reopening the page — hierarchy flattened, charts vanished after saving (Sep 2020, Jul 2019).

Do: configure zoom, collapsed groups, and column layout. Close everything; reopen the next day (or in a private window on another machine).

Pass: the view is exactly as you left it.

Why: one app demanded third-party cookies for LinkedIn, Facebook, and Google Tag Manager domains just to run (2020–2022); others store your project data on the vendor's own servers.

Do: open DevTools → Network while using the app. Note every non-Atlassian domain. Cross-check the vendor's privacy page: where is data processed and stored?

Pass: no trackers, no surprise domains, and a clear answer on data residency. On the Atlassian Marketplace, "Runs on Atlassian" means data never leaves Atlassian's infrastructure.

Why: reviewers abandoned apps over 2–3 seconds of lag per change and initial loads measured in hours (2020, 2016).

Do: after the safety tests pass on the throwaway project, point the app at your largest real project (read-only!). Time first load, then drag a bar and count the delay.

Pass: first view in seconds, drag response effectively instant at your real project size.

Why: averages hide decay — an app can hold a decent lifetime rating while recent reviews say "it only worked for 3 months out of the 6" (May 2025). And when something breaks, support speed is the product.

Do: sort reviews newest-first and read ten. Then send support one real question and start a timer.

Pass: recent reviews match the average, and a human answers your question in one business day.

Why I wrote this

I built SteadyGantt after reading those 500+ reviews — it's designed to pass every test above (read-only until you confirm a diff, everything autosaves, one-click undo, no external servers). But the checklist is deliberately tool-agnostic: whichever app you pick, run these tests first. The hour costs nothing; the incidents above cost real teams real data.

You can run tests 1–8 against SteadyGantt right now.

The live demo is the full app with sample data — no install, no signup.

Try the live demo →