Exploratory Testing with Charters and Session Notes
On this page
Exploratory testing is learning about the software, designing tests and running them at the same time, each informing the next. You notice the date field accepts typing as well as the picker, so you try typing 31 February; that leads you to wonder about the time zone of the confirmation email, and so on. It is the most effective way to find bugs in new and changed features, and also the easiest to do badly, because "I clicked around for an hour" is impossible to review or repeat.
The fix is a light structure that keeps the freedom: a charter that says what you are exploring, a time box, and notes that show what you covered.
Writing a charter#
A charter is a one-sentence mission. A widely used template is:
Explore target with resources to discover information.
Examples:
- Explore the new split-payment option with amounts at and around the card limits, and a declined second card to discover how partial failures are handled and shown.
- Explore the address form with long, non-Latin and oddly formatted addresses from our top five countries to discover validation and layout problems.
- Explore the order history page with an account that has 300 orders and one that has none to discover performance and empty-state issues.
A good charter is narrow enough to finish in one session and broad enough to leave room for judgement. "Explore checkout" is too broad: you will skim everything and dig into nothing. "Check that the coupon field accepts SAVE10" is too narrow: that is a test case.
Where do charters come from? The test ideas from your test design (chapter 1), the risky parts of a new ticket, recent bug reports in the same area, customer support tickets, and questions developers could not answer during refinement ("what happens if the payment provider times out?").
Which is the best charter for a 60-minute session on a new file upload feature?
The first names a target, the resources and the kind of information wanted, and fits in an hour. The second is so broad that nothing is explored in depth; the third is a single scripted check with no room to explore.
Time-boxed sessions#
Run exploratory testing in uninterrupted sessions of 45 to 90 minutes, one charter per session. This approach, known as session-based test management, makes exploration plannable ("we need about six sessions for the new checkout") and reviewable.
During a session:
- Stay on charter, but note detours. If you spot something interesting outside the charter, write it down as a new charter instead of following it. That is how you build the next session's plan.
- Vary one thing at a time where you can, so that when something breaks you know what caused it.
-
Use heuristics to generate ideas when you run dry. A few that work on almost any feature:
- Goldilocks: too big, too small, just right (lengths, quantities, file sizes).
- Interrupt: go Back, refresh, close the tab, lose the network, let the session expire, mid-flow.
- Repeat: double-click, submit twice, open the same item in two tabs.
- Zero, one, many: no items, exactly one, more than a page full.
- Follow the data: create something, then find every place it appears (lists, emails, exports, search) and check it is the same everywhere.
- Change the user: a different role, a different language, a different time zone.
- Stop when the time is up, even mid-idea. Write the idea down.
Session notes#
Notes are what turn an hour of clicking into evidence. Keep them as you go, in plain text, using a few fixed headings:
CHARTER Explore profile picture upload with very large, very small,
wrong-type and renamed files to discover how limits and
errors are handled
TESTER Anna DATE 2026-11-12 DURATION 60 min
SETUP Staging build 4.18.0, Chrome latest on Windows 11,
Safari latest on macOS
NOTES
- 9.8 MB JPEG: uploads, preview OK (limit is 10 MB per help text)
- 10.2 MB JPEG: rejected, message "File too large (max 10 MB)" - good
- 0-byte file named a.jpg: spinner never stops, no error -> BUG-1
- virus.exe renamed to virus.jpg: rejected as "Not an image" - good
- HEIC from iPhone: rejected on Chrome, accepted on Safari, preview
blank on Chrome afterwards -> BUG-2, asked dev whether HEIC is supported
- Cropping dialog: Esc closes it but keeps the uncropped image (question)
BUGS BUG-1 empty file hangs upload; BUG-2 HEIC inconsistent
QUESTIONS Is HEIC supported? Should Esc in cropper cancel the upload?
NEW IDEAS Explore upload on slow network (3G throttling)
Explore replacing an existing picture
TIME SPLIT 70% testing, 20% bug investigation, 10% setup
Three parts matter most:
- What you covered (and, implicitly, what you did not). A reviewer can see that nobody tried replacing an existing picture.
- Bugs and questions, kept separate. A question is not a bug until someone confirms the expected behaviour; "Esc keeps the uncropped image" might be intended.
- New ideas, which feed the next sessions.
The debrief#
After the session, spend ten minutes going through the notes with the developer or another tester. The debrief is where questions get answered (the developer might say HEIC is meant to be rejected everywhere, which turns BUG-2 into "Safari accepts a format we do not support"), bugs get filed, and the next charters get prioritised. It is also the cheapest training a new tester can get.
Halfway through a session on the upload feature, you notice the notification settings page has a typo and a toggle that does not save. What should you do?
Off-charter findings are valuable, but following every detour means the charter never gets covered and nobody knows what was tested. Writing it down keeps the information and turns it into planned work.