Task walkthrough · MuninCompare
Compare two HAR files to find GA4 and Adobe Analytics changes after a release
A frontend release can quietly drop an analytics event, change a parameter or fire a tag twice. Record the same journey before and after the change, compare the two HAR files on your own computer and review only what changed.
A Perunlight product workflow for MuninCompare, our paid Chrome extension. Written with AI assistance and checked against the accepted MuninCompare 1.0.1 test run (Chrome for Testing 151 on Linux, fictional HAR files). Published .
Choose a comparison route before installing anything
Start with the task you need to complete. The free practice resources and the paid Chrome extension are separate options with different scopes.
Learn the comparison vocabulary without installing a tool
Use the interactive lesson for fixed fictional captures. It demonstrates the example; it does not load your own HAR files.
Compare local GA4 captures with a free command-line tool
Choose the free CLI practice kit if its narrow GA4 scope fits and you can run Node.js. This separate MIT-licensed tool was tested with Node.js 22.22.2; it is not the MuninCompare extension and does not cover Adobe Analytics or batches.
Read the extension workflow before purchasing
Download the free illustrated manual and four fictional HAR files. This ZIP contains no extension. Reading it does not require purchase.
Use the graphical Chrome extension for your own authorized captures
MuninCompare costs $0.99 once. Read the setup guide first: installation is unpacked through Chrome Developer mode, which a managed browser may block. The extension supports its documented GA4 and Adobe Analytics workflow; a report is not proof that tracking is correct.
Watch the 46-second comparison
If the extension is already installed, allow about five minutes to work through this example. Installation is a separate step covered by the setup guide.
- Download and extract the free illustrated manual and examples ZIP. It contains four fictional HAR files and the manual; it does not contain the extension.
- Load
changes-before.haron the left andchanges-after.haron the right. - Compare the summary: removed 1, added 1, changed 3, unchanged 4 and repeated 1.
- Use Explain to inspect the removed
newsletter_signup, theadd_to_cartcount going from 1 to 2, and thecucurrency change from EUR to USD. The recording temporarily enables redacted values to show that currency change. - Turn redacted values off, keep property IDs off, and export the default redacted HTML report. Open it locally and review it before sharing.
The recording uses newly opened fictional files and makes no analytics requests. It demonstrates the accepted interface and result format; it is not evidence of a customer using the product or of events reaching an analytics service.
When a before/after comparison helps
Use it when a change could touch tracking although nobody meant to change analytics: a redesign, a checkout refactor, a consent banner update or a tag manager clean-up. You want a short list of differences to review, not every request the page made.
MuninCompare reads the standard GA4 and Adobe Analytics collection requests in two HAR files, groups them by event and reports what was removed, added, changed or repeated. Other traffic is counted but not analysed.
Explore three fictional tracking changes
Free practice · Fictional data
Open a case, compare the requests, then reveal what to investigate. These are fixed examples from the free HAR practice kit.
01Missing event1 → 0 requests
An event disappeared
The same add-to-cart step is meant to happen in both journeys. Compare the request counts.
Before
baseline.harpage_view- 1 request
view_item- 1 request
add_to_cart- 1 request
4 total requests · 3 supported
1 outside scope · 0 unsupported · 0 malformed
After
missing.harpage_view- 1 request
view_item- 1 request
add_to_cart- 0 requests
3 total requests · 2 supported
1 outside scope · 0 unsupported · 0 malformed
Show what to investigate
add_to_cart appears once before and zero times after. The CLI reports removed.
Next check. Check that both journeys included the same click and consent choices. Then inspect the tag or handler on the changed page. The files alone cannot tell you why the request disappeared.
02Repeated request1 extra repeat
One request became two
The fictional after file contains an extra identical add-to-cart request. Find the row that changed.
Before
baseline.harpage_view- 1 request
view_item- 1 request
add_to_cart- 1 request
4 total requests · 3 supported
1 outside scope · 0 unsupported · 0 malformed
After
duplicate.harpage_view- 1 request
view_item- 1 request
add_to_cart- 2 requests
5 total requests · 4 supported
1 outside scope · 0 unsupported · 0 malformed
Show what to investigate
add_to_cart grows from one to two supported requests. The CLI reports changed and repeated_after: 1: one extra identical occurrence, not one event group.
Next check. Inspect the click handler and tag triggers for an unintended second send. Repetition may also be intentional. This request evidence does not prove that Analytics recorded a duplicate conversion.
03Changed parameterEUR → USD
Same count. Different currency.
The number of requests stays the same. Look at the currency annotation on view_item.
Before
baseline.harpage_view- 1 request
view_itemFictional cu: EUR- 1 request
add_to_cart- 1 request
4 total requests · 3 supported
1 outside scope · 0 unsupported · 0 malformed
After
parameter-change.harpage_view- 1 request
view_itemFictional cu: USD- 1 request
add_to_cart- 1 request
4 total requests · 3 supported
1 outside scope · 0 unsupported · 0 malformed
Show what to investigate
view_item stays at one request, but its fictional cu parameter changes from EUR to USD. The CLI reports changed; it does not print those parameter values.
Next check. The currency annotations here come from the authored HAR files. Check whether the currency change was intended, then inspect the payload. Equal request counts alone would miss this difference.
Each supported request in these examples carries one event. The narrow GA4-style comparison excludes Adobe, batches and custom collectors. The files use invented .example hosts and responses; these are not browser captures. Request counts do not prove delivery to Analytics, tracking correctness or consent.
Four fictional HARs, a local Node CLI, source, tests and a worked lesson. Run the examples yourself with Node; tested on Node 22.22.2 / Linux. This browser example is a teaching diagram, not the MuninCompare extension.
Try three changes with the free HAR practice kit
Download and extract the free MIT-licensed practice kit, then open a terminal in its folder. It contains a local Node CLI, source and tests, a README lesson, a diagram and four newly authored fictional HAR files. It runs without a purchase or account. Node must already be installed; the tested version is 22.22.2 on Linux.
node har-diff.mjs fixtures/baseline.har fixtures/duplicate.har
node har-diff.mjs fixtures/baseline.har fixtures/missing.har
node har-diff.mjs fixtures/baseline.har fixtures/parameter-change.har- Duplicate case: add_to_cart changes from 1 to 2 supported requests; repeated_after is 1 extra identical occurrence.
- Missing case: add_to_cart is removed.
- Parameter case: view_item is changed. Open the fictional HAR to see the cu value change from EUR to USD; the CLI report withholds parameter values.
These practice files are separate from the Northwind Tea examples below. They use reserved .example hosts and invented responses, not browser captures. The CLI reads a narrow GA4-style query/form format with one event per supported request. It counts unsupported, malformed and outside-scope requests. Adobe, GA4 batches and custom collectors are outside its coverage. See the bundled README for exact rules, error messages and the full example output.
The CLI compares files locally without uploading them or rewriting them. It withholds raw URLs, parameter keys and values, and custom event names from its report. It is a teaching aid with an 8 MiB / 20,000-entry limit per file; real customer captures and other platforms are unverified. Request evidence alone does not prove delivery, correctness or consent. Review even the reduced report before sharing.
Download the free HAR practice kit. This kit was written with AI assistance by the Codex coordinator and independently reviewed before publication. MuninCompare remains a separate paid graphical extension; its illustrated manual and example ZIP are linked at the end of this article.
Capture two comparable journeys
- Pick one short, scripted journey on a site you are authorised to test, for example home, product page, add to cart, basket. Use a test account and test data.
- In Chrome, open DevTools, select the Network panel and tick Preserve log so requests are kept across page loads. Walk the journey on the current build.
- Save the capture with Export HAR (sanitized). According to Chrome’s DevTools documentation, the sanitized export leaves out the
Cookie,Set-CookieandAuthorizationheaders. - Repeat exactly the same journey on the new build, with the same consent choices and the same browser profile, and export a second HAR.
- Name the files so they cannot be swapped, for example
checkout-before.harandcheckout-after.har.
The comparison is only as good as the two journeys. MuninCompare cannot tell whether they were equivalent: an extra click or a different consent choice shows up as a difference the release did not cause.
Treat HAR files as sensitive
A HAR file records what the browser sent and received. Even a sanitized export can still contain full URLs and query strings, form fields, response bodies and analytics parameters that identify a person. Store it like any other sensitive file and delete it when you are done.
MuninCompare reads the files in its own tab, in memory. The extension has no permissions or host access and makes no network requests. While parsing, it discards headers, cookies, response bodies, the bodies of non-analytics requests and full URLs. A report leaves the tab only when you press Download, and by default contains counts, redacted event names, parameter names and changed/unchanged markers.
Redaction replaces common identifiers such as e-mail addresses, URLs, tokens, UUIDs and client IDs, but it cannot recognise every kind of personal data, for example a name inside a page title or a custom eVar. It is not anonymisation: review every report before you share it.
Read the comparison
- Removed: an event fired before the change and not after. The most likely sign of a broken tag, so review these first.
- Added: an event that only fires after the change. Expected if the release added tracking; otherwise find out where it came from.
- Changed: the same event with a different parameter value or a different number of occurrences. Explain shows what differs.
- Repeated: identical occurrences of one event in a capture, a possible double firing.
- Unchanged: the same events with the same compared values.
- Coverage: what was not analysed. Adobe Target, Adobe Web SDK and requests to custom collection domains are listed as unsupported, and ordinary page traffic is counted as unrecognised. Unsupported does not mean clean.
Volatile values such as client and session IDs, timestamps and cache busters are ignored, and so is the order of requests. Two runs of the same journey can therefore come out unchanged.
Worked example: a fictional checkout journey
MuninCompare includes fictional HAR files for an invented shop, Northwind Tea, on the reserved domain tea.example.com. The before file records home, product, add to cart and basket: 18 requests, including GA4, Adobe Analytics, Adobe Target, a server-side style GA4 request, a look-alike host and ordinary page requests. The after file records the same journey after a fictional tag change.

What the comparison reports
- Removed 1: GA4
newsletter_signupfired before the change, not after. - Added 1: GA4
view_cartfired only after the change. - Changed 3: GA4
view_item(the currency parametercuchanged; with redacted values shown, EUR → USD), GA4add_to_cart(one occurrence became two) and an Adobe page view whoseeventsvalue changed on the product page. - Repeated 1:
add_to_carthas two identical occurrences in the after capture, a possible double firing. - Unchanged 4: GA4
page_view,scrollandbegin_checkout, and the Adobe add-to-cart link call. - Coverage: in each capture, one Adobe Target request and two GA4-style requests on non-standard hosts are listed as unsupported and not analysed; ordinary page requests are counted as unrecognised (5 before, 6 after).
Client IDs, session IDs, timestamps, cache busters and the order of requests differ between the two files and did not become differences. A second fictional pair records the same journey twice with new volatile values in a different order: all 8 event groups come out unchanged.
Result in the accepted 1.0.1 test run: for the changed pair, removed 1, added 1, changed 3, unchanged 4 and repeated 1; for the reordered identical pair, no differences. The redacted HTML report opens offline and contains no scripts.
Work through the results in this order
- Removed events: likely regressions.
- Changed parameters on key events such as add to cart, sign-up or purchase.
- Repeated events: possible double counting.
- Added events: confirm someone meant to add them.
- Coverage: anything unsupported needs a separate check.
Common mistakes
- Comparing different journeys. Different pages, clicks or consent choices create differences the release did not cause.
- Capturing with blocking extensions on. A blocked collection request looks like a removed event. Use the same clean test profile for both captures.
- Sharing the raw HAR. Share the reviewed report instead, and only with people entitled to see the analytics data.
- Reading “no differences” as “tracking is correct”. The comparison shows change, not correctness or consent compliance.
- Ignoring coverage. Server-side tagging hosts, Adobe Web SDK and Adobe Target are not analysed; check them another way.
Free and manual alternatives
- Chrome DevTools on its own: filter the Network panel to your analytics requests and read each payload in both builds. Fine for one or two events; slow and easy to get wrong for a whole journey.
- GA4 DebugView: Google’s DebugView shows the events and user properties Analytics collects from a device in debug mode, in real time. It helps while you test one build, but it is not a before/after comparison.
Limits and next steps
- MuninCompare is installed as an unpacked extension in Chrome with Developer mode on. It is not a Chrome Web Store listing, and a managed browser policy may block it.
- It was tested in Chrome for Testing 151 on Linux with fictional HAR files. Other browsers, other operating systems, real customer journeys and HAR exports from your own sessions were not part of that test.
- Only standard GA4 and Adobe Analytics collection requests are analysed. Adobe Web SDK, Adobe Target and custom collection domains are counted as unsupported.
- It does not capture traffic, connect to analytics services, certify tracking or consent, or prove that two journeys were equivalent.
Next: if you have MuninCompare, load the fictional pair first to see the result format, then capture one short authorised journey before and after your next release. The MuninCompare setup guide covers installation, loading files and exports.
Perunlight product
Compare your own captures with MuninCompare
MuninCompare is a paid Chrome extension: $0.99 USD, one-time purchase, version 1.0.1, installed unpacked rather than from the Chrome Web Store. It compares two local HAR files and exports redacted HTML, CSV or JSON reports. Tested in Chrome for Testing 151 on Linux.
Get MuninCompare · $0.99Read the setup guide
This illustrated-manual ZIP contains no extension. The MuninCompare graphical comparison needs the paid extension; the separate CLI practice kit above is free.