JSM export reconciliation runbook
Wezy CSV vs Jira Dashboard: Reconcile Customer Demand
Investigate differences between a Wezy voting CSV and its JSM dashboard. Match scope, timestamps and request identities before comparing counts or demand scores.
In short
Compare the same accessible request set at documented times. Check filters and pagination first, reconcile request identifiers second, and only then compare the demand metrics or historical periods. Equal row counts alone prove nothing.
In this guide
When a Wezy CSV differs from the JSM demand dashboard, reconcile scope and request identity before comparing totals. The displayed page, a filtered export, published Jira fields and a historical comparison are not automatically the same dataset. A difference can be explainable without being dismissed as harmless.
This runbook investigates a mismatch. For turning a reconciled export into an executive narrative, use the separate monthly voting report guide.
Preserve the evidence before refreshing everything
Keep the downloaded file unchanged in a restricted location. Record when it was prepared and received, which account exported it, the project, edition, active scoring policy and applied filters. Capture the dashboard state without including private request details in a broadly shared screenshot.
Then identify what the comparison actually is: current Wezy dashboard versus current CSV; historical period versus current values; or Wezy panel versus Jira fields/search. Those require different explanations. The official export guide says analytical CSVs follow the applied filters and the exporting account’s permissions. It does not promise an unfiltered dump of every request on the site.
Step 1: make the scopes comparable
| Dimension | Check on both sides | Common source of a false mismatch |
|---|---|---|
| Project | The same service project | Comparing similarly named projects |
| Request filters | Type, status, minimum votes and score | A saved mental picture of old filters |
| Advanced filters | Organization/segment when used | One view selects a narrower audience |
| Account and access | The actual viewing/exporting role | An administrator sees more than the analyst |
| Pagination | Whether all relevant pages were inspected | Treating the first screen as the total set |
| Time | Preparation, receipt and observation times | Votes or settings changed between observations |
| Period | Current state or complete covered UTC dates | Comparing rolling/current values with a past snapshot |
Use Show more requests where needed and Refresh results deliberately, noting the new observation time. An empty filtered ranking is not evidence of no demand across the site. These behaviors are described in Wezy’s dashboard documentation.
Do not widen the exporting account’s permissions just to force parity. Instead, compare the authorized intersection or ask an appropriately authorized owner to investigate the broader scope.
Step 2: compare request identities, not just row counts
Two datasets can both contain four requests and still disagree:
Dashboard set: JSM-101, JSM-102, JSM-103, JSM-104
Export set: JSM-101, JSM-102, JSM-104, JSM-105
Only in dashboard: JSM-103
Only in export: JSM-105
This fictional example demonstrates why “four rows on each side” is not a reconciliation. Match the actual request identifier, not the title: titles can change or be repeated.
Inspect the real downloaded schema before counting. A file with several analytical rows per request needs its dimensions understood before it can become a one-row-per-request comparison. Do not sum repeated current vote totals or silently keep the first row to make duplicates disappear.
For a minimal check, normalize copies into an explicitly team-defined format containing request_key and votes. These are worksheet column names, not claimed Wezy export headers. Preserve the raw file and document how the mapping was made.
Run a local comparison on explicitly normalized copies
Download the local reconciliation helper. It uses Python’s standard library, makes no network requests and does not modify either input. First map copies of your evidence into this exact worksheet contract:
request_key,votes
JSM-101,8
JSM-102,0
JSM-103,
The blank value remains unknown; it is not converted to zero. Run python jsm-demand-reconcile.py observed.csv exported.csv. The output lists request keys found only on one side, differing vote values and unknown values. Duplicate keys, malformed rows and invalid counts fail with an error instead of being silently summed.
A request title or a guessed source column must not replace the actual request identifier. The helper deliberately refuses to infer Wezy’s native export schema. Check the manual mapping and scope first; matching normalized files do not prove that the original export was complete or captured at the same instant.
Step 3: compare metrics only on matched requests
For each shared request, compare current V, O, R30 and the applicable score under the same policy. A vote count and an organization count are different quantities. The weighted organization metric is not applicable in Standard, and a blank or unavailable value must not be turned into zero.
A simple investigation table is enough:
| Request | Difference observed | Evidence needed next |
|---|---|---|
| Present only on one side | Membership mismatch | Filters, status, access and observation time |
| Same request, different V | Active support mismatch | Current saved state and export timing |
| Same V, different O | Attribution mismatch | Represented organizations under the same scope |
| Same inputs, different S | Policy or calculation mismatch | Active formula/version and applicable weights |
| Current values versus historical values | Period mismatch | Complete UTC coverage and original formula context |
Do not label every discrepancy “indexing delay.” That explanation is relevant when Jira fields or search lag the Wezy panel; it does not by itself explain a mismatch between every possible pair of app views.
Step 4: handle historical data as a separate comparison
Wezy Advanced can add supported historical dimensions when the selected UTC dates are completely covered. Changing today’s weights does not retroactively make earlier calculations use the new formula. Exported organization names reflect currently observed names, not certified historical names. These constraints are documented in history and trends.
A team using local midnight for one period and UTC midnight for another has not compared identical boundaries. Record the dates rather than presenting an unexplained percentage change. If either period is unavailable, retain an explicit missing state and stop that part of the comparison.
Re-prepare a controlled export
Once the intended scope is known, use Export requests → Prepare CSV → Save CSV. An expired local download link should be prepared again; it is not a reason to change permissions or reinstall the app. Permissions and licensing are checked when the file is delivered.
Compare the newly received file against a newly recorded dashboard observation. If customers can continue voting during that interval, document that limitation. Do not claim a transactional point-in-time snapshot across independently refreshed interfaces unless you have evidence for it.
Close the investigation with a reason, not a forced match
Classify the outcome as explained scope difference, explained time difference, explained policy/period difference, or unresolved discrepancy. Record the affected request identifiers, owner and next action. A second export that happens to match does not prove why the first differed.
For an unresolved case, send Wezy support a sanitized reproduction: app version, edition, role, filters, timestamps and the mismatching metric. Do not initially attach a full personal-data export or private customer conversation.
The analytical export in Wezy Customer Voting is useful when its scope remains explicit. Administrative JSONL recovery is a different operation; follow the recovery checklist rather than treating a reporting CSV as a complete backup.
Published by the team behind Wezy Customer Voting. Product capabilities are linked to their official documentation; worked examples are illustrative.
Let customers vote on your JSM portal.
Install Wezy from Atlassian Marketplace, open one request for voting and see the first vote in Jira.