Anatomy of a Jacard Analysis.
Each Analysis brings together a finding, the evidence, and the fix Jacard can run once you approve. Every number links to the raw data and compute behind it. This page shows the format, block by block, on a refund example.
View source appendix ↓The finding
The conclusion comes first, in plain language. No hunting through charts to find the point.
refund review
Auto flagged by JacardRefunds doubled last week. One supplier caused 83% of them.
The refund rate jumped from 2.1% to 4.6% over the last two weeks. Text-mining the refund reasons points at one inbound batch from one supplier: “stitching came apart”, “arrived cracked”. The spike started nine days ago, two days after that batch reached the warehouse.
written by Jacard · reviewed by you
Traceability
Every number carries its source, query, and latest sync time.
Refund rate
4.6%⌁q01
was 2.1% two weeks ago
Refunded last week
$18.4K⌁q02
up from $8.9K
From one supplier
83%⌁q03
supplier B, batch 2214
SKUs affected
3⌁q04
91% of the claims
The evidence
The chart shows exactly where the signal begins and what drives it.
The spike has one source.
Fig. 1 · Daily refunds for 21 days, split by supplier. The batch lands, the refunds follow.
View exact data · 21 days
| day | supplier B | others | total |
|---|---|---|---|
| day 1 | $0 | $430 | $430 |
| day 2 | $40 | $465 | $505 |
| day 3 | $0 | $395 | $395 |
| day 4 | $25 | $505 | $530 |
| day 5 | $0 | $440 | $440 |
| day 6 | $30 | $480 | $510 |
| day 7 | $0 | $410 | $410 |
| day 8 | $45 | $455 | $500 |
| day 9 | $410 | $435 | $845 |
| day 10 | $640 | $495 | $1,135 |
| day 11 | $860 | $460 | $1,320 |
| day 12 | $1,080 | $385 | $1,465 |
| day 13 | $1,260 | $470 | $1,730 |
| day 14 | $1,420 | $515 | $1,935 |
| day 15 | $1,590 | $445 | $2,035 |
| day 16 | $1,750 | $405 | $2,155 |
| day 17 | $1,720 | $490 | $2,210 |
| day 18 | $1,680 | $435 | $2,115 |
| day 19 | $1,590 | $475 | $2,065 |
| day 20 | $1,500 | $415 | $1,915 |
| day 21 | $1,440 | $450 | $1,890 |
sources⌁q02⌁q05
Actions
Jacard recommends the next move, then waits for your approval.
Pause the three affected SKUs
Takes them off the storefront and holds outbound orders until the QC check clears. About $2.1K/day of sales paused against $2.6K/day currently refunding.
pause math⌁q04⌁q05· runs only after a human approves
File the supplier claim
A drafted claim for $12.9K against batch 2214, with the refund evidence attached.
claim draft⌁q03⌁q04· runs only after a human approves
QC-hold the next inbound batch
Batch 2215 from the same supplier arrives in 6 days. This holds it for inspection before it ships to customers.
inbound schedule⌁q06· runs only after a human approves
Or describe your own action
Type what you want done. Jacard drafts the execution plan and waits for your approval.
The appendix
Every receipt resolves to the system and query that produced it.
Organizational memory
Approvals and corrections improve what Jacard investigates next.
It gets better with use.
Refund-reason mining now runs weekly for every new inbound batch because your team approved this one. When a signal clears the bar your team has set, the Analysis arrives before anyone asks. Approvals, corrections, outcomes, and your standards feed the same memory. It is cited so you can inspect it and editable so you can correct it.
Run it on your data.
Jacard builds Analyses like this from your own data. We're accepting a limited set of beta signups until July 31st.