Composite case study
Ridgeline Supply Co. / Multichannel ecommerce
From a Monday spreadsheet to a daily inventory check
A 14-person DTC brand connected its sales and inventory data to a nightly reorder check while the ops manager kept every purchasing decision.
The business
A concise operating snapshot before the implementation started.
- Annual sales
- About $9 million
- Team and range
- 14 people and 45 SKUs
- Sales channels
- Shopify, Amazon, and wholesale
- Supplier lead time
- 75 to 90 days
Where the workflow broke
The reorder decision lived in a Google Sheet updated by hand, usually on Mondays and sometimes not for two weeks. The best-selling cookware set sold out on Amazon before the sheet caught the demand change, and a competitor took its search position during the 11-week wait for new stock.
Ordering heavier to avoid another stockout created the opposite problem. About $340,000 was tied up in slow-moving inventory. The team already had Claude Code, but it was only being used for one-off scripts.
What we built
The work started by mapping where the reorder decision really lived, then connecting the tools the team already used instead of adding another subscription.
Map the real inputs
Bring together Shopify and Amazon sales, 3PL inventory, supplier minimums, and lead times that had fallen six months out of date in the spreadsheet.
Build the nightly check
Use Claude Code to calculate rolling sales velocity and reorder points for every SKU and channel, including lead-time variance and seasonal demand.
Turn exceptions into one clear view
Show days of stock beside the reorder point on one internal page, then send a Slack alert when a SKU crosses its line.
Tune the alerts and transfer ownership
Replace the noisy flat threshold with SKU-specific seasonal patterns, then spend two weeks helping the ops manager understand and extend the system herself.
How it was checked
Codex independently backtested the pipeline against three months of sales and inventory history. It caught a real bug: units reserved for an incoming Amazon FBA shipment were being counted as available stock, which would have delayed a reorder trigger by about a week. The logic was fixed before it touched a live decision.
What stayed human
The system checks the same signals every night and surfaces a reorder point. The ops manager still applies judgment, approves each purchase, and can change the seasonal rules when the business changes.
The engagement
- Timeline
- About eight weeks
- Investment
- Roughly $28,000
- Context
- Less than the company had paid a freelancer the year before to keep patching the same spreadsheet macros.
Measured results
Two reorder cycles in
- Reorder check frequency
- BeforeWeekly, when time allowed
- AfterEvery night, automatic
- Stockouts on the top 20 SKUs
- BeforeAbout one every 5 to 6 weeks
- AfterZero since launch
- Cash tied up in slow-moving stock
- Before$340,000
- After$215,000
- Ops time on inventory
- BeforeAbout 6 hours a week
- AfterUnder 1 hour a week
The cookware set that had stocked out the previous year reached the next camping season with stock intact. The new logic flagged its seasonal ramp in June, inside the supplier's lead time.
Why it held up
The tool did not replace the ops manager's judgment. It gave her the same read on inventory every day and made the slow changes in demand visible early enough to act. The business already owned the AI; the missing part was wiring it into the operating decision.
Have a workflow with the same kind of friction?
We can map the real process, identify the review points, and agree on what evidence should count.