Retail
Store rollouts that still run on checklists and handoffs
We turn store upgrades and releases into pipelines your team starts, with a check after every stage and a tested rollback. At a global retailer with 900+ stores, one self-service upgrade pipeline cut store-server upgrades from 6 to 8 hours to about 3 per server.
- Your tools stay
Store release 2.14
Started by the team at 07:30
- Wave 101040117012301310146015201680175
- Wave 202030219022402380241025702630286
- Wave 304050412041804270433044904510470
- Wave 405020514052605310547055805620589
- Wave 506070611062906340640065306710688
- Wave 607150722073607430759076407810797
Release bot#store-releases
Wave 2 passed its checks, so wave 3 started. Stores in later waves keep trading until their turn.
What slows releases
Where retail releases get stuck
Store upgrades pass from team to team
One team prepares the package, another books the store windows, and a third runs it server by server. Every handoff is a wait.
Only a few stores can be upgraded at once
When every upgrade is run and watched by hand, the team can only handle a few stores at a time. A rollout to the whole fleet drags on, and stores run different versions in the meantime.
Peak season makes every change a risk
Before peak, nobody wants to touch checkout or the stores. Changes queue up behind the freeze and go out together afterwards, in one release that is harder to check and to roll back.
What changes
From a checklist between teams to a pipeline your team starts
A Self-Service Release Build automates the release flows that cost you the most, end to end, and hands them to your team. Plan on 4 to 8 weeks.
-
TodayEach team runs its part of a store upgrade by hand, then passes it to the next.
After a BuildHand-run steps and handoffs become one pipeline per release flow, started by your team.
-
TodaySomeone watches each upgrade and decides by hand whether the store is fine.
After a BuildEvery stage checks its own result, and rollback is tested before go-live.
-
Today“Which stores have the new version?” gets asked in chat, on calls and in the stand-up.
After a Build“Where is it now?” gets answered by status messages in chat or the ticket.
-
TodayThe full procedure sits with one or two people, and they are needed most at peak.
After a BuildThe one person who knows how becomes a runbook, a trained team, and a pipeline your team has already changed on its own.
Results
What changed at Nike, across 900+ stores
Store-server upgrades ran by hand across five teams, only two servers at a time. We replaced that with one self-service upgrade pipeline, let developers start their own staging deploys, and added Jamf integration with macOS signing and notarization for mobile delivery.
| Nike | Before | After |
|---|---|---|
| Store-server upgrade, per server | 6 to 8 hours | about 3 hours |
| Store servers upgraded safely at the same time | 2 | 8 |
| Test build to staging devices | 5 to 10 minutes, by a release engineer | about 1.5 minutes, started by the developer |
| Mobile app and bot provisioning | 45 minutes of hands-on work | about 3 minutes |
| Recurring device-profile repair, several times a week | by hand | one click |
| Quarterly test-driver update | about 2 hours | under a minute |
| Engineering time on this work | 600+ hours a year back | |
“Previously, moving CI workflows between Dev, Staging, and Prod required manual, uncommitted code changes. Mateusz implemented a solution that reduced this to a single line of code, streamlining the process and enabling faster iteration for both developers and our team.”
Reading
More on releasing in retail
- ArticleThe Retail Delivery Squeeze: Shipping Faster Without Breaking PeakWhy the weeks you most need to ship are the weeks a bad change costs most, and how platform work lets speed and stability rise together.Read the article
- ArticleThe Technology Trends Reshaping Fashion - And the Engineering That Ships ThemAI styling, virtual try-on and connected products are now always-on services, and shipping them safely takes small batches, automated tests and fast rollback.Read the article
- ArticleHow Cloud Computing Enhances Online Fashion RetailElasticity for sale peaks has to be engineered and load-tested before the drop, and cloud cost only stays sane with governance.Read the article
Where to start
Start with your last release
The free Release Check shows where your last release spent its time and what that costs a year, in under 5 minutes and without a call.
Contact
Talk to the engineer
Book 30 minutes and bring your last release, or send a message. You talk with Mateusz directly.
