Treat mobile as “hire someone to make a few screens” and the project burns on splash art. Run it as engineering. Each stage has a sign-off: who uses it, actions that must finish, how it connects to the back office, whether a weak field network works, and who answers users after listing. Pick the entry before native versus mini program. See Mini program or APP as the main entry. Cost and cycle are in What APP, mini program and H5 cost, and how long they take.
Split employees, customers, stores and suppliers. How many times a day, whether there is a network, and whether you need push decide APP, mini program, or H5. When H5 is enough is in Is H5 enough on its own; offline is in No network on site—do you need offline.
List the three to five actions that must finish on the phone. Cut decorative pages. Employee, member, store and supplier homes are not the same. See Employee features people will actually use, What belongs on a member mini program home, How to design store redemption and reordering.
Share accounts, stock and membership with the web back office. Isolate by role so stores and warehouses do not leak. The rule is in Why one back office keeps multi-end data from forking.
Walk weak networks, scan, pay and redeem on a real device—not only a simulator. Measure the download and login funnel too. See What to do when nobody downloads the APP.
Listing, review, versions, and who answers users go into the handoff. Certificates and store accounts must sit under the company. Rebuild versus extend is in Rebuild the old APP, or extend it.
The usual stall is not technology. It is that nobody inside can sign “this step can drop paper.” The floor wants a short path, marketing wants a brand home, IT wants it identical to WeCom. At kickoff name who confirms roles, who cuts tasks, who signs visuals—two rounds max. More than two usually means the main entry is unset. After launch, use the checklist to see if anyone uses it. See Launched, unused: how to diagnose. The field gap with and without an entry is in Field work with a mobile entry versus without.
Phase-one launch is not the end. Ticket states change, stores are added, certificates expire. Put maintenance in the contract or an internal role. That is cheaper than another rebuild talk. If the supplier side comes in phase two, reserve the APIs in phase one. See How supplier orders and settlement should run.
Do not treat “design approved” as the big milestone. Real milestones: a field device submits a document, a clerk redeems a coupon, a supplier confirms a PO. Polish visuals after those actions hold. If an old APP must coexist, put the migration list before step 5 so launch day does not stop the floor. A template mini program skips the thinking in step 2 and doubles rework in every stage after. See A template mini program is almost as bad as none.
Phase one does not ship a mall, live streaming, or a forced company-wide APP download. Shrink scope so the floor can finish. Changes go in writing so verbal grid-adds do not roll the IA back into a template. On launch day check that test accounts cannot mix into real redemptions. Walk a real device, real network and real codes. That is closer to acceptance than signing in a demo.
A week after release, ask crews and store managers “what do you tap after you open it?” If they cannot answer, you are still on a brochure. Compare Brochure mini program versus a work tool Then put the home screen back.
Roles and places, task list, APIs and permissions, device acceptance, release and operations. Send who will use it and what must finish on the phone. We schedule as engineering, not as icon work.