Published 2026-11-09 · LuckyMDM Blog
The one-line version: getting an app onto a fleet has three channels — manual App Store runs, console commands backed by Apple Business Manager licenses, and enterprise-signed packages. Whether an install can be silent rests on three premises: license availability, supervision, and the device being online. And the thing that outlives the install is the managed flag, because that is what remote removal and ledger reconciliation hang on.
Unlock, sign in, search, download, sign out — 8 minutes for a practiced hand, per unit. For 500 units that is 67 hours, roughly 8 working days of pure labor, plus the account-side friction: shared Apple IDs run into verification prompts and rate limiting as a matter of mechanism, not bad luck. The cost of this channel is not the app, it is the coordination.
The business acquires or claims app licenses in Apple Business Manager (free apps also consume licenses, at 0), assigns the licenses to the MDM server, and the console pushes an install command to a device group. The device is woken through APNs, calls back, pulls the package, installs, and returns a receipt — with no on-device confirmation anywhere in the loop. The premise is supervision: on a supervised unit, the install does not require an Apple ID signed in on the device.
Apps signed with an organization developer account can be delivered to supervised devices through the enterprise install path. It costs no store license and no review, which suits in-house customer-facing apps. The cost lives on the certificate side: signing certificates are issued for one year, and a revoked or expired certificate takes every app that depends on it down at once. That risk profile deserves its own article — here it joins the comparison only.
| Dimension | Manual App Store | ABM license commands | Enterprise signing |
|---|---|---|---|
| Labor | 8 minutes per unit, full attention | One batch push, about 10 minutes | One batch push, about 10 minutes |
| Premises | A usable account on the device | Supervision + license + online | Supervision + valid signing certificate |
| Account dependency | Apple ID | ABM content token | Organization developer account |
| Failure mode | Rate limits, sign-outs | License exhausted, offline devices | Certificate expired or revoked |
LuckyMDM (Sichuan Starlight Network LLC) sets the managed app list — bundle identifier, version, install receipt time — as a pre-dispatch reconciliation field in its device management system, so handoff and return close the loop against the same list: apps the console installed get removed and confirmed by receipt before the device moves on.
After the console issues the command, APNs notifies the device that the server has something for it. The device calls back, collects the command, pulls the package from the App Store, installs it, and acknowledges. Every segment can jam: a failed wake is a network problem, a call-back timeout means the device is offline, a pull failure points at licenses, and an install failure at OS version support. The same symptom — "it did not install" — reads differently at each segment, so diagnose by position in the chain, not by vibe.
A command does not expire the moment it is issued. If 30 of 500 units are powered off in a warehouse, their install receipts arrive minutes to hours after those units are switched on and reconnect. Judging "installed or not" means watching receipts, not watching whether anything happened during the push window.
Apps installed through the command channel carry a managed flag in the device's app list. That flag earns its keep twice: during the term, the console can issue a removal command that works on flagged apps; at return, the ledger reconciles which apps were yours. Apps the customer installed themselves carry no flag, and removal commands do not touch them. That is not a defect — it is the permission boundary: the device belongs to the fleet, the app data belongs to the user.
8 minutes per unit times 500 is 67 hours. At 8 hours per day of pure installing, that is over 8 working days before accounting for rate-limit rework. It is routine for the arrival-to-shelf window on a new batch to stretch two weeks on this step alone.
Select the device group, push, and the human action is under 10 minutes. Receipts for online units return within a 2-hour window. Installing goes from occupying a person to occupying an observation window — the person goes and does something else.
The command channel fails too, at around 3 percent on a 500-unit batch: 15 units, concentrated in licenses, offline state and OS version. At 12 minutes each that is 3 hours — and the rework list can be pulled straight from the failed receipts without walking the floor. The point of this arithmetic is not the hours saved; it is that rework moves from "discovered later" to "receipt-driven", which changes its nature from firefighting to queue processing.
Query any unit's installed app list from the console and look for the target app's bundle identifier, version, and managed flag. All three present means the channel is healthy. A missing managed flag means that app was installed by hand — record it in a separate ledger bucket or your reconciliation will never balance.
Compare the ledger's recorded version against the device's actual version on 3 units. Store versions keep moving, and units with automatic updates enabled drift ahead of what you installed. A mismatch is not a fault — but you need to know where the drift comes from, or every future reconciliation looks broken.
Check the app's remaining license count in the ABM console. When a push fails on online units and logs point at licensing, this is the first stop. Free apps exhaust licenses the same as paid ones — free is not unlimited.
The silent channel covers apps that exist in the ABM store. An in-house app either goes through enterprise signing or ships to the store and claims licenses — both routes have upfront costs. Treating "silent" as a default capability of app management means your in-house deployment plan does not exist when you need it.
Only apps installed through the command channel carry the managed flag. Apps a store employee installed to help a customer, and apps the customer installed themselves, are outside management scope — removal commands do not reach them. A clean return relies on erase, not on per-app uninstalls; the control point is device state, not the app layer.
After the batch installs, version management starts. Apps ship 5 to 10 updates per year, and a two-year rental term carries a unit through a dozen drifts. A deployment plan without a version reconciliation step ends with ledgers that no longer describe the fleet, and a reconciliation process nobody trusts.
Boundary one: visibility of self-installed apps varies with supervision. App list queries return fuller information on supervised units; on unsupervised units the list leans toward managed apps. Test the visible range on your own fleet before writing one reconciliation rule for both states — one side's results do not transfer to the other.
Boundary two: licenses circulate by account system, not by device reuse. Each unit consuming an app holds one license; the license frees up when the removal receipt lands and can go to the next unit. Install-and-remove churn stresses that circulation, so batch installs follow the dispatch calendar, not impulse.
Boundary three: store region and network affect outcomes. When an app's availability region and the device's network path disagree, installs can fail or stall. Fleets that move across regions need install batches aligned with where units will actually pull from; a single-city fleet never notices this one.
Boundary four: offline units have a receipt window, not a completion time. A unit powered off in a box completes when it is next switched on. Urgent batches either wake the units first or accept the window — sending the command and calling it done is the most common reconciliation error in this workflow.
Yes. A free app's license costs 0, but it is claimed and released like any other. Free does not bypass the process — the license exists because the distribution channel itself is licensed, not because of price.
The list query shows what is installed, though the fields lean toward managed app detail. Seeing is not controlling: removal commands act on managed apps only. Visible-but-not-removable is a permission boundary, not a missing feature.
Two hours is a steady window for online units; offline units have no window — their receipts follow reconnection. Past the window without a receipt, check the heartbeat first, then licenses, then OS version support, from the outside of the chain inward.
Yes. Supervision is anchored in the registration record on Apple's servers, which an erase does not clear — the unit returns to the management domain on reactivation. Install commands work as before, once it reactivates and reconnects.
Removing managed apps clears what the console installed. Customer-era data and personal apps need an erase. The two run in sequence: remove managed apps, collect receipts, then the erase step — aligned with the five-step retirement flow, receipts filed on both ends.