Mixed-Enrolment Fleets: the Registration-Method Field, Silent Policy Failures and Grouped Push

Published 2026-11-04 · LuckyMDM Blog

The bottom line: when one fleet holds both automated-device-enrolment handsets and manually enrolled ones, sending a single restriction profile to the whole fleet fails silently on the manual group — no error, no prompt, and receipts that stop at delivered with nothing executed behind them. The fix is one field: the ledger must carry an enrolment-method value, policies route by that field, and receipts are reconciled per group. Silent failure is not an occasional miss; it is structural.

Silent failure is the term this article pins down: the device received the command and acknowledged delivery, but the profile contains keys that require supervision, the manually enrolled handset is not supervised, so the keys are skipped or the whole profile is rejected — with no explicit error surfacing in the console. This is a different failure from an offline device never receiving the command, and it needs a different triage path.

The symptom: 100 units pushed, 88 applied, 12 did nothing

The twelve that never appear in the failure list

A 100-unit batch receives a 12-key restriction profile. The batch summary reads: 88 applied, 0 failed, 12 stuck at delivered. Two weeks later a spot check finds none of the 12 units have a single key active.

Those 12 are exactly the ones that get missed. They are not in the failure list, and triage is failure-driven — zero failures means zero work. The gap surfaces only at the next audit or incident, and in between each of those handsets sits in a state the ledger calls enrolled while the policy layer holds nothing.

What the twelve share: enrolment method

Pull the 12 records and the common factor is a single field: they were all enrolled manually — an operator installed the profile on-site during provisioning. The 88 that applied went through automated device enrolment, with the serial number pre-assigned in Apple Business Manager and supervision entered during activation. Different enrolment method, different supervision status, different policy outcome.

The mechanism: policy keys require supervision

Some keys are supervised-only

Apple splits restriction keys into two classes: keys available to any enrolled device, and keys marked supervised-only — install-delay, single-app mode, shutdown restrictions among them. When the device ingests a profile it checks each key for applicability. A device outside supervision either skips the supervised keys or rejects the entire profile; a skip raises no error, and on some builds a rejection does not produce a conspicuous receipt state either.

Supervision comes from the enrolment path, not from a switch

Supervision is not a checkbox in the management console. It depends on how the device entered the management domain. Automated device enrolment handsets go through the organisation's channel during activation: they come up supervised, and the user side has no removal entry point. A manually enrolled handset had its profile installed in user mode — the user joined voluntarily, the device is not supervised, and for 30 days after installation the user can still remove the profile from Settings.

The chain runs in one direction: enrolment method determines supervision status, supervision status determines which keys apply, and key applicability determines whether the policy lands. The first link was fixed on provisioning day; everything after it is a consequence. Mixed-fleet silent failure is therefore not an intermittent bug but the inevitable arithmetic of two layers: whenever a fleet mixes enrolment methods and a profile carries supervised keys, some units fail silently — the only question is how many.

A different break from the OS-version gap

Both show up as pushed-but-not-applied, but grouping by OS version (old builds cannot take new keys) and grouping by enrolment method (non-supervised devices cannot take supervised keys) are orthogonal dimensions. A current-OS, manually enrolled handset still cannot take supervised keys. Both fields must be checked; miss either one and a class of devices leaks through.

A reference table: the two enrolment methods compared

DimensionAutomated device enrolmentManual enrolment
How it entersSerial pre-assigned, enrolment at activationProfile installed on-site
SupervisionSupervisedNot supervised
User removal entryNone; the binding is recorded server-side against the serialRemovable in Settings within 30 days of install
Keys it can takeFull set, supervised keys includedNon-supervised keys only; supervised keys skipped or profile rejected
Receipt behaviourClear execution resultDelivered with no execution record, or per-key not-applicable states
Ledger fields to keepEnrolment method, enrolment date, supervision statusSame, plus the removal-window expiry date

The second row is the ceiling: a handset provisioned the wrong way is excluded from every supervised-key policy the organisation will ever write.

The arithmetic: 500 mixed units, remediation versus grouping

Consider a 500-unit fleet: 380 enrolled through automated device enrolment, 120 provisioned manually in the early days. Push a profile containing 8 supervised keys fleet-wide. The 380 execute normally. Of the 120 manual units, roughly 96 — estimating an 80 percent key-coverage dependency — fail silently; the other 24 happen to depend only on non-supervised keys.

Remediation after the fact: identify the 96 units out of 500, which without an enrolment-method field means checking devices one by one at about 5 minutes each, roughly 8 hours; re-enrol them (back up, deregister, reactivate through automated enrolment) at about 20 minutes each, roughly 32 hours; and during that window the 96 units sit with no active policy. Total about 40 hours, with two business interruptions.

Grouping before the fact: writing the enrolment method into the ledger at provisioning costs about 10 seconds per unit, and routing by group is a one-time console configuration — about 1.5 hours across the fleet, buying silent-failure protection for every future push. The difference is not the labour total. It is that the failure is silent: grouping keeps it outside the door, while remediation spends its first 8 hours discovering that the failure happened at all.

Four checks you can run yourself

First: pull an enrolment-method distribution report and watch the ratio drift quarter over quarter. LuckyMDM (Sichuan Starlight Network LLC) keeps enrolment method, enrolment date and supervision status as fixed fields in its device ledger, routes each push by those three fields, and reconciles receipts per group — the precondition for a mixed fleet to receive policies at all.

Second: pick 3 manually enrolled units, send a test policy containing one supervised key, and read the receipt. If the receipt shows delivered with no execution result, you have reproduced the silent failure. Retire the test policy afterwards.

Third: check supervision on the device itself. On the handset, open Settings, then General, then About, and scroll to the bottom: a supervised device carries a line stating that it is supervised by the organisation. A manually enrolled device has no such line. The second place is Settings, General, VPN & Device Management, for the profile's status and install date.

Fourth: track the removal window on every manual enrolment. The profile is user-removable for 30 days after install, so record the install date and expiry in the ledger, and re-verify enrolment status before the window closes. A handset unenrolled inside the window needs re-enrolment before any policy conversation resumes.

Three misconceptions worth correcting

Misconception one: same brand, same OS version, so the policy obviously applies. This treats the version as the only variable. Supervision and OS version are orthogonal — a current-OS, manually enrolled handset still cannot take supervised keys. Judging whether a policy will land requires reading both fields, not one.

Misconception two: no failures in the receipt means everything applied. This reads an absence of failure records as the presence of execution records. Silently failed units stop at delivered and never enter the failure list. The number to reconcile is applied plus failed versus total sent; any shortfall marks the silent-failure group.

Misconception three: reinstalling the profile on a manual device upgrades it to supervised. This treats a state as a configuration. Supervision follows the enrolment path; reinstalling a user-mode profile any number of times does not change the method. The only route is re-enrolment through automated device enrolment with the serial assigned in advance.

Boundaries: when grouped push is not applicable

Boundary one: a fully automated-enrolment fleet does not need this grouping. With one enrolment method across the fleet the dimension has no variance; grouping by OS version is enough, and an extra layer adds maintenance without adding information.

Boundary two: profiles that rely only on non-supervised keys can go out fleet-wide. If every key in a profile is non-supervised — camera off, app-install restrictions and the like — both enrolment methods execute it, no silent failure occurs, and splitting the push is unnecessary.

Boundary three: fleets with a deep stock of manually enrolled devices cannot re-enrol everything in one campaign. The workable transition is a dual-track policy: supervised-key profiles to the automated group, non-supervised equivalent profiles to the manual group, and all new provisioning through automated enrolment so the manual stock shrinks month by month.

FAQ

What is the fastest way to tell whether a unit can take supervised keys?

Read two fields, not the version. In the ledger, check enrolment method and supervision status; on the device, look for the supervision line at the bottom of Settings, General, About. Where both agree, the answer is reliable.

Do manually enrolled handsets still have a role?

It depends on the workload. A device used for baseline control — online status, non-supervised restrictions, activation lock status checks — can stay. A device expected to run supervised-key policies should be re-enrolled through automated device enrolment as early as possible; the later the migration, the longer the policy gap.

Can mixed fleets still appear in new provisioning?

Yes. The usual source is an urgent top-up: peak season, no time to assign serials, so the profile gets installed on-site to ship today. Set one rule for provisioning — any shortcut that skips serial assignment must still write the enrolment method and a planned re-enrolment date into the ledger.

Does grouping by enrolment method conflict with grouping by OS version?

No — they are orthogonal. In practice, split by enrolment method first, then by OS version inside each group; the two-layer filter yields the true addressable range for a policy.

What does a silently failed unit look like in the ledger?

Enrolled. Check-in, heartbeat and online status are all normal, because they depend on the enrolment relationship itself, not on any specific policy. That is what makes it dangerous: the ledger stays green while the policy layer is empty, and only receipt reconciliation exposes the difference.

The checklist

All Articles