What an OTA update actually is
An OTA (over-the-air) update is new firmware delivered to a device already in the field across its network connection, with no physical intervention by the user or installer. For a security product — a door sensor, a PIR detector, a gateway — OTA is the difference between a fix that ships in days and a recall that ships in trucks. A contact sensor with a sleep-current bug, a gateway with a TLS-stack weakness: both are solved in firmware, and both are worthless as fixes if you cannot reach the installed base.
The update channel is part of the product whether you design it or not. A device with no signed-update path is, in practice, a device you can never patch — and for a security brand that is a liability you inherit on day one.
Why OTA is sharper for security hardware
Security devices sit at the edge of the home and hold the most trust. A vulnerability in a door sensor is not just a convenience bug; it is a confidence bug. That raises the bar on two things a brand should ask its manufacturer about before signing: how updates are authenticated, and how a bad update is recovered.
Signed, authenticated updates
A fielded device must reject any firmware that is not genuinely from the manufacturer. Updates are cryptographically signed so the device only accepts images from a key it already trusts. This is the same root-of-trust logic as Matter device attestation, applied to the update channel rather than to first commissioning — the device proves the new image is ours before it ever runs it.
Rollback and staged rollout
A good update process never pushes to 100% of devices at once. It stages — a small batch, watch, then widen — and it keeps the previous good image so a failed update can fall back rather than brick the unit. A factory that cannot describe its rollback path has not really thought about the installed base.
What "Matter-certified" means once you update
A Matter certificate is bound to a specific certified software and hardware configuration; changing that configuration can require re-certification or at least a re-test before the device may keep carrying the Matter mark. This is the part brands underestimate. The connectivity module carries a certified software stack, and your application firmware runs alongside it. A change that alters the certified behaviour — cluster behaviour, commissioning flow, the attestation path — can break the certificate. A responsible manufacturer treats firmware releases as certification events: a documented delta review against the original PICS, and a test pass before the update is broadly pushed. We cover the cost and timeline of certification in our article on Matter certification budgeting.
The factory's OTA duties (not just the brand's)
Shenan Yangguang builds the device end to end — schematic, prototyping, firmware R&D, certification cooperation and mass production — so the OTA duty stays with one party rather than being bounced between a module vendor, a cloud vendor and an assembler. Concretely, the factory owns: building the signed image, holding the signing keys in a controlled environment, running the staged rollout, monitoring field failures, and keeping the update server reachable for the committed lifecycle. When those duties are split across three suppliers, the brand is the one who answers the support ticket when none of them will.
Lifecycle: the commitment that actually matters
Device lifecycle support is the period after shipment during which a manufacturer commits to ship security fixes and functional maintenance for a product. A security product sold in 2026 should not be orphaned in 2027. The commitment belongs in the contract, with a defined minimum support window and a defined notice period before end-of-life. We are deliberately cautious about quoting a single number here: the right window depends on the market, the device class and the channel, and we confirm it per project in writing rather than printing a figure that becomes a promise we cannot keep.
Two update planes: device firmware and the app
A common confusion: the device firmware and the smartphone experience are updated by different planes. The device firmware is the factory's job, signed and certified as above. The companion experience — dashboards, notifications, account settings — lives in a white-label app + cloud stack, updated on its own cadence by the app and cloud team. A brand should map both planes, because a security fix that lands in firmware but breaks in the app is still a broken experience for the user.
What to put in an OTA clause with your ODM
Five lines worth writing down: signed updates with key custody, staged rollout with rollback, a certificate-delta review per release, a minimum lifecycle window, and a field-failure monitoring path. None of these are exotic; all of them are missing from the quotes that win on price alone.
Reading an OTA capability sheet
Before signing, ask the manufacturer for a one-page OTA capability sheet and read it like a spec. It should state the signing scheme (what key type, where the private key is stored), the maximum image size the device can hold for a staged fallback, the rollout batch sizes and the monitoring signal the factory watches during a push. A sheet that says only "we support OTA" is not a specification — it is a promise with no mechanism behind it. The capability sheet is also where the certificate-delta review appears: it should name who re-tests the device against the original PICS after a firmware change, and under what conditions a change is treated as a new certification rather than a maintenance release.
The support window against the installed base
The lifecycle window only means something against the size of your installed base. A brand with ten thousand gateways in homes needs a monitoring and rollback discipline that a pilot of two hundred units does not. We are deliberately cautious about quoting a single support number: the right commitment depends on the device class, the market's expectations and the channel you sell through, and we confirm it in writing per project rather than printing a figure that becomes a promise we cannot keep. The point is not the number — it is that the number exists, is written down, and is owned by one engineering team rather than assumed by three suppliers who will each point at the other when a field fails.