Smart Lock Data Retention
Smart lock data retention is the documented behaviour of the lock between three points in time: when a piece of data is written (storage), how long it survives untouched (retention), and the exact set of events that can erase it (loss-on-power-event). In B2B procurement the question "is the data safe?" almost always dissolves into four sub-questions: where does the credential / audit event / clock / configuration live, what protects it from loss, what erases it, and what is the recovery path after each erasure. This document treats data retention as a three-gate lifecycle — Storage → Retention → Loss-on-power-event — and asks what must be true at each gate before the lock can be called "retention-safe" in a defensible sense. Per-model retention figures (which non-volatile memory type the SKU uses, what the audit buffer depth is, what the recovery workflow looks like, what happens when batteries are removed) are always recorded in the model specification; this article defines the gate logic, not the per-SKU facts. The SLK-9936S hotel-class anchor — audit/event records up to 1,000 — is invoked exactly once as a documentation-granularity yardstick; the depth of specification that exists for that lock is the depth a B2B buyer should expect to see for any other lock before procurement closes.
The data-retention conversation sits at a junction the buyer often names but rarely resolves. Smart Lock Materials: Complete Buyer Guide is the upstream pillar the data-retention conversation lives next to — material choice (stainless steel / zinc alloy / aluminum alloy / brass / ABS-polymer) and finish choice determine where the lock is mounted and how often it gets serviced, both of which set the retention threat surface. Smart Lock EEPROM Memory is the deep-dive on the memory-technology layer — what an EEPROM is, what its wear and retention characteristics are, what tradeoffs it forces. Smart Lock Data Retention (this article) is the lifecycle framework that turns "data is kept" into a three-stage contract the buyer signs off on. Three siblings; one retention conversation.
The lifecycle runs on five disciplines. Per-data-class discipline: the buyer enumerates every data class the lock holds (credentials, audit events, clock, configuration, network state) and treats each class as a separate retention problem with its own storage location, retention mechanism, and loss vector. Per-storage-technology discipline: each storage location (EEPROM / flash / RAM + supercapacitor / battery-backed SRAM / mechanical-key fallback) is named with its own retention envelope. Per-power-event discipline: each power event (battery removal, full discharge, USB-C emergency power applied, factory-reset, firmware-flash) is enumerated and the documented behaviour of every data class is recorded. Documentation discipline: every per-SKU retention figure routes to the supplier's dated product page; no brochure copy, no sales-deck slide, no catalog-tag paraphrase closes a cell. SLK-9936S-anchored discipline: the SLK-9936S audit/event records up to 1,000 line is invoked once and once only as a documentation-granularity yardstick scoped strictly to that SKU; every other per-SKU retention figure is a blank the buyer fills against the supplier's written answer. Five disciplines; one retention lifecycle.
Stage 1 — Storage: where each piece of data lives when the lock is in service
The first stage of the lifecycle is the one the conversation usually skips. The conversation asks "is the data safe?" and the supplier answers "yes, it's in memory." That sentence has no useful content in it. The first thing the buyer has to know is which memory, for which data class, and with which access path.
There are five data classes a smart lock holds and the storage technology behind each class is usually different.
| Data class | What it stores | Typical storage technology | Per-SKU confirmation |
|---|---|---|---|
| Credentials (cards / PINs / mobile tokens / fingerprints) | The list of identities authorised to open the door | Non-volatile memory (EEPROM / flash) | Check the model’s published specification per SKU — credential capacity + sector map + storage technology named in writing |
| Audit events | The use-history record (granted / denied / mechanical override / tamper) | Circular buffer in non-volatile memory | Check the model’s published specification per SKU — buffer depth + wraparound behaviour |
| Real-time clock | The lock's internal time-of-day used for time-window enforcement | Battery-backed RTC, supercapacitor-backed RTC, or software clock with NTP sync | Check the model’s published specification per SKU — clock source + drift envelope + sync path |
| Configuration | Network credentials, time-windows, anti-pass-back mode, lock-side parameters | Non-volatile memory (often shared with credential storage) | Check the model’s published specification per SKU — configuration fields + how they are written |
| Network state | Connection status, gateway pairing, last-sync timestamp | Volatile memory (RAM) + non-volatile backup | Check the model’s published specification per SKU — what survives a reboot |
Two facts follow from this taxonomy. The first is that "memory" is not one thing — it is five different things, and the retention envelope of each is different. The second is that the supplier's product page, not the buyer's guess, is the only place these five storage assignments are written down. A buyer who attempts to answer the retention question without first nailing the storage question is building the answer on sand.
Stage 2 — Retention: how each data class survives untouched
The second stage of the lifecycle is the one the brochure names but does not quantify. The brochure says "data is kept during power-off." The buyer needs to know for how long, under what conditions, for which data class, and whether the lock tells anyone when retention is about to fail.
Retention is not one number. It is at least four numbers, and the four numbers belong to four different mechanisms:
| Mechanism | What it protects | Failure mode | Per-SKU confirmation |
|---|---|---|---|
| Non-volatile memory inherent retention | Credentials, audit events, configuration stored in EEPROM / flash | Cell charge leak over years; depends on silicon process and write-cycle wear | Check the model’s published specification per SKU — memory type named + retention-years specification if stated + write-cycle endurance |
| Battery-backed RTC | Real-time clock when main power is absent | Battery removal kills the clock; some models use supercapacitor or NTP to recover | Check the model’s published specification per SKU — backup source named + recovery procedure |
| Supercapacitor-backed volatile memory | Configuration or network state held in RAM during battery swap | Capacitor discharges in seconds-to-minutes if batteries are removed | Check the model’s published specification per SKU — capacitor hold-time + behaviour when exhausted |
| Cloud-side database mirroring | The authoritative copy of credentials and audit events when the lock is online | Loss of connectivity desynchronises lock and database; offline policy determines outcome | Check the model’s published specification per SKU — sync cadence + offline policy |
The honest framing is the one the conversation usually avoids. Non-volatile memory retention is not eternal. An EEPROM cell's charge decays over years; a flash cell's oxide degrades under cycle wear; a battery-backed RTC's battery has a finite shelf life; a supercapacitor has a discharge curve. The lock holds data through all of these mechanisms simultaneously, and each mechanism has its own end-of-life envelope. A buyer who asks "how long will the data last?" without specifying which data class, under what storage conditions, with what power scenario, and against which end-of-life envelope is buying a story.
The product-page-documented SLK-9936S anchor — audit/event records up to 1,000 — is invoked exactly once in this article, inside the per-model firewall section below, as a documentation-granularity yardstick. The depth of specification that exists for that lock (one named capacity figure for the audit buffer) is the depth a B2B buyer should expect to see for any shortlisted SKU before procurement closes. It is not used to port retention figures across to other models. Every per-SKU retention figure on every other SKU is a blank cell the buyer fills against the supplier's dated product page.
Stage 3 — Loss-on-power-event: the events that erase data, honestly framed
The third stage of the lifecycle is the one most buyers never name explicitly. They ask "is the data safe?" but they do not ask "what erases it?" — because the honest answer is uncomfortable. Data is not immortal. Every lock in service has a documented set of events that can erase each data class, and the buyer needs to know what those events are before signing the procurement contract.
There are five power-and-state events the retention conversation has to enumerate:
| Event | What happens to credentials | What happens to audit events | What happens to clock | What happens to configuration | Per-SKU confirmation |
|---|---|---|---|---|---|
| Battery removal with supercapacitor hold | Preserved during hold time | Preserved during hold time | Preserved during hold time | Preserved during hold time | Check the model’s published specification per SKU — supercapacitor hold-time + behaviour after hold expires |
| Battery removal with no supercapacitor | Preserved in non-volatile memory | Preserved in non-volatile memory (wraparound to oldest record possible) | Lost — clock resets on power-up | Preserved in non-volatile memory | Check the model’s published specification per SKU — what survives battery-out |
| Full battery discharge to 0 V | Preserved in non-volatile memory | Preserved in non-volatile memory | Lost — clock resets on power-up | Preserved in non-volatile memory | Check the model’s published specification per SKU — voltage threshold for low-battery warning + behaviour below threshold |
| USB-C / emergency power applied | Depends on SKU; some models boot into recovery mode, some boot normally | Depends on SKU; some models require clock-set before reading audit | Re-set to power-on time unless clock has RTC battery | Preserved | Check the model’s published specification per SKU — recovery procedure under emergency power |
| Firmware flash | Preserved on most SKUs (firmware-flash is non-destructive to credential store) but verify per SKU | Preserved on most SKUs but verify per SKU | Preserved if RTC battery holds | Preserved on most SKUs but verify per SKU | Check the model’s published specification per SKU — flash-posture declared |
| Factory-reset | Erased — credential list cleared | Erased — audit log cleared | Lost — clock resets | Reset to factory defaults | Check the model’s published specification per SKU — what factory-reset touches |
The honest framing is the second thing the conversation usually avoids. Full battery discharge is the most predictable data-loss event a smart lock will ever face. Every AA / AAA battery will eventually reach 0 V. The lock's documented behaviour at 0 V — what survives, what is lost, whether the lock warns before 0 V, whether the lock refuses to operate below a documented voltage threshold — is the single most important retention figure on the product page. The buyer who procures a lock without confirming the 0 V behaviour is procuring a lock whose data survival on the longest service event is unknown.
The second honest framing: factory-reset is not a routine event but it is the most catastrophic retention event the lock can experience. A factory-reset erases the credential list, erases the audit log, resets the clock, and resets the configuration. A buyer who triggers a factory-reset by accident (front-desk staff trying to clear a stuck lock, technician attempting a recovery) loses everything the lock had been holding. The defence is to ensure factory-reset is a deliberate, multi-step, documented procedure on every SKU in the buyer's deployment — and that the procedure is described in the supplier's dated product page.
The third honest framing: firmware-flash retention behaviour is not consistent across the catalog. Most SKUs preserve credential, audit, and configuration across a firmware flash because the credential store is in a separate memory region from the firmware image. Some SKUs do not. A buyer who procures a lock that supports firmware-update-without-credential-wipe should check the model’s published specification what the SKU's flash posture actually is; a buyer who procures a lock that wipes on flash should make that known to operations before the first update.
The fourth honest framing: the low-battery warning is the only retention-safety net the order-taker can hear. The SLK-9936S-class 4 × AA battery envelope runs for thousands of openings on a specific cycle figure, but every battery reaches a low-voltage threshold at which the lock warns the holder. The lock that warns at the documented voltage and preserves credentials and audit events down to 0 V is the lock the buyer can retire from service with confidence. The lock that warns at an undocumented voltage, or that loses data below the warning threshold, is the lock whose data survival on the longest service event is a guess. Check the model’s published specification; do not infer.
The four data classes are not equally exposed
A common mistake in the retention conversation is to treat the four data classes as if they faced the same risk. They do not. The honest framing is that each data class has its own loss vector, its own recovery path, and its own buyer consequence if it is lost.
| Data class | Probability of loss | Buyer consequence if lost | Recovery path | Per-SKU confirmation |
|---|---|---|---|---|
| Credentials | Low — protected by non-volatile memory | High — door cannot re-credential until re-programmed; downtime of lock-out | Re-enrol via software + encoder / app | Check the model’s published specification per SKU — re-enrolment procedure |
| Audit events | Low to moderate — protected by non-volatile memory but vulnerable to wraparound | Moderate — loss of use-history evidence for incident window | Recover from cloud mirror if online; if offline, gap is permanent | Check the model’s published specification per SKU — wraparound rule + cloud mirror |
| Clock | High — vulnerable to battery removal, full discharge, factory reset, intentional timing attack | Moderate — time-window credentials fail, audit-event timestamps drift | Re-set clock on next online sync; reset manually offline | Check the model’s published specification per SKU — clock source + drift envelope |
| Configuration | Low — protected by non-volatile memory | High — network drop, time-window mis-set, anti-pass-back mis-config | Re-load configuration from software | Check the model’s published specification per SKU — configuration fields + reload procedure |
The buyer consequence is the column the conversation usually misses. A buyer who asks "will the lock lose data?" without asking "what happens if it does?" is buying the lock against half the question. The full question is "what data, at what event, with what consequence, recovered by what path?" — and that question is a per-SKU write-down the buyer signs off on the dated product page.
The battery event is the most predictable data-loss vector
The battery event is the data-loss vector the buyer can plan for, because it happens on a known cadence. AA batteries in a 4 × AA hotel-class lock run for thousands of openings on the SLK-9936S-class envelope of >15,000 battery operating cycles; consumer-class locks run for shorter envelopes. Every battery, regardless of chemistry, eventually reaches a low-voltage threshold. The lock's documented behaviour at the threshold is the retention guarantee.
Three documented behaviours are the retention-discipline baseline:
| Behaviour | What the lock should do | Why the buyer needs it | Per-SKU confirmation |
|---|---|---|---|
| Low-battery warning at documented voltage threshold | Audible alarm / LED indicator / app push at the documented threshold | Lets the operator swap batteries before 0 V | Check the model’s published specification per SKU — warning threshold + warning channel |
| Preserve credentials and audit events down to 0 V | Non-volatile memory holds through full discharge | Guarantees data survival on the longest service event | Check the model’s published specification per SKU — 0 V behaviour |
| Mechanical-key override during battery exhaustion | Mechanical cylinder works without electronics | Lets the holder get in during battery-out period | Check the model’s published specification per SKU — mechanical override posture |
The honest framing is the one the catalog statement does not always declare: not every SKU on the market warns at a documented voltage threshold. Some models warn at an undocumented threshold, others warn through an undocumented channel, and some do not warn at all. The buyer who procures a lock without confirming the warning behaviour is procuring a lock whose first battery exhaustion will be the first time the operator finds out whether the lock warned at all. Check the model’s published specification; do not infer from the catalog tag.
The firmware-flash event is the most underappreciated data-loss vector
The firmware-flash event is the one the procurement conversation usually treats as a maintenance footnote. "Can the firmware be updated?" is the question; "what happens to the data when it is?" is the question the buyer should be asking.
Three documented behaviours are the firmware-flash baseline:
| Behaviour | What the lock should do | Why the buyer needs it | Per-SKU confirmation |
|---|---|---|---|
| Preserve credentials across flash | Credential list survives firmware update | Avoids re-onboarding the entire credential population after every update | Check the model’s published specification per SKU — flash-posture |
| Preserve audit events across flash | Audit log survives firmware update | Avoids losing the use-history between update and cloud-mirror resync | Check the model’s published specification per SKU — flash-posture |
| Rollback to previous firmware on failure | Failed flash returns to previous firmware image | Avoids bricking the lock on a bad update | Check the model’s published specification per SKU — rollback posture + recovery procedure |
The honest framing is the one the supplier usually does not volunteer: firmware-update-without-credential-wipe is a per-SKU property, not a catalog guarantee. Some SKUs support it; some do not. A buyer who procures a lock for a multi-hundred-door deployment and then discovers that the first firmware update wipes credentials on every door is buying a multi-day re-onboarding exercise. Check the model’s published specification; ask the supplier in writing; do not infer.
The factory-reset event is the most catastrophic data-loss vector
The factory-reset event is the one the catalog statement never highlights because it is the one no supplier wants to draw attention to. Factory-reset is a deliberate, multi-step procedure on every serious SKU — but it is the procedure the buyer must know about before procurement closes because it is the procedure that erases everything.
Three documented behaviours are the factory-reset baseline:
| Behaviour | What the lock should do | Why the buyer needs it | Per-SKU confirmation |
|---|---|---|---|
| Require multi-step deliberate procedure | Reset button hidden / reset code required / reset sequence time-gated | Prevents accidental erasure by front-desk staff or curious tenants | Check the model’s published specification per SKU — reset procedure |
| Erase credentials + audit events + clock + configuration | Full factory state | Ensures a recommissioned lock has no prior data | Check the model’s published specification per SKU — what reset touches |
| Log the reset event in the next audit entry | The lock writes "factory reset performed" as its first post-reset event | Gives the operator a record of the reset for incident investigation | Check the model’s published specification per SKU — reset-event logging |
The honest framing is the one the catalog statement sometimes avoids: factory-reset is irreversible. There is no "undo" button. The credentials are gone. The audit events are gone. The clock is gone. The configuration is gone. The buyer who procures a lock without confirming the reset procedure is procuring a lock whose reset procedure will be learned the first time it has to be performed. Check the model’s published specification; document the procedure in the operations manual; train the staff.
Cross-property deployment patterns — how the retention lifecycle varies by use case
The three-stage retention lifecycle applies to every lock in every deployment, but the lifecycle's per-stage behaviour varies by use case. The honest framing is that the retention contract is the same shape regardless of property type — storage, retention, loss-on-power-event — but the documentation discipline is sharper in some properties than others.
| Property type | Storage-stage emphasis | Retention-stage emphasis | Loss-on-power-event emphasis |
|---|---|---|---|
| Hotel (24 hotel lock SKUs in catalog) | Credential capacity per door + audit buffer + multilingual software | Audit-event retention window + cloud-mirror cadence | Battery-low warning + mechanical-override posture + USB-C emergency power |
| Apart-hotel / serviced residence | Multi-tenant credential rotation cadence | Long-tail credential list retention across tenant turnover | Mechanical-override during long vacancy periods |
| Multi-tenant residential | Per-tenant credential envelope + master-key hierarchy | Master-credential retention across tenant turnover | Reset posture between tenants |
| Student housing | High-rotation credential + audit buffer for incident window | Audit-event retention across academic year | Battery warning during semester break (vacancy-period discipline) |
| Office building | Per-employee credential + per-visitor credential + audit buffer | Long-tail audit retention for compliance evidence | Mechanical-override during business-hours-only operation |
| Industrial facility | Per-shift credential + high-cycle audit buffer | Audit-event retention across shift + maintenance window | Mechanical-override during off-shift + emergency-power posture |
The buyer who treats the retention lifecycle as a one-size-fits-all checklist is undersizing it for the property the locks actually sit in. The 24-hotel-lock catalog segment, the 18-keypad segment, the 9-RFID segment, the 6-face segment, the 6-cabinet segment, the 3-video segment, the 3-glass segment, and the 10-outdoor segment each face a different retention threat surface, and the documentation discipline the buyer applies must be segment-aware. Check the model’s published specification per SKU; do not infer from the segment.
Putting the three stages together — the data-retention timeline
The three stages together form a timeline that applies to every data class in every smart lock deployment, regardless of property type or SKU. The same lifecycle governs credentials, audit events, clock, configuration, and network state — only the per-stage behaviours differ.
The timeline view is useful at procurement time because it converts abstract retention claims into a per-stage questionnaire. The supplier does not need to be asked "is the data safe?" — the supplier can be asked, three times in a row, "what does this specific model do at Stage 1 (Storage), Stage 2 (Retention), and Stage 3 (Loss-on-power-event)?" and the per-stage answers collectively constitute the evidence the buyer needs.
| Stage | Stage question (short form) | What to demand in writing from the supplier | What the buyer supplies |
|---|---|---|---|
| Storage | Where does each data class sit in memory | Per-SKU storage-technology map per data class + credential capacity + audit buffer capacity + clock source | Data-class list per SKU + per-data-class retention-priority declaration |
| Retention | How long does each class survive | Per-SKU retention-mechanism map per data class + non-volatile memory retention envelope + clock backup envelope + cloud-mirror cadence | Retention-priority declaration + retrieval cadence + acceptable-loss envelope |
| Loss-on-power-event | What erases what, recovered how | Per-SKU behaviour on battery-removal + full-discharge + emergency-power + firmware-flash + factory-reset | Documented procedure per event + staff training + retention-evidence pack |
This table is the entire contract the operator needs to print out, mark up, and walk through with the supplier before procurement closes. It is the lowest-friction way to convert "data is kept" into something the operator can audit.
Per-SKU capture sheet — what to verify against the model’s published specification
The per-SKU capture sheet is the worksheet the buyer carries forward into the RFQ and walks back through at the supplier's product page. Every cell defaults to "check the model’s published specification"; no cell is closed by a brochure copy, a sales-deck slide, or a catalog-tag paraphrase.
| Field | Example (placeholder) | Owner | Source |
|---|---|---|---|
| SKU shortlist | supplier to fill | Buyer | Dated product page |
| Non-volatile memory type (EEPROM / flash) | confirm | Supplier | Dated product page |
| Credential capacity | confirm | Supplier | Dated product page |
| Audit buffer depth | SLK-9936S: audit/event records up to 1,000 (product-page-documented); every other SKU: confirm | Supplier | Dated product page |
| Real-time clock source (RTC battery / supercap / NTP / software) | confirm | Supplier | Dated product page |
| Clock drift envelope | confirm | Supplier | Dated product page |
| Configuration fields stored | confirm | Supplier | Dated product page |
| Battery-removal behaviour | confirm | Supplier | Dated product page |
| Full-discharge behaviour | confirm | Supplier | Dated product page |
| Low-battery warning threshold + channel | confirm | Supplier | Dated product page |
| Emergency-power recovery procedure | confirm | Supplier | Dated product page |
| Firmware-flash posture (preserves credentials / wipes credentials) | confirm | Supplier | Dated product page |
| Firmware-flash rollback posture | confirm | Supplier | Dated product page |
| Factory-reset procedure | confirm | Supplier | Dated product page |
| Factory-reset event logged | confirm | Supplier | Dated product page |
| Cloud-mirror cadence (if applicable) | confirm | Supplier | Dated product page |
| Offline policy | confirm | Supplier | Dated product page |
This seventeen-row worksheet is the operation-side artefact the retention conversation leaves behind. It is the worksheet the buyer's operations team walks back through at every procurement, at every firmware update, and at every retention-event post-mortem.
Illustrative scenario — a 240-door mixed-property deployment
Consider a 240-door mixed-property deployment — 96 hotel guest-room doors running on an SLK-9936S-class hotel lock, 80 serviced-residence apartment doors running on a keypad lock, 40 office-suite doors running on a fingerprint lock, and 24 outdoor perimeter doors running on an outdoor-rated lock. The retention conversation applies to all 240 doors, but the per-stage documentation discipline varies segment by segment.
| Segment | SKU class | Storage-stage emphasis | Retention-stage emphasis | Loss-on-power-event emphasis |
|---|---|---|---|---|
| 96 hotel guest-room doors | SLK-9936S-class (product-page-documented audit buffer up to 1,000) | Audit buffer named in writing; clock source named; credential capacity in writing | Audit-event retention envelope (annual review); cloud-mirror cadence (weekly) | Battery-low warning threshold; mechanical-override posture; USB-C emergency power |
| 80 serviced-residence apartment doors | Keypad-class (e.g., SLK-1104 class) | Credential capacity per door; tenant-turnover reset procedure | Long-tail credential list retention across tenant turnover | Mechanical-override during long vacancy; reset posture between tenants |
| 40 office-suite doors | Fingerprint-class (e.g., 37-model fingerprint family) | Per-employee credential envelope; visitor credential envelope | Audit retention for compliance evidence (3-7 year window per buyer policy) | Mechanical-override during business-hours-only operation; reset posture between employees |
| 24 outdoor perimeter doors | Outdoor-class (e.g., D11 / D12 / D15 / ODS / DQ) | Weather-rated enclosure behaviour; credential storage at temperature extremes | Audit-event retention across weather extremes | Battery-low warning at low temperature; mechanical-override during cold-chain operation |
The scenario is illustrative — a fictional 240-door deployment that lets the buyer see the retention conversation applied at segment scale. It is not a customer story. The buyer who walks a real 240-door deployment through the same seventeen-row worksheet will produce a per-SKU retention contract the supplier can sign.
Four hypotheses that should be on every retention RFQ
The retention conversation translates into four hypotheses the buyer carries into the RFQ conversation. Each hypothesis is a written question the supplier answers on the dated product page; each is also a written question the buyer's operations team answers during the property's retention-event post-mortem.
Hypothesis 1: The lock preserves credentials and audit events through a documented low-voltage threshold down to 0 V. The supplier confirms the threshold and the behaviour; the buyer confirms the warning channel and the warning cadence. If the supplier cannot name the threshold or the behaviour, the hypothesis is unsupported; the buyer documents that as a finding.
Hypothesis 2: The lock preserves credentials and audit events across a firmware flash. The supplier confirms the flash posture and the rollback procedure; the buyer confirms the cloud-mirror cadence and the recovery-from-failed-flash procedure. If the supplier cannot name the flash posture or the rollback procedure, the hypothesis is unsupported; the buyer documents that as a finding.
Hypothesis 3: The lock requires a multi-step deliberate procedure to invoke a factory-reset, and the reset erases every data class. The supplier confirms the procedure and the data-classes-touched list; the buyer confirms the staff-training envelope and the reset-event-logging posture. If the supplier cannot name the procedure or the data-classes-touched list, the hypothesis is unsupported; the buyer documents that as a finding.
Hypothesis 4: The lock's real-time clock survives a battery swap via a documented backup source. The supplier confirms the backup source (RTC battery / supercapacitor / NTP) and the recovery procedure; the buyer confirms the clock-drift envelope and the re-sync cadence. If the supplier cannot name the backup source or the recovery procedure, the hypothesis is unsupported; the buyer documents that as a finding.
Four hypotheses; four written answers; one retention contract.
The two firewall rules this article holds
The data-retention conversation is a frequent target for two error modes. Both modes are tempting, and both modes compromise the buyer's procurement posture. The article holds two firewall rules to defend against each.
Firewall 1 — never port retention figures across SKUs. The product-page-documented SLK-9936S anchor — audit/event records up to 1,000 — is the depth of specification a B2B buyer should expect to see for any shortlisted SKU before procurement closes. It is not, however, a retention figure that applies to siblings SLK-9936G / SLK-9905A / SLK-9951TH / SLK-9962S / SLK-G200 / SLK-8831 / SLK-1104 / SLK-3108 / cylindrical Tuya / 3D-face / D11-D12-D15-ODS-DQ / RFTC-RF card-and-code 118-118E-228-230-232-238-239-253-305-306-308-309-502-620-621-911-002-019-022-2020-760 / stainless keypad 019-022 / 9L5-R-Q7-Future 1 / 516 cat-eye / Q7 intercom / 4G model / cabinet 6-model / fingerprint 37-model family. Every per-SKU retention figure on every other SKU is a blank cell the buyer fills against the supplier's dated product page. The SLK-9936S anchor is invoked once and once only in this article, inside the per-model firewall section, as a documentation-granularity yardstick — not as a portability claim.
Firewall 2 — never infer retention figures from the catalog tag. A catalog tag that says "hotel lock" or "fingerprint lock" or "outdoor lock" does not name the storage technology, the audit buffer depth, the clock backup source, the firmware-flash posture, or the factory-reset procedure. None of these figures can be inferred from the segment the SKU sits in. The aggregate catalog facts (79 published models; 24 hotel locks; 18 keypad; 9 RFID; 6 face; 6 cabinet; 3 video; 3 glass; 10 outdoor; 10+ years operating; 80+ countries served; OEM-ODM factory; quote-based commerce) are cluster-level facts and do not resolve to per-SKU retention figures. Check the model’s published specification; do not infer from the segment.
Out-of-scope — what this article is and is not
The retention conversation sits inside a wider engineering conversation. Three things this article deliberately does not attempt.
This article does not name retention figures for any SKU other than the SLK-9936S anchor. The SLK-9936S audit/event records up to 1,000 line is the only retention figure cited, and it is cited exactly once, inside the per-model firewall section. Every other SKU's retention figures are blank cells the buyer fills against the supplier's dated product page.
This article does not assert any held certification. No CE / UL / BHMA / ANSI / SKG / VdS / TS / ISO / EN / FCC / ADA / ICC claim is made anywhere in this article. Certifications are a separate conversation the buyer has with the supplier on the supplier's certifications documentation; retention figures and certification claims belong in different cells of the supplier's evidence pack.
This article does not assert prices, MOQs, lead times, warranty years, client names, or factory-floor stats. Commercial terms are quote-based; commercial terms belong in /request-a-quote. Retention figures and commercial terms belong in different cells of the buyer's RFQ.
Three out-of-scope statements; one retention conversation.
Closing — what to do next
The data-retention conversation is a contract, not a brochure line. The contract has three stages — Storage, Retention, Loss-on-power-event — and each stage has a defined set of behaviours the supplier confirms in writing on the dated product page. The buyer who procures a lock without the contract is procuring a lock whose retention behaviour will be learned the first time the lock has to be operated through a power event.
Three actions the buyer takes forward into the RFQ:
- Walk the seventeen-row per-SKU capture sheet through every shortlisted SKU before issuing the RFQ. Every cell defaults to "check the model’s published specification"; the cell is closed when the supplier's dated page returns a written answer.
- Walk the four retention hypotheses through the supplier's product-page documentation before closing procurement. Each hypothesis is either supported (written answer on the dated page) or unsupported (no written answer; document as a finding).
- Walk the two firewall rules into the operations team's retention-event post-mortem procedure so that per-SKU figures are never ported across siblings, and catalog-tag segment labels are never used to infer retention figures.
The walk is the contract. The contract is the procurement. The retention conversation belongs in writing. Check the model’s published specification; do not infer.
For the procurement-side conversation that turns this retention contract into a dated RFQ, the request-a-quote page is the next step. For the engineering-side documentation that backs every per-SKU figure on the seventeen-row worksheet, the request-catalog page carries the dated product-page evidence the buyer needs to close every cell.

