Restaurants data retention: A working diary of how one restaurant group set retention periods and privacy controls for smart lock records across sites.

Veröffentlicht am: Nov 21, 2023

Data Retention and Privacy for Smart Locks in Restaurant Groups

Data Retention and Privacy for Smart Locks in Restaurant Groups

Every building type resolves into door classes, and restaurants data retention is organised around them. This is a diary, not a policy. Across eleven working days I sat with a restaurant group that was putting electronic locks into kitchens, stores and staff doors across a portfolio that straddled two jurisdictions. My job was narrow and slightly awkward: find out what happens to the records a lock produces once the event is over.

  • Separate the door families — sort the door population into classes that behave alike, starting from restaurants data retention
  • Map each to its grade — state what day one needs before anything is quoted
  • Test the fit — record the measurement that closes restaurants data retention
  • For restaurants data retention, the order below is fixed and restaurants is the reference point.

The question I was handed had a legal flavour but an operational centre. "Restaurants data retention is not on my list of things I own," the property director told me on the first morning. "It sits between the system supplier, the area manager and whoever reads the incident report, and it has been nobody's job for two years." That sentence set the order of my diary. If nobody owns the deletion step, the records grow until somebody else finds them.

What follows is the log of that fortnight: the sites I visited, what I found in the admin exports, the arguments I had with the group's own people, and the working rules that came out of it. I have kept the group anonymous, the sites unnumbered and every example illustrative. The point is not the particular group. The point is that restaurants data retention decisions get made in kitchens and back offices rather than in a policy manual, and they fail in the same places every time.

CONFIRM-WITH-LOCAL-DATA-PROTECTION-REVIEWER: which retention period applies to staff unlock events at your sites, and whether evidence of deletion has to be kept separately from the event itself.

Day one at the kitchen door: four fields and a hundred service

Direct answer. The first site was a forty-cover room with a front door that carries the brand, a kitchen door on a keypad, a dry store on a code, and a staff entrance on a fob. The property manager had the keys in a drawer and an office printer no one had printed from since the rollout. The standing question in this vertical is which door fails at the worst moment and what the manual route is. The answer here was the kitchen door

What interested me was not the hardware but the record. When the manager showed me a week of activity, the report came back thin: a door identifier, a timestamp, the credential presented, and the outcome. No photograph, no audio, no narrative. That thinness is good news, and it is also fragile, because the record on its own says very little and the group had already added the rest. The same question as restaurants data retention, read from a different angle, is covered in Guest, Resident and Staff Experience of Smart Locks in Restaurant Groups. If the answer has to survive a supplier meeting, Total Cost Thinking for Smart Locks in Restaurant Groups is the next one to read. What can be supplied is listed in the smart door locks.

The operations manager had built a second export with first names joined in from the rota. She wanted to see who was opening the dry store late at night. It worked well, and it also created a monitoring system nobody had authorised and nobody could delete. The door record had gone from a mechanical fact into a personnel record without anyone passing a gate. The second take on restaurants data retention is Illustrative 6-300 Sites Smart Lock Programme for Restaurant Groups: Scope Notes.

Volume matters here as much as content. A door opened at every service change means a row at every service change. My own planning arithmetic, not a measured figure, puts one kitchen door at roughly six thousand rows a year; across a group of sites the same arithmetic multiplies, which is why the retention period has to be set at group level rather than site by site. Door types differ across the estate, and the compatibility work in Door Types in Restaurant Groups: Kitchen, Store and Staff Doors Compatibility Notes changes which doors can carry a reader at all, which in turn changes which records exist. The reference set starts with Fingerprint Smart Locks for Commercial Property Owners.

Confirm on the dated product page for the specific SKU what the door record contains and where it is held, because the manager's console is a supplier system and the field list is a supplier statement.

The export that sat on the shared drive

The group had never decided to keep monthly exports. It had simply never deleted them, and the difference between those two things shows up the first time someone asks what data the group holds.

At the head office I was shown a shared drive with a folder per site and a file per month. Each file was a spreadsheet of every unlock event across that site's doors, with credential identifiers instead of names. The naming convention was clean, the folder structure was tidy, and the oldest file went back several years. Nobody had a reason to keep them and nobody had a mechanism to remove them. The retention decision had moved out of the supplier's platform and into the group's own storage, where the group owned it completely and governed it not at all.

Two things made it worse than a tidy archive. First, the roster file that mapped credential identifiers to people lived in a different system, so the spreadsheet was anonymous until someone joined it to a name. That meant the deletion rule for the join file mattered as much as the rule for the event file, and nobody had written either. Second, the exports had been shared out. A site manager had downloaded a copy for a dispute, a duty manager had emailed one to the area manager, and a printed copy sat in a folder behind the till at one site. Deletion inside the platform would have left every one of those untouched.

I wrote a single sentence in the diary that evening and it became the spine of everything after it. A retention schedule that only names the supplier's database is half a schedule. The full scope is the platform, the export, the spreadsheet, the inbox, the shared drive and the paper.

Site two: a contractor asking who opened the store at 23:14

The second site had a different complication, and it is the one that makes restaurants different from an office. The people who present credentials at a restaurant door are not only employees. There are agency kitchen staff, cleaners on a contract, engineers on a maintenance call, suppliers delivering, and a front desk handing out temporary access to people nobody in the building has met.

At this site a contract cleaner had been asked, after a break-in in the corridor, to produce her own activity history. She asked the site manager for a copy of the records showing when the corridor door was opened. The site manager's answer was honest: she had no idea how to produce one, no idea what would appear in it, and no idea how long it would take.

That question sits inside a web of relationships the group had never mapped. Whether a credential holder is a direct employee, an agency worker or a visitor changes which set of records exists and who holds them. The agency may keep its own records of who it sent, the group keeps its own events, and the person asking wants a joined view that neither system was built to produce. Contractor and Third-Party Access for Restaurant Groups covers the access mechanics on the same doors; this diary picks up where that one stops, at the moment somebody asks for the record behind the credential.

The group took two decisions here. The first was that a person outside the payroll could be issued a credential that carries a contractor reference rather than a name, with the join held in the agency's own record. The second was that a question about activity gets answered with a defined pack, produced by a named person, rather than with an improvised export assembled under pressure.

The request that arrived at 04:00 on a Tuesday

Three days after my visit, a subject access request for smart lock records arrived at the group's head office. It was forwarded to the site, then to the area manager, then to me. Total elapsed time before anyone who knew the system was involved: two days. The request itself was reasonable and narrow. The group's problem was not the question; it was that there was no path to answer it.

I sat with the site manager while she tried. She found the console, found the report, found the date range. Then she hit the first real question: the export contained events for every person who used that door, so the answer to one person's request included rows about other people. Redaction by hand, on a spreadsheet, with no rule about which columns matter. The second question was how far back the platform would search, and the answer came back as a support ticket. The third was where the group stored the resulting file once the answer had been sent. Nobody had decided, so it went into a personal mailbox, which meant it would be there until somebody tidied a mailbox.

The fix was not technical. It was a runbook: a single owner, a standard evidence pack, a defined field list, a redaction rule that keeps only what answers the question, a filing location, and a review of the answer before it leaves. The deadline and the grounds for refusal are the reviewer's call, not mine.

Audit Trail Requirements for Restaurant Groups: What Records Matter sets out which records earn their place in a trail. This diary takes the other half of the problem, which is how long each of those records stays and how it is disposed of when its time is up.

The rules the group signed off, written as a table

The output of the second week was a single sheet of paper that the property director, the operations lead and the group's reviewer each initialled. It is a working schedule, not a legal instrument, and the periods in it are the group's own planning assumptions rather than requirements. The column that mattered most was the last one, because a retention rule without a disposal method is a sentence of good intentions.

RecordWhere it is createdGroup planning assumption for keepingWhy the group wants that windowWho may see itWhat happens at the boundary
Kitchen and store unlock eventsLock and group admin console90 daysService disputes and rota check-backsProperty manager, operations leadDeleted automatically, with a monthly count of what was removed
Front door manual override and mechanical open recordsLock and console, where recorded180 daysBrand, insurance and premises questionsDuty manager onlyRows deleted, monthly total retained without identifiers
Contractor and agency credential eventsCredential system30 days after the credential expiresAgency engagement window and dispute handlingArea managerCredential expiry triggers removal of the join, event rows age out on the shorter period
Credential-to-person mappingRota and payroll systems30 days after a leaver's exitEmployment handling and enquiry responsePeople team onlyJoin detached, counts kept without names
Evidence packs issued in response to a requestGroup mailbox or filing location12 monthsDemonstrates what was provided and whenData-protection reviewerManual review, then deletion
Group monthly export spreadsheetsShared drive12 months rollingGroup reporting and site disputesProperty operations leadsFile deleted, one monthly count row retained
Device telemetry such as battery and connectivity statusSupplier platform12 monthsFleet maintenanceMaintenance contractorSupplier-side retention confirmed with the supplier
Camera and delivery-platform logsSeparate systemsOut of scope for this scheduleNot controlled by the lock programmeSite and area managersKept apart from lock records; correlation refused

Two decisions behind the table are worth pulling out. The first is that contractor credentials were given the shorter period because the underlying engagement ends, and an event trail that outlives the engagement is harder to justify than one that ends with it. The second is that the roster join was given its own line because it is a different record with a different owner, and it is the join that turns an anonymous event into an identifiable one. Permission design around these records is a separate piece of work, covered in Access Policy for Restaurant Groups: Duties Managers Permissions Design, and it is worth reading alongside this schedule because the two are easy to build and easy to contradict.

The record chain, drawn once so nobody argues from memory

On day nine I drew the chain on a whiteboard and asked the group to correct it. The version on the board became the version in the policy annex, because the argument died when the picture was finished.

erDiagram
  SITE ||--o{ DOOR : contains
  DOOR ||--o{ ACCESS_EVENT : produces
  ACCESS_EVENT }o--|| CREDENTIAL : references
  CREDENTIAL }o--|| PERSON : maps_to
  ACCESS_EVENT }o--|| RETENTION_RULE : governed_by
  RETENTION_RULE ||--o{ DELETION_JOB : creates

The point of drawing it is that the chain has two separate deletion points. Deleting an access event does nothing to the credential-to-person mapping, and the mapping is the part that identifies people. Deleting the mapping without touching the events leaves a set of rows that can still be linked through a credential number held in an export somewhere. The group decided, for that reason, that the join was deleted first and the events were allowed to age out afterwards. That order is also why the retention period on the join was set shorter than the period on the events.

The second point is that a retention rule has to create a job, and the job has to be attached to a person. In the diagram the job hangs off the rule, not off the site. That is what prevented the outcome described next.

Options for each record type: keep, count, or stop recording

Before signing the schedule, the group worked through five options for each record type. It is a genuine comparison rather than a formality, because the cheapest option is often the one nobody proposes in the meeting.

OptionWhat an investigator can still doWhat an employee can still doPrivacy exposureOperational burden
Keep every event indefinitelyTrace any access ever recordedBe traced against any historic accessHighest, because every record can be joined to a personLow day to day, high when a deletion is finally demanded
Keep events for a fixed window, then deleteReconstruct a recent service or disputeSee a limited recent windowModerate and boundedNeeds a scheduled job and a monthly count
Keep counts only, delete raw rows quicklyProve how many openings occurred, not whoCannot be identified from the countLowCounts must still be produced somewhere
Do not record the door at allNothing; fall back to keys and sign-offNothingMinimal, but no operational evidenceManual route takes the load, and it must be real
Record against a credential, keep the join elsewhereFollow a credential to a person only with the joinCannot be identified without the joinLow by default, rises if the join is looseDepends on discipline over a second system

The group chose the second option for the kitchen and dry store, the third for the front door, the fifth everywhere until the rota join, and the fourth for one door where a mechanical override already existed and the master key hierarchy was sound. The reasoning behind the front door is worth noting: the front door carries the brand and the building, and its activity is rarely the thing a dispute turns on, so a count is usually enough. The reasoning behind the fourth is in Smart Lock Key Control and Master Key Decisions for Restaurant Groups, which covers what the mechanical layer needs to do if the electronic record is not carrying the load.

The decision that surprised the room was choosing counts for the front door. The operations lead argued for full retention on every door by default, on the grounds that a group should never be seen to record less than it could. The counter-argument was that a record nobody has a use for is still a record someone must protect, answer for and eventually delete.

The delete that stopped running when the admin account changed

On the last morning of the second week I asked for the deletion evidence, and the operations lead went looking for it. There was a monthly report, produced by the platform, listing rows removed. It stopped eleven months earlier. Nobody had noticed, because the report is only interesting when it is missing.

The cause was mundane. The supplier had migrated the admin platform, the scheduled job had not been carried across, and the console no longer removed anything. The rows had continued to arrive, because arrival is the default. A retention schedule written in a document is a statement of intent; a retention schedule with no evidence of disposal is a document that can be quietly untrue.

Two changes followed. The monthly count of removed rows became something the operations lead read, not something the system emailed into an ignored folder. And a quarterly check was added: the group exports its own row totals from the platform and compares them with the platform's removal report. Where the numbers disagree, the group asks the supplier to explain before the next quarter closes. It is unglamorous, and it is the only mechanism I saw that would have caught the failure.

The same principle was applied to everything outside the platform. Exports are deleted from the shared drive on a schedule and removed from mailbox archives by the people who sent them. Printed copies are pulled from site folders at the same boundary. Where a site is being decommissioned, the credential store and the event history that belongs to it are disposed of with the site rather than migrated into the next tenant's estate; the re-keying and end-of-life handling is covered in End-of-life and Re-Keying Planning for Restaurant Groups. Credential data deletion is only credible when it reaches every copy, and most copies are not in the system you bought.

flowchart TD
  A[Event created at the door] --> B{Joins to a named person}
  B -- Yes --> C[Personal record, personal period applies]
  B -- No --> D[Operational record, shorter period applies]
  C --> E[Delete at the boundary and keep the count]
  D --> E

Where the records sit once they leave the site

The group's head office was in one jurisdiction and most of its sites were in another. On day ten we sat down with the map of where data physically lived, and it took an hour to draw because nobody had drawn it before.

Some of it was obvious: the admin console sits with the supplier, somewhere the group could not name. Some of it was not: the supplier's support function had a separate system, and diagnostics uploaded from a site when a reader misbehaved could contain site names, door names and a technician's notes. The group had no visibility of that store and no idea of its lifetime. Some of it was uncomfortable: an area manager in the second jurisdiction could, from a shared dashboard, see unlock activity for doors in the first. Some of it was a commercial question: what happens to a site's records if the site is sold, and does the buyer inherit the event history of the staff who left years ago.

Multi-site smart lock governance turns out to be a retention problem wearing a procurement costume. The group buys once and then keeps records at every site it buys for, and the portability of those records between sites and jurisdictions is a property of the whole programme rather than of any one door. The integration options decision and the choice between standalone and networked locks both move this question, in opposite directions: one adds systems that hold their own copies, the other decides whether the trail sits at the door or on a platform. Before the group signs either, it wants to know where the record is created, where a copy lands, and who can delete it at each of those points. Integration Options for Smart Locks in Restaurant Groups Operations covers the systems side, and Choosing Between Standalone and Networked Smart Locks for Restaurant Groups covers where the trail lives.

CONFIRM-WITH-LOCAL-DATA-PROTECTION-REVIEWER: which framework governs the central store for staff records in the second jurisdiction, whether the group is treated there as a controller for its own employees' records, and what the group must do when a site is sold.

The records nobody had on the inventory

Before the final session I asked the group to list every place a door event could sit. They listed the console. I kept asking until the list reached ten, and the list became the record inventory that now travels with the programme.

The roster file with credential identifiers and names. The shared drive export. The attachment in an email. The spreadsheet on a site manager's laptop. The printed page in a folder behind the till. The supplier's support ticket. The credential-to-person mapping in the rota system. The camera that covers the kitchen corridor and the delivery platform's own record of who collected an order at that door. A tablet left on a pass, logged in as the duty manager, whose session kept the door history on screen.

Two of those are worth stopping on. The first is correlation. Where a camera covers the kitchen door, the unlock event and the video are two records of the same person, and joining them produces a surveillance record regardless of what anyone intended. The answer is not to remove the camera; it is to keep the records apart and to keep the door record for the shortest window the operational question requires. The second is the tablet. A shared device left logged in can expose more history than the export ever did, and it belongs in the inventory with a name against it.

None of this is a device problem. It is a restaurant group privacy review that starts from the record rather than from the sensor. The practical output was a one-page inventory with an owner, a location and a period for each row, and a rule that any new system joining the programme adds a line to that page before it goes live. Credential sharing and hygiene on the same doors is covered in Smart Locks and Credential Sharing and Hygiene: A Practical Control Plan, and the two pieces of work share the same root: an unowned credential becomes an unowned record.

What I would put in the evaluation pack now

The group started the programme with a door list and a budget line. It finished with a record inventory, and I would put the following into any evaluation pack written for a restaurant group.

The retention schedule travels as a schedule to the specification, not as a document that lives in a folder. It is attached to the contract and it names a party on each side. The record inventory exists before the first lock is fitted, because the inventory is what tells you what the schedule has to cover. Each record has an owner who is a person rather than a team, and that person knows how to produce a copy of it on request. The evaluation asks the supplier where the record is created, where a copy lands, how the removal job is evidenced, and what happens to the record when a site leaves the estate. It asks for the credential model, because a credential that carries a name changes every retention line. The RFQ Template for Smart Locks for Restaurant Groups is where the supplier-side questions belong, and Questions to Ask a Supplier Before Choosing Locks for Restaurant Groups is the list I work through before I send one. The credential design question belongs in Smart Locks for Restaurant Groups: Credential Model that Survives High Turnover and Hygiene Constraints.

The pack also carries two things that are easy to leave out. A decommissioning route, because a group that replaces locks in a phased programme runs two record stores at once for a period, and the old one has to be emptied rather than abandoned; the phasing logic sits in Phased Retrofit: Smart Locks Across a Restaurant Groups Portfolio. And a cross-jurisdiction map, because a group that treats each site as its own privacy problem will eventually answer a question about one kitchen in the wrong place.

It is worth pairing this with smart lock commissioning project planning. That is the same question restaurants data retention has to answer here.

The supplier meeting, rewritten as questions

The last thing I did was sit with the group and turn the schedule into questions for the supplier, because a schedule written by the buyer is a statement of intent until the supplier can meet it.

The last question was the one that surprised the supplier team. I asked what their own support staff were allowed to see when a customer reported a reader that failed. Their answer was accurate, careful, and narrower than I had expected, and it went into the pack as a fixed line.

The same question is worked through in more detail in the smart lock motor selection. That is the same question restaurants data retention has to answer here.

Closing note from the last service change

I left on a Friday, at the changeover, watching a chef key the kitchen door in and a second chef key it out twenty minutes later. Two rows, added to a file, joined to a rota name, kept for ninety days, then deleted by a job that somebody reads. That is the whole programme in one gesture, and it is worth saying plainly: the privacy position of a restaurant group's lock estate is decided less by the lock than by whether that job runs, who owns it and what the group does with the roster sitting next to it.

If your group has reached the point where the retention question has been handed to you rather than answered for you, that is the right moment to talk to us. Bring your door list, your record inventory and your cross-jurisdiction map, and we will work through them with you before any hardware is named — request a quote is the way in, and the conversation is about your records rather than our features.

Ihr Warenkorb
Ihre Erfahrung auf dieser Website wird verbessert, wenn Sie Cookies zulassen. Cookie-Richtlinie
Chat on WhatsApp