The Check That Quietly Stopped

Every venue has a list of things that are supposed to happen.

The fridge temperature gets logged. The float gets counted at open. Somebody walks the fire exits before doors. The ID scanner gets used on everyone who looks under thirty.

⚠️ The dangerous checks are not the ones nobody does. Everyone knows those are missing. The dangerous ones are the checks that used to happen, stopped, and left behind a sheet that still looks exactly like a check is happening.

The short version

• A check can stop without anyone deciding to stop it
• The paperwork usually outlives the check by weeks
• "The sheet is filled in" is not evidence that anyone looked
• Two signs give it away: identical entries and entries written at the wrong time
• The fix is to check the check, not the thing being checked

Why checks die quietly

How a check starts	How a check ends
A rule, a meeting, a named person	Nobody decides. It fades.
Everyone knows it is happening	Nobody notices it stopped
A fresh sheet	The same sheet, still being filled in

A check almost never ends with a decision. It ends when the person who cared about it leaves, or gets busy, or when it gets moved to "whoever is on." After that it keeps its shape for a while — the clipboard is still on the wall, the column still gets ticks — while the actual looking stops.

One: the full sheet that means nothing

This is the most common form.

Monday     4°C   ✓
Tuesday    4°C   ✓
Wednesday  4°C   ✓
Thursday   4°C   ✓
Friday     4°C   ✓
Saturday   4°C   ✓

⚠️ A real fridge does not read exactly 4°C for six days. It drifts. It reads 3 after a delivery and 5 after a busy Saturday with the door open.

A column of identical numbers is not a well-behaved fridge. It is somebody writing the number they expect to see.

The general sign

Real readings   vary a little
Copied readings do not vary at all

⇒ This applies to float counts, glass counts, cellar temperatures — anything measured. If it never moves, nobody is measuring it.

Two: the entry written at the wrong time

The second sign is timing.

Fridge check    supposed to happen   before open
Written         at close, all at once

⚠️ A check written at the end of the night, for the whole night, was not a check. It was a memory, filled in to make the sheet complete.

How to spot it

Look at the handwriting and the pen. Six entries in the same ink, the same slant, the same pressure, were written in one sitting. Six real checks through a night look slightly different each time.

⇒ This is not about catching anyone. It is about knowing whether the check you are relying on is real.

Three: the check that moved to "whoever is on"

A check with a named owner tends to survive. A check owned by a role tends to fade.

"Mia walks the exits before doors"     - survives
"Whoever opens walks the exits"        - fades

⚠️ "Whoever is on" means that on any given night, the person responsible can reasonably believe it is someone else. Two people each assume the other did it. Nobody did.

The fix

One   Every check has one named person per shift
Two   The name is said out loud at the start of the shift
Three The named person signs, not "whoever"

⇒ This is the same rule that fixes unanswered messages — covered in the enquiry nobody knows you missed. A job without a name has an owner of nobody.

Four: the automated check that silently died

This one is newer and growing. It applies to anything that runs by itself: a temperature sensor, a till alert, a booking reminder, a camera that is supposed to record.

⚠️ An automated check that stops working usually does not tell you it stopped. It simply goes quiet — and quiet looks exactly like "nothing is wrong."

A real example

We run an automated process that forwards incoming messages to the person who answers them. It checks every thirty seconds and writes a small line saying it is alive.

One night it stopped. It had been writing its "I'm alive" line faithfully — and then it simply was not running any more. Nobody noticed for eight hours, because no messages arrived in that window, so there was nothing missing to notice.

What we had        a process that wrote "I'm alive"
What we lacked     anything reading that line
Result             8 hours down, noticed by accident

⇒ Writing a heartbeat is only half a check. Something has to read it. We had built the half that writes and not the half that looks.

The rule for anything automated

An automatic check needs a second check
that notices when the first one goes quiet

That second check can be simple — a person who looks at a light, a daily message that says "all fine," anything whose absence you would notice.

Five: how to test whether a check is alive

You do not need to audit everything. You need to spot-test.

One   Pick a check you rely on
Two   Ask the named person when they last did it
Three Go and look at the thing yourself, right now
Four  Compare what you see to what the sheet says

⚠️ Step three is the one that matters. Reading the sheet tells you whether the sheet is complete. Looking at the fridge tells you whether the fridge is cold.

The gentle version

If you do not want this to feel like an inspection, do it together.

"Let's check the fridge together - what does it read right now?"

⇒ Then compare that number to what was written earlier tonight. If they match to the degree, twice in a row, the check is probably real.

Six: which checks to test first

Not all checks carry the same weight.

Test first   anything where failure hurts a person
             - fire exits, ID checks, allergy notes, barred-guest flags
Test next    anything where failure costs money quietly
             - float counts, stock, till reconciliation
Test last    anything where failure is visible fast
             - lights, music, cleanliness

⚠️ The first group is also where a dead check does the most damage and is least likely to be spotted. A missing ID check produces no visible sign until the night it matters.

⇒ On the specific case of a tool that stopped being used: the ID scanner you stopped using.

Seven: why "we've never had a problem" is not reassurance

The absence of incidents is exactly what a dead check produces — right up until the incident.

Live check    catches problems, so you rarely see them
Dead check    catches nothing, so you also rarely see problems

⚠️ From the outside, both look like "no problems." You cannot tell them apart by the incident rate. You can only tell them apart by looking.

The uncomfortable arithmetic

A check that has quietly stopped is not neutral. It is worse than no check, because people still believe it is happening and stop taking the extra care they would take if they knew it was not.

Numbers worth keeping

One   How many checks you rely on (list them)
Two   How many have one named owner per shift
Three How many you have personally spot-tested this month
Four  How many showed identical or batch-written entries

⚠️ Three is usually zero. That is not a criticism — it is the normal state, and it is exactly the gap this article is about.

⇒ On choosing which of these to put in front of the team: the number you check first.

The card for the office wall

────────────────────────
  Is this check alive?

  1. Who owns it tonight? (a name, not a role)
  2. When did they last actually do it?
  3. Go and look yourself - right now
  4. Does it match the sheet?

  Watch for:
   - numbers that never change
   - a whole night written in one sitting
   - "whoever is on"
────────────────────────

Common objections

"Our staff are honest"

⚠️ This is not about honesty. Most dead checks are not faked on purpose. They are filled in from memory by someone who believes the thing is fine — which it usually is, right up until it is not.

"The sheet is always complete"

⇒ That is the symptom. A complete sheet tells you somebody filled it in. It does not tell you somebody looked.

"We'd notice if something went wrong"

⇒ Some failures are loud. Others — an expired ID, a warm fridge, an allergy note not read — are silent until the one night they are not. The silent ones are the reason the check exists.

"We don't have time to audit"

⚠️ You are not auditing. You are spot-testing one check a week. Five minutes, and it catches the pattern long before an inspector or an incident does.

"The automated system would tell us"

⇒ Only if something is watching it. An automated check that stops usually goes quiet rather than raising an alarm — and quiet looks like everything is fine.

What to do this week

One   List every check you rely on. All of them.
Two   Next to each, write one name - or "nobody".
Three Pick the most important one with "nobody". Give it a name.
Four  Spot-test one check yourself. Compare to the sheet.
Five  If anything runs automatically, ask: what tells us when it stops?

Summary

• Checks stop without anyone deciding to stop them
• The sheet outlives the check — a full sheet proves someone wrote, not that someone looked
• Identical numbers and batch-written entries are the two tells
• "Whoever is on" means nobody; give every check a name
• Automated checks need something that notices when they go quiet
• "We've never had a problem" looks the same whether the check is alive or dead

A check that stopped is worse than no check at all, because everyone is still relying on it.