Things go wrong in adult social care. Not because services are careless, but because people with complex needs are being supported by human beings, in real buildings, around the clock. A month in which nothing at all went wrong is not a month that happened. It is a month you did not hear about.

So when the log is empty, the question is not whether the service improved. It is why nothing reached you. Four possibilities, and they are not mutually exclusive.

• Do staff recognise a near miss? Something that could have caused harm and did not is the cheapest lesson available, and the easiest to walk past, because nothing happened.
• Do they have a realistic opportunity to report? A form that takes twenty minutes at the end of a twelve hour shift is not a system. It is a choice you are asking a tired person to make.
• Do they know how? Agency and bank staff often do not, and nobody tells them because nobody expects them to need it.
• Do they feel safe to? This is the one that matters most, and the one you cannot establish by asking directly.

There is a fifth reason, and it is more sympathetic than it first appears. Staff take pride in a shift where nothing went wrong. Not on my shift is a phrase most managers will recognise, and it is usually said with justified satisfaction. But when a clean shift is a source of pride, reporting starts to feel like admitting yours was not one. Nobody has to be hiding anything for that to suppress what reaches you.

Recognising, reporting and learning is how safety, experience and outcomes improve. A service that hears about more is not a service where more is going wrong. It is a service that can do something about it.

Recording is data capture. Learning is what happens when an action changes practice for the better. Between the two sits the step that gets skipped: establishing what actually caused it.

Take a medication error. It gets reported properly, the same day. The action is that the member of staff repeats medicines training and completes two competency observations. Both are done, signed and filed within the fortnight. Every part of that looks like a service responding well.

Now ask whether the person already knew how to administer medicines safely. If they did, nothing in that action has touched the cause. The paperwork is complete, the practice is unchanged, and the service has learned nothing except how to close an action.

Competent people make errors for reasons that have nothing to do with knowledge:

• interrupted mid-round and picking up where they thought they were
• two people whose names or packaging look alike
• a chart that is ambiguous about a variable dose
• a round scheduled across breakfast and personal care
• one senior covering two floors
• a supplier changing a box so it no longer looks like the old one

Retraining a person for a system problem returns them to the same system, only now with an entry on their record. It also means a member of staff quietly carries the cause for something the service produced, which is its own reason to get this right.

Checking whether the action was completed will not catch this. Nor will checking whether it was embedded. Both will confirm, accurately, that you did the wrong thing successfully. The tell arrives later, when the same error reappears with a different member of staff and gets read as a training issue across the team rather than as evidence that the cause was never found.

This is the most reasonable-sounding of the six. A good response is worth recording and worth praising. It is not evidence that there was nothing to learn.

The response is also the part that gets written up in the most detail, because it is the part staff can describe confidently and the part they were trained for. The lesson is usually earlier than that, in the minutes before anybody was responding to anything.

A person is found on the floor. They are helped up safely, checked for injury, observed and recorded properly. That is a correct response and the record will show it. Now work backwards from it.

• What footwear were they wearing, and were they their own?
• What were they carrying, and were they carrying it because nobody was free to help?
• Where were they going, and why at that moment?
• Was anything on the floor, and had it been there before today?
• What was the lighting like, and does it change at that time of day?
• Had they been waiting for support, and for how long?
• Has anything changed recently: a medicine, a new chair, a different room, an infection?

The most under-used source here is the person themselves. They were there and nobody else was. Even where communication is limited, they may be able to show you what they were reaching for, or where they were heading. A record that describes the fall in detail and contains nothing the person said about it has taken a statement from everybody except the witness.

Then the timing question, which survives every textbook response. How long were they on the floor before anybody noticed? If the honest answer is that nobody knows, the lesson is about your noticing rather than your response.

The same logic applies to a near miss. A medicines error caught before administration means staff did exactly the right thing, and the lesson is in whether the catch was the system working or one alert person having a good day: our article on getting PRN protocols right covers a common version of this.

A team meeting is a genuinely efficient way to reach people. You have an audience in one room, sitting down and, with luck, engaged. That is more than most weeks offer.

But a lesson is not learned in a meeting. It is discussed there and shared there, and both of those are worth doing. They are also theory. The lesson itself exists in what happens on the floor afterwards, and a meeting cannot tell you anything about that.

Three separate assumptions are stacked inside that sentence, and each one can fail on its own.

• That everyone who needed to hear it was in the room. Night staff, weekend staff, bank and agency staff usually were not.
• That hearing something changes what somebody does at four o'clock on a difficult shift eight weeks later.
• That the cause was staff knowledge at all, rather than staffing levels, equipment, a care plan that was wrong, or a procedure that made good practice difficult.

This is also the question an inspector arrives at quickly. Not whether you shared the learning, but what changed after you shared it, and how you satisfied yourself that a lesson shared became a lesson learned. Minutes answer the first part of that and none of the second.

Sharing at a meeting is a fast way to close a box. It is not a way to find out whether anything is different. And if nobody checks, nobody knows, until it happens again and you learn it from the incident rather than from your own oversight.

A policy change may well be required, and sometimes it is exactly the right action. The failure is treating it as the whole action.

A policy says what is expected and why. A procedure says how, by whom and by when. Both are theory. They are statements of intent about practice, and practice is a third thing, the only one of the three that a person using your service actually experiences.

Take a real example. You update the policy so that every new admission with a high risk of choking is referred to speech and language therapy for a swallowing assessment. That is a good change and it should be made. It does not, by itself, mean a single referral happens.

Some of what makes it happen belongs in the procedure, and should be written there. If yours does not answer these, that is worth fixing:

• who makes the referral, by role
• where the choking risk is identified in the first place, and by whom
• at what point in the admission, and what triggers it
• who notices if it was missed, and when
• and what happens when the admission arrives at seven on a Friday evening

Strengthen all of that, though, and you still have a document. You have moved from a vague intention to a precise one, which is genuine progress and is not the same as a referral being made.

The test is the next admission. Somebody arrives with a high risk of choking. Was the referral made, by whom, on what day, and can you establish that without asking anybody whether they remembered? That is the only evidence the change worked. Until it happens, you have a better procedure and no information.

Policy revision is also the easiest action to complete, which makes it the one that gets chosen. It can be done at a desk, it produces a dated artefact, and it can be evidenced immediately. Changing what happens on a night shift is none of those things.

Worse, adding to a policy can reduce the chance of compliance. A procedure that was already long enough to be skimmed does not become more likely to be followed by growing. If the honest cause was that the procedure was impractical, lengthening it addresses nothing.

Closing an action feels good. On a list of forty open items, marking one complete is a genuine relief and it feels like progress. That feeling is the risk, because it arrives whether or not anything improved.

Complete, tested and embedded are three different states, and action plans commonly record only the first.

• Complete means the task was done. The training was delivered, the plan was rewritten, the equipment was ordered.
• Tested means somebody went back afterwards and checked practice against the change. Read the records, asked the staff, watched a shift.
• Embedded means it held over time without the manager standing over it, and would still hold if that manager were on leave.

This is why SMART-TE adds Tested and Embedded as sign-off stages after the standard SMART fields. An action closed at complete has recorded intention, with a date and a name against it, which is what makes it one of the purest forms of false assurance available.

Here is how it plays out. The action from an incident was to raise it at a team meeting and share the learning. The meeting happened. The action was marked closed. Two weeks later the same incident occurred. Nothing about that action was dishonest and nothing about it worked.

Two weeks is a short loop, which makes that a cheap lesson. The same failure stretched over months is how a known risk becomes a prosecution. In one published case a person had attempted to leave through windows on several occasions, staff were aware each time, and the fall that followed left them unable to walk: lessons learned from a serious CQC prosecution.

Testing also does a second job that is easy to miss. Going back to check gives you the chance to ask whether the action was addressing the right thing at all. That is the only point in the process where a wrong diagnosis gets caught, which matters because a well-executed action against the wrong cause will close cleanly every time.

Recurrence is the tell. When the same theme reappears three months later, the explanation is often not that the action failed but that nobody ever checked whether it worked. More on this in closing the loop on actions.