How to spot HR bottlenecks in a growing company

A company of fifteen runs on a spreadsheet and a good memory. A company of fifty does not. And the problem is that nobody announces the day you cross that line: there is no alarm, only an accumulation of small failures that one ordinary month stop being anecdotes.

A bottleneck in HR is any point where information stops, is duplicated or goes missing before it reaches whoever has to decide. You find them by measuring three things: how long a piece of data takes to become available, how many times the same data is typed in, and how many decisions are taken without it.

This article is a step-by-step diagnosis. It is not here to convince you of anything: it is here so that by the end you know exactly where your operation jams and what to measure to prove it.

Why bottlenecks appear as you grow, and not before

With fifteen people, the person running HR knows everybody. They know who is on holiday because somebody told them in the corridor, they know Marta starts later on Tuesdays and they know the Saturday shift is always covered by the same team. The information lives in their head and it works.

Past forty or fifty people that stops scaling, for three reasons at once:

• Memory stops being reliable. Not through any failing, but through volume: fifteen exceptions can be remembered, eighty cannot.
• Intermediaries appear. It is no longer one person deciding, it is area managers who need the same information and do not have it.
• Legal obligations change tier. From fifty people the equality plan, the whistleblowing channel and the works council all apply, and each demands documentation nobody used to ask for.

What fails is not the team. It is that a system designed for fifteen is still running at fifty, and nobody has had time to change it.

Step 1: map the journey of one piece of data, end to end

Take one real absence from last month. Somebody asked for three days. Now answer, with names and tools:

1. Where did they ask? Email, WhatsApp, a form, in person.
2. Who approved it, and where is the record of that approval?
3. How did whoever builds the rota find out those days needed covering?
4. How did it reach payroll?
5. Where is that data today if an inspection asks for it tomorrow?

If at any of those five steps the answer is “I told them” or “I have it in a spreadsheet”, you have just found a bottleneck. It is not a failing of the person: it is a point where the data depends on somebody remembering.

Do the same with a clock-in and with a shift swap. Three journeys mapped is usually enough to see the pattern.

Step 2: count how many times the same data gets typed in

It is the most revealing measure and the most uncomfortable. One approved holiday day usually gets written in four places: the email it was asked in, the shared calendar, the rota and the sheet that goes to the accountant.

Every new copy multiplies two costs:

• Time. Measurable: time yourself for a week on copying data from one place to another.
• Divergence. Four copies of the same data are four versions that can stop matching, and when they do nobody knows which one is right.

Centralising employee data is not a luxury for large companies: it is what takes that number from four to one. And the indicator for knowing you have managed it is simple, count the places the same piece of data lives in.

Step 3: measure the delay between something happening and somebody knowing

This is the bottleneck people see least, because it produces no visible errors: it produces decisions taken late.

Three questions with numerical answers:

• If somebody does not clock in on time today, when does their manager know? At the time, at the end of the day, or at month end?
• If sick leave starts on a Monday, when does whoever plans Thursday’s rota find out?
• At month end, how many days does it take to have the hours ready for payroll?

A month’s delay in attendance tracking means corrections are made to a period already closed, when there is nothing left to correct. That is where the hours that do not add up appear, and the arguments nobody can settle, because both sides remember different things and there is no record.

Step 4: look for shifts covered twice, or not at all

Shift management has an unmistakable symptom: last-minute swaps get negotiated over messaging and never reach the official rota. The result is two people turning up for the same shift, or nobody.

To measure it, compare a month of published rota against actual clock-ins. Every difference is a change that happened outside the system. If there are more than a handful, the rota has stopped being the source of truth and become a forecast nobody consults.

There is a legal consequence too: the difference between what was planned and what was recorded is the first thing an inspection compares, and repeatedly changing schedules without notice can amount to a disguised substantial change of conditions.

Step 5: ask what happens if the key person goes on holiday

It is the quickest test of all. In many SMEs there is one person who is the system: they know who is off, who has days owing and why that shift is always done by the same people.

If that person goes away for two weeks and the operation suffers, the bottleneck is not in a process: it is that the knowledge was never written down. It is the most expensive of all because it stays invisible until it shows up, and by then it is late.

Step 6: check what documentation you could produce tomorrow

This step turns the diagnosis into something verifiable. With nothing prepared, answer whether you could produce today:

• The time record of any employee from the last four years.
• The monthly hours summary for anybody on a part-time contract.
• The pay register with means and medians broken down by sex.
• The history of one specific absence with its approval date and who approved it.

All four are mandatory, and in all four cases not having them is a breach regardless of whether anybody was harmed. If gathering them means asking three different people for things, that is not a compliance problem: it is the same bottleneck, seen from outside.

What centralising actually solves, and what it does not

Centralised HR management attacks one specific cause: that the data is written once, at the moment it happens, and everything else reads it from there. A clock-in goes in once and feeds the rota, the hours calculation, the absence report and the legal record without anybody typing it again.

That removes the bottlenecks in steps 2, 3, 4 and 6. That is a lot, and it is measurable.

But it is worth being honest about what it does not fix. HR software does not decide a shift policy nobody has written, it does not replace a conversation about why people leave, and it does not turn a habit into a criterion. If step 1 reveals that absences are approved depending on who asks, centralising will make that arbitrariness more visible, not smaller. Which is useful, but it is not the same thing.

A diagnosis in one afternoon

The six steps take an afternoon and require buying nothing:

1. Map the journey of an absence, a clock-in and a shift swap.
2. Count how many places the same piece of data lives in.
3. Measure the delay between something happening and somebody knowing.
4. Compare a month of published rota against actual clock-ins.
5. Ask what breaks if the key person leaves.
6. Check what mandatory documentation you could produce tomorrow.

When you finish you will have a list of concrete points, not a feeling. And that list is what turns “our processes are a mess” into something you can prioritise, budget for and, above all, check again in three months.