Which HR process should a 30-person company fix first?

A founder of a thirty-person firm listed his HR problems for me in this order: the appraisal format was wrong, two people had left without proper handovers, and somebody's leave balance had been miscalculated for eight months. He wanted to start with appraisals, because the appraisal conversation had gone badly the previous week.

Appraisals run once a year. The leave balance error had been running every month for eight. That is the whole of the prioritisation problem: the thing that hurt most recently is rarely the thing costing most.

Two questions, in this order

Rank every HR process you can think of against two questions. You do not need precision — you need the order, and the order is usually obvious once both questions are asked together.

1. 1

   How often does it run?

   Attendance runs daily. Leave runs weekly. Payroll runs monthly. Joining runs whenever you hire. Appraisals run once or twice a year. A broken process that runs daily is doing damage every day; a broken one that runs annually has eleven months before it matters again.

2. 2

   What does one failure cost?

   Count three currencies, not just money: rupees out of the door, hours out of a manager's week, and the chance that somebody good leaves over it. A leave balance error costs little in rupees and a great deal in trust. A missed statutory enrolment can cost in all three.

Multiply roughly. A daily process with a moderate failure cost beats an annual one with a high failure cost almost every time, because it has already failed three hundred times this year.

The usual answer at thirty people

Run the exercise honestly and most thirty-person companies land on the same order. Not because businesses are identical, but because the processes that touch everyone every month are the same everywhere.

• Attendance and leave. Runs daily and monthly, feeds payroll, and produces an argument you cannot win without records. Almost every other number in HR depends on it being right.
• Joining. Runs every time you hire, and every record you later cannot find was one that should have been collected in the first week.
• Payroll input. Runs monthly, and is where attendance and joining errors finally become visible — usually to the employee, in their salary.
• Exit. Runs less often but fails expensively: assets not returned, access not revoked, a settlement argued over weeks after the person left.
• Confirmation and movement. Runs quietly and is missed quietly, which is why probation dates slip past unnoticed.
• Appraisal. Runs once a year. Genuinely important, and almost never the right thing to fix first.

If your list comes out differently, trust your list — but check you ranked by frequency and cost rather than by irritation.

What "fixed" means

A process is not fixed because somebody wrote a policy. It is fixed when four questions have written answers that a manager can find without asking you.

1. Who does it?          A role, not a person's name.
2. Who approves it?      With a limit, where money or time is involved.
3. By when?              A date in the month, not "promptly".
4. Where is the record?  One place, named, that outlives the approver.

If any answer is "it depends" or "ask the founder", it is not fixed.

That is deliberately small. A thirty-person company does not need a manual; it needs four answers per process, written where managers look. The manual comes later, if at all.

Do one, all the way

The most common mistake after prioritising is to start all of them. Four half-implemented processes leave you worse off than one finished, because employees now face four sets of half-rules and fall back on asking whoever seems friendliest.

1. 1

   Pick one process

   The top of your ranked list. Resist adding a second because it seems related.

2. 2

   Answer the four questions

   An afternoon, not a project. Write them on one page.

3. 3

   Run one full cycle

   One month of attendance, or three actual joiners. A process that has not survived a real cycle has not been tested.

4. 4

   Fix what the cycle exposed

   There will be two or three things. There always are, and they are cheap to fix now.

5. 5

   Then start the next one

   Not before. The discipline of finishing is what makes the second one faster.

When you genuinely cannot tell

Sometimes the list is long and the ranking is not obvious, usually because the complaints you are hearing are several symptoms of one underlying gap. That is the case an HR process audit is for: it follows a real month through every process you already run and reports which of the complaints are the same problem wearing different clothes.

It is also worth doing when the answer looks obvious but keeps not working. If attendance has been "fixed" twice and still produces arguments, the problem is usually upstream of attendance — in who is allowed to approve a correction, or in the fact that nobody announced the rule. An audit finds that faster than another attempt at the same fix.

And if the process you land on is joining, start narrow. Getting employee onboarding right removes a surprising share of later problems, because the documents you cannot find at exit are the ones nobody collected at entry. If it is attendance, the same logic points at attendance, shift and leave, which is where the working-day data every other number depends on comes from.

One more thing about appraisals

Appraisals almost always come last in this exercise, and founders almost always resist that. It is worth saying why they can wait: an appraisal is a conversation about a year of work. If attendance, leave and confirmation records are unreliable, the appraisal has nothing solid to stand on anyway. Fix the record first, and the conversation gets easier on its own.

A worked example

A thirty-two person engineering firm listed five problems: appraisals were inconsistent, two exits had gone badly, leave balances were disputed, a new joiner's salary had been wrong, and nobody could find an old appointment letter. Five problems, five potential projects, and a founder with no spare week.

Ranked by frequency and cost, the list collapsed. Leave balances ran monthly and fed the wrong salary, so those two were one problem. The missing letter and the bad exits were both the absence of a single filing place, so those were one problem. Appraisals ran once a year and depended on records the other two produced.

Five problems became two, in an order: fix leave and attendance so payroll stops inheriting errors, then fix where documents live. The appraisal was left alone for a year and was easier when it came round, because there was finally something recorded to discuss.

That collapse — from a list of complaints to two underlying gaps — is the useful output. It is also why it is worth doing the ranking on paper rather than in your head: the relationships between the complaints only become visible when they are written in a column.

If your own list will not collapse like that, it is usually a sign the complaints are genuinely separate, or that you are too close to them. That is the point at which an outside pass over one closed month earns its keep, which is what an HR process audit is.