Skip to content
Internal audit: how to make it worth the cost
Boardroom practice

Internal audit: how to make it worth the cost

Ten practices that decide whether an internal audit tells the board something new, written for CEOs, CFOs and directors.

Most internal audits fail quietly. Reports arrive on time, findings get closed, and the board learns nothing it did not already know. The cause is rarely the auditor's skill. It is how the function is set up, what it is allowed to examine, and how the board reads what it receives. The ten practices below are professional judgement, except where a law is named.

01Plan from what the board takes on trust

Most audit plans are a list of departments. A better starting point is a list of things the board accepts without checking: the reported margin, the value of stock, that receivables will be collected, that statutory dues are paid, that customer data is safe.

For each one, ask who checks it today. The answer may be a management review, the statutory auditor, an outside certification, or nobody. Internal audit belongs where the answer is "nobody" or "only the person who does the work". This also shows where two parties are checking the same thing.

02Check who holds the three levers

A reporting line on an organisation chart proves little. Independence depends on three things: who can remove the internal auditor, who approves the fee or budget, and who can change the plan or edit a draft report. If the head of an audited function holds any of these, findings in that function will soften.

For listed companies, the SEBI listing regulations place the appointment, removal and remuneration of the chief internal auditor under audit committee review. An unlisted company can adopt the same practice by choice.

03Audit decisions, not only transactions

Voucher testing shows whether junior staff followed the process. Larger losses usually sit in decisions senior people take around the process: discounts beyond policy, a vendor chosen without bids, a credit limit raised for one customer, a write-off, a year-end journal entry, the terms of a related-party deal.

The plan should name exceptions and overrides as a subject in their own right. Whether these ever come into scope is the real test of independence.

04Test every record where the data allows

A sample of a few dozen invoices from many thousands will rarely find a problem. Where the accounting system holds the data, a test can run across every record. Useful examples:

  • the same invoice paid twice
  • a vendor sharing a bank account, address or phone number with an employee
  • purchase orders split to stay just under an approval limit
  • entries posted on holidays or backdated
  • vendor bank details changed and then changed back

None of this needs expensive technology. It needs access to the data and someone who asks for it.

05Read for causes, not for the count

Twelve findings across four locations may be one problem, such as nobody owning the vendor master, or one person holding too much system access. Fixing each finding closes the report and leaves the cause in place, so the findings return next year.

Ask the auditor to group findings by cause and to say which two or three causes matter most. The number of findings measures nothing.

06Make "risk accepted" a decision someone signs

Management may decide not to fix a finding. That is legitimate. But the decision should be signed by someone senior enough to own the consequence, and the board should see the list.

The same applies to action dates that keep moving. And "closed" should mean the auditor retested the fix, not that a manager reported it done. Without these three rules, the easiest way to close a finding is to stop discussing it.

07Bring internal audit in before a change, not after

A new accounting system, a new plant, an acquisition or a new sales model all rebuild the controls. Controls are cheapest to get right while the change is being designed. A review of access rights, approval flows and data migration before go-live costs far less than an investigation later.

The auditor should advise at this stage, not design or approve. An auditor who signs off a design cannot review it independently afterwards.

08Do not turn the auditor into a second approver

Some companies ask internal audit to check every bill before payment. It feels safe. It creates two problems. Line staff relax, because someone else will catch the error. And the auditor becomes part of the process, so nobody independent is left to review it.

If a pre-payment check is needed, place it inside finance and let internal audit test whether it works.

09Choose the model on purpose

ModelStrengthWatch for
In-house teamKnows the business and its peopleCareers depend on the managers being audited; skills can be narrow
Outside firmSpecialist skills and fresh eyes; easier to say hard thingsThin knowledge of the business; junior teams; a low fixed fee that buys checklist work
MixedAn in-house head with outside specialists for areas such as IT or fraudNeeds a head senior enough to direct the specialists

One legal limit applies to every model. Under Section 144 of the Companies Act, 2013, a company's statutory auditor cannot provide its internal audit.

10Review the internal audit itself once a year

Four questions are enough. Did it tell us something we did not know? Did any finding change a decision, a process or a responsibility? How long did reports take after the fieldwork ended? How much of the effort went to our biggest risks?

The audit committee, or the board where there is no committee, should also meet the internal auditor once a year without management present. Three questions are worth asking in that meeting: what were you unable to look at, what changed between your draft and final reports, and where were you refused data?

An internal audit that never reports anything uncomfortable is either auditing a perfect company or not looking hard enough.

This article is for general information and is not professional advice.

Keep reading