Decentralized Management

What decentralized management means

Decentralized management (اللامركزية الإدارية) is the transfer of the right to take particular decisions from the top of an organisation to the levels below it, under a written and standing rule. The decision then belongs to the position, rather than being a favour granted by whoever holds the position above it.

The definition concerns where the right to decide sits. It does not concern how many people report to each manager, and it does not concern how the organisation has been divided into units. That distinction is what makes decentralization measurable: as long as the right is written down, as a decision rule of the kind formalization covers, it is possible to count the decisions settled where they arise and the ones referred upwards.

Decentralized management is not occasional delegation

Decentralization is also not a single level that an organisation moves to in one step. Between full centralization and full decentralization there are graded levels: the lower level is consulted and the higher level decides; then the lower level proposes and the higher level approves; then the lower level decides and the higher level is informed; and finally the lower level decides and nobody is informed. An organisation may sit on either of the two middle levels, and disagreement can arise when the two parties each believe they stand on a different level of the four.

Delegation is a personal arrangement. A manager entrusts a particular employee with a particular matter, keeps the right, and takes it back whenever the manager chooses. Decentralization is a general rule attached to the position rather than to the person holding it, so whoever occupies the position finds the right there, and whoever leaves the position leaves the right behind.

The difference has practical effects, not only verbal ones. An organisation that runs its affairs through delegation can slide back into centralization whenever a manager changes or an incident occurs, because nothing written stands in the way. Where the rule has been written down, amending it takes a visible decision, and that makes any return to centralization visible rather than silent.

The spending limit as the practical measure of decentralized management

One clear form of the rule is the spending authority limit. Take an organisation that handles 1,000 purchase requests a year, where the branch manager’s spending authority is SAR 5,000, and assume that 820 of the requests fall below that limit:

• Decided where they arise: 820 requests.
• Referred to head office: 180 requests, or 18 percent of the total.

Now raise the limit to SAR 20,000, and assume that the requests referred upwards fall to 60, or 6 percent. The number referred has dropped by 120 requests, which is two thirds of the number previously referred.

The distribution of requests by value is assumed for this example, and in practice it varies from one line of business to another. What does not vary is the shape of the relationship. Raising the limit reduces the number of requests referred upwards by a larger proportion than it reduces their value, because the requests that stop being referred are those between the old limit and the new one, while every request still referred sits above the new limit. Where large requests are few in number and high in value, an organisation that raised its limit and cut the number of referred requests by two thirds may have removed only a small part of the value referred upwards.

The figures in this example and in the examples that follow are assumed in order to show the calculation, and the calculation of the effect is not a recommendation of any figure. We have found no published reference for an appropriate authority limit under decentralized management in the Saudi market.

The return on decentralized management is time, and time can be counted

Assume that referring a request to head office adds 4 working days to the time it takes. The first case then produces 720 working days of delay a year, summed across the referred requests (180 requests at 4 days each), and the second produces 240 (60 requests at 4 days each).

These days appear in no cost record, because they involve no spending. They can nonetheless be the real price of centralization for an organisation: goods that arrived late, maintenance that waited, an offer that lapsed. Anyone weighing centralization against decentralization should put these days on the scale; otherwise centralization appears to cost nothing.

The cost on the other side: decisions spread across units

Take six units buying the same item, each from its own supplier. Assume that each unit buys the same quantity and that the prices they obtained were 100, 102, 104, 108, 110 and 112. The average paid is 106, and the best price any of them obtained is 100.

If all six units bought at the best price, what they pay would fall by 5.66 percent of their actual spending (the gap between 106 and 100, taken as a share of 106). This is the gain that centralization promises in purchasing, and it is a real gain.

The gain depends on a condition that may go unstated: the best price has to be genuinely available to every unit. A unit in another city may be out of that supplier’s reach, and a unit buying a small quantity may not be offered that price. Whoever calculates the saving on the assumption that every unit reaches the lowest price has calculated a saving that cannot be collected in full.

The degree of decentralization differs within one organisation

Asking whether an organisation is centralized or decentralized is a mistake. The correct question is: for which decision? A single organisation may leave operating expenditure to its branches and keep hiring decisions at head office, or it may do the reverse.

A useful document here is an authority table that states, for each type of decision, who decides, who is consulted and who is informed, which is the idea behind a RACI matrix. Without such a table the right remains implicit, and where it sits is discovered at the first disagreement.

The problem grows sharper where reporting lines cross, as in a matrix structure. An employee with two managers does not know which of them decides on a request unless that has been written down, so the request goes up to the person above both of them, and centralization returns without anyone having intended it.

Staff decisions under decentralized management show their effects late

The reasoning above holds for purchasing because whatever is bought can be compared by its price. Decisions on hiring and pay are different. The effect of spreading them across units does not appear on an invoice; it surfaces a year or two later, in a comparison between two employees.

Consider two units that each filled the same job on their own authority, the first at SAR 9,000 and the second at SAR 11,000. The difference is SAR 2,000, which means that the second unit pays 22.2 percent more than the first for the same work (2,000 divided by 9,000).

The difference does not stay hidden. Once one of the two employees moves to the other unit, or the two meet on a joint project, the difference becomes known, and it then has to be either justified by a real difference in the work or addressed by raising the lower salary. Addressing it raises the permanent cost, since the increase stays in the pay from then on.

For that reason an organisation can distinguish between two kinds of decision: one left to the unit, and one left to the unit within a common framework set by head office. The unit chooses whom to hire, for example, while the table of pay and grades stays uniform. That is the point at which the role of the HR function itself is settled.

The conditions decentralized management depends on

• Information at the point of decision. A person given the right to buy who cannot see the price available to others decides with less than head office had to decide with.
• A written rule with an explicit limit. A line such as “Simple matters are decided at the branch” is not a rule, because simplicity is a matter of judgement and becomes disputed at the first mistake.
• Feedback on outcomes to the decision maker. A decision maker who does not see the effect of their decisions in their own figures does not decide better as time passes. Tying results to the manager who controls them is the work of responsibility accounting.
• Review by sample after the event. Prior approval does not serve here, because prior approval is centralization itself under the name of review.

Decentralized management beside span of control, divisional structure and teams that direct their own work

• Span of control. The span covers how many employees report to one manager; decentralization covers what that manager is entitled to decide. A manager with thirty direct reports and no spending limit reaching SAR 1,000 has a wide span and is highly centralized at the same time.
• Divisional structure. A divisional structure covers the axis along which the organisation is divided; decentralization covers which rights have been transferred to the divisions. Division into units without any transfer of rights produces units in name only, with head office deciding on their behalf.
• Teams that direct their own work. A team of this kind is an arrangement for carrying out work within a small group, whereas decentralization is a rule for distributing the right to decide across the levels of the organisation. An organisation can be decentralized without having a single team that fits that description.

What undermines decentralized management

• Transferring the right while keeping the approval. A written authority limit followed by a request to be “kept informed”, without whose reply the decision cannot go ahead, is asking permission under another name.
• Retreating after a single mistake. A mistake in one decision is answered by withdrawing the right from everyone, so the rule is lost, and the withdrawal leaves its mark on the trust of those who stay.
• Leaving limits unreviewed as prices change. A limit set years ago becomes narrower than intended as prices rise, so the share of referred requests grows without anyone having decided that it should.
• Transferring the right without the resources. Someone who is asked to decide but has not been given a budget to decide within becomes responsible for a result whose means they do not control. That is worse than open centralization, because it is called authority and it is not.
• Decentralizing what needs to stay common. Leaving data definitions and classification rules, such as the chart of accounts, to each unit produces reports that cannot be combined, so head office has to rebuild them.

Before the right to a decision is transferred

Four lines are written down: the decision itself, its limit if it has one, who is informed after it is taken, and how it will be reviewed later. Without the first, the right expands to what was not intended; without the second, judgement stays open to dispute; without the third, head office learns of the decision from its result; and without the fourth, the decision does not improve and its weak points go undiscovered.

Which decisions to transfer first is not settled by the definition of decentralized management. That choice differs with the line of business and the size of the organisation.

After a year, one thing is measured: the share of requests referred to head office even though they fell below the limit. That share is a truer guide than the authority tables, because it measures what people do rather than what they are permitted to do. A rise in the share means that the right was transferred on paper but not in practice, and one possible cause is that the person given the right finds nobody to stand behind them if they get a decision wrong.

This is an explanation of the concept, not legal advice.

Related terms

• Formalization
• Responsibility Accounting
• Budget
• Cost Center
• Internal Control