Benchmarking against the best, not the average
Most benchmarking compares you to a sector average, one indicator at a time. Frontier methods compare you to the organisations that actually achieve the most with the least — across everything you consume and everything you deliver, at once.
| Traditional benchmarking | Frontier methods | |
|---|---|---|
| Reference point | The sector average | The best observed performance |
| What is compared | One indicator at a time | All inputs and outputs at once |
| Weights | Chosen in advance | Derived from the data, in each organisation’s own favour |
| Result | A ranking | An improvement potential, and the organisations that define it |
Because each organisation is evaluated under the weights most favourable to itself, a gap to the frontier is difficult to dismiss: it means there is no weighting of your own outputs under which you would look efficient.
The methods we work with
Ibensoft works with the two established frontier methods: Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA). They are regarded as the most advanced approaches in benchmarking because both measure against the frontier — the best performance actually observed in the sector — rather than against the middle of the field.
DEA builds the benchmark directly out of observed performance. Nothing is fitted, no cost function is assumed and the target it sets for an organisation is a weighted combination of real, identified organisations.
SFA takes the statistical route. It estimates a frontier and separates genuine inefficiency from random variation and measurement error — which matters when data is noisy, or when a single difficult year would otherwise be read as poor management.
The two are complementary. A result that holds under both is a result worth acting on.
What the methods answer
How much better could we run?
The size of the improvement potential, and which parts of the operation it sits in — not a score, but a quantity.
Who should we learn from?
The specific organisations that define your benchmark. Not an anonymous average, but identified units, and how much each of them contributes to the target you are measured against.
Should these units be merged?
Frontier methods separate the gains from consolidation into what comes from scale, from a better mix of activities, and from simply running the combined organisation well. It is the question a regulator asks about a proposed merger, and the one a group asks about its own subsidiaries.
Are we improving?
Measured across several years, the methods distinguish the whole sector’s frontier moving from your own position relative to it — the difference between improving and merely keeping up.
Why regulators use them
In sectors with little or no competition — electricity and gas networks, water, parts of health care — market pressure does not reveal what efficient operation actually costs. Regulators therefore use frontier methods to establish what a well-run organisation should be able to deliver for, and to set requirements on that basis, so that customers receive the efficiency gains competition would otherwise deliver on its own.
European energy regulation is the clearest case: frontier analysis underpins the revenue frameworks for network operators in numerous countries.
For the organisation on the receiving end, the same analysis is the best available guide to meeting the requirement — and the only rigorous basis for showing that a particular comparison does not hold.
Beyond regulated sectors
The same methods apply wherever comparable units can be measured on what they use and what they deliver.
Industry associations
Where an authority benchmarks a sector, the association is the counterparty in that conversation, and needs its own analysis of the same data to take part in it on equal terms. An association can also commission the analysis on behalf of all its members, so that each one receives an evaluation reflecting its own conditions rather than a sector average.
Municipalities
A comparison between municipalities is only fair once differences in population and local conditions are accounted for. Done properly, the analysis shows what the municipality itself added — not what its circumstances gave it.
Organisations with many units
Branches, offices, plants and subsidiaries can be compared with each other by the same logic, inside a single organisation and with no regulator involved. The methods have for example been applied to bank branches and to member-owned financial institutions.
Where we come in
These methods are demanding to run well and easy to run badly. Small choices — which outputs enter the model, how outliers are treated, whether returns to scale are assumed constant — move every number that follows. Our work is to run them properly, to state the choices we made, and to make the result usable by the people who have to act on it.
That work builds on more than 25 years of research in efficiency analysis. The point is to make it available: insights that used to require a specialist, in the hands of the utilities, municipalities and public services that need them.
Behind Ibensoft
- Peter Bogetoft — Professor of Economics, Copenhagen Business School
- Author of Benchmarking with DEA, SFA, and R and Performance Benchmarking: Measuring and Managing Performance (Springer, 2011 & 2012)
- Written more than 100 scientific articles and published 7 books
- Developer of open-source R software and of Ibensoft's proprietary interactive benchmarking software
- Has advised regulators, industry associations and companies across numerous sectors, including energy, water, health care and financial services, in Europe and North America