Centralised or decentralised IT?

Pool software developers and save money, or disperse them to business units and improve efficiency? That's the stark choice CIOs have struggled with for years. But McKinsey's Diogo Rau has an alternative approach.

IT departments have struggled for decades with the centralisation/decentralisation dilemma. Should application developers be aggregated into one large group to control costs? Or should they move into business units to get closer to internal customers and accelerate development?

Careers advice from Tessa Hood:

You won't get promoted looking like that

For many companies, the issue has been more than academic. Those with centralised applications development controlled costs - but watched competitors beat them with better customer service, sales and marketing applications.

Conversely, at many companies with decentralised development, application costs have exploded and islands of automation have mushroomed.

From studying several leading-edge companies and helping dozens of others improve their IT organisations, the answer is a structure that captures the benefits of both centralisation and decentralisation but does not force management to choose one or the other.

We call it the IT demand-supply model. Several major organisations in financial services, logistics, telecommunications and software have adopted the model. While dramatically improving the speed and quality of new systems, they have kept IT costs in check.

Their experiences show IT organisations how to accomplish goals that appear mutually exclusive: directly serving business units and maintaining economies of scale and compatibility in technology.

In industry after industry, companies that built computer systems faster than competitors have gained profound business process advantages.

But while such successes are now commonplace, they obscure difficult work behind the scenes: how these companies organised their development resources to be responsive to business demands yet remain cost-effective.

To deliver new systems faster, many companies in IT-intensive industries have placed applications development teams into the business units they serve.

But they often pay a high price for decentralising development: splintered development activities that make it difficult for functions or divisions to connect systems and share information, and massive duplications of effort and thus high IT costs.

How can companies structure applications development to be both cost-efficient and responsive to business unit needs? A number of companies have addressed the problem by separating the demand and supply elements of applications development.

This model is most useful in companies with legacy environments that often split their technology platforms into such categories as mainframe, distributed systems, and databases.

Exclusive column: The Naked CIO

See what this CIO really thinks…

The Naked CIO: Boadroom stereotypes

The Naked CIO: IT staff disloyalty

The Naked CIO: Cut the bull

The Naked CIO: Animal farm

The Naked CIO: Offshore - or off their trolley?

The Naked CIO: Shadow of the job axe

The Naked CIO: Identity crisis

The Naked CIO: Innovation - same old story

By demand, we mean people who work with business unit and functional managers to understand and shape their system's requirements. They become the voice of the internal customer - clarifying their system needs and taking responsibility for new business processes that depend on new information systems.

They help the business prioritise IT investments and track their progress. They also commission IT projects and select and manage company or external resources that build the systems.

The IT supply organisation provides the resources for developing new applications. It also sets technical strategy: defining the technical architecture so that different applications initiatives across functions or business units can share information.

By centralising the supply organisation, the IT demand-supply model maintains the efficiencies necessary in applications development and the ability to design systems that share information across functional and business unit boundaries. The model frees business units from having to manage a broad range of IT suppliers, external and internal.

There are three typical choices to organising supply resources. The first is the tower model. Here the demand side works independently with supply groups that develop applications and run the technology infrastructure.

The second way to structure supply resources is the competency model. Application resources are organised in broad functional groups such as marketing, manufacturing and R&D.

In addition, there is an infrastructure group. This allows supply groups to build competencies across business units and help their managers standardise on common applications software.

This model is most relevant in companies with synergies across business units - for example, firms that sell different products to the same customers, or the same products to different customer segments.

The lifecycle model is a third way to organise IT supply resources. It is suited to companies that need to make optimal design decisions at the inception of big applications development projects - telecommunications providers are a case in point.

In this model, supply resources are organised by phases of the software development process: design, build and run. Solution managers form the core of the demand organisations.

Typically, they are project managers or business analysts who know both business and IT, are adept at drafting system requirements, and strong enough to tactfully challenge their customers.

The organisation structure of solution managers may mirror the business unit they support, or it may match the portfolios of systems and applications.

Some demand organisations add a strategy role to co-ordinate all IT support for a business in a forward-looking manner. Some organisations pool project managers in a delivery role for the implementation phase - an effective tactic for IT organisations that outsource heavily.

No matter which demand-supply model a company implements, all are likely to bring several advantages in applications development.

The first is providing technology purchasing expertise to business units. The second is much tighter co-ordination of requests for technology services across business units. That lowers costs through volume discounts and eliminating duplicative purchases.

Thirdly, applications are built faster and better because IT suppliers are working with a more technologically sophisticated customer: the demand groups.

As information technology becomes vital to the success of nearly every business, IT and business managers will need to structure their applications development to meet the challenge. Breaking apart IT demand from supply provides the benefits of decentralisation and centralisation while minimising their downside.

Diogo Rau, an associate principal who focuses on IT governance issues, is based in McKinsey's San Francisco office. This column is based on the work of McKinsey's London location and its head of practice, Paul Willmott.