The State of MDM

See also: Mastering master data management slideshow

It is now a decade since a nascent technology market called master data management (MDM) emerged, aiming to tackle the inconsistency that dogs most large organisations when it comes to handling key shared data such as customer, product and asset.

A survey run by market research firm the Information Difference back in 2008 showed that, on average, a large company has six different systems claiming to be the one and only source of customer data, and nine such systems claiming to be the trusted source of product data.

This lack of consistency in shared data is more than an inconvenience: it makes it very difficult for an organisation to understand its business performance.

Answering basic questions like “what is my most profitable product line?” or “which assets are the costliest to maintain?” proves challenging if there are multiple definitions and codes for products and assets, and if different parts of the business allocate costs differently.

Of course stamping out such issues was what ERP was supposed to do, but quite evidently did not.

One major company I work with admits to having 650 separate major applications, just one of these their ERP system.

Very few global companies have a single ERP instance: dozens of separate instances within a company are common, hundreds not unheard of.

Given this diversity of operational transaction systems, the idea of MDM is to establish a dedicated hub that can act as a trusted single source for master data.

There are different approaches, but the goal is always to end up with a single authoritative source of data about customers, products, locations, people, assets and the like.

This source can be used to populate other applications, such as a corporate data warehouse, which can in turn rely on this data to be of high quality and to be the definitive version.

A host of technologies sprang up over the last ten years providing technology to address this need, some wider in scope than others, but all with the core intention of fixing the problem of inconsistent and poor quality master data.

The same market research firm conducted a survey in the summer of 2012 to try and update this picture of how MDM has evolved, successful projects implementing it, and the overall cost.

This survey followed on a similar one in 2008, the idea being to see what differences, if any, have emerged.

One immediate difference that appeared was in the longevity of the MDM applications. The average MDM application had been live for four years, whereas in 2008 the average was barely over a year.

The success rate changed too: in 2008 the survey respondents from large companies claimed that their MDM implementation was broadly successful in only 54 per cent of cases, but by mid-2012 this had improved to 81 per cent.

It is clear that MDM implementations are maturing, and that people are getting better at understanding what works.

The scope of such projects has widened: 59 per cent of the 2012 survey respondents had opted for an enterprise-wide scope for their MDM project, as distinct from restricting it to a single data domain (such as customer), geography or business unit.

One thing that did not change is that the scope of MDM projects stretches beyond single data domains such as product, with the 2012 survey showing that half of the respondents had over four data domains covered by their MDM implementation.

This is actually similar to the 2008 result, which at the time was controversial, since most vendors at that point were pushing products that dealt with a single data domain, usually either product or customer.

I recall having numerous debates with vendors at that time arguing with me that multi-domain MDM implementations were not the way to go, mostly because their own products could only properly support a single data domain.

This stance was self-serving and in clear contrast to the survey data and my own experience in implementing large MDM projects.

By 2012 virtually the whole industry has acknowledged reality, and the same vendors that were vigorously denying multi-domain was necessary a few years ago are now pushing it as the only way forward.

Just 16 per cent of projects were single domain in 2012.

The 2012 survey shows that an MDM project needs a mean of ten (median five) full-time equivalent staff for the initial implementation, with a mean of seven (median four) staff required full time to support it.

Obviously the numbers depend on the size of the company, but it can be see that this is a non-trivial level of commitment.

The major barriers to success in MDM have not changed: internal politics and poor data quality were the most commonly cited issues.

Broadly, the survey shows an encouraging improvement in maturity and success rates over the last four years.

MDM is growing up.