If you are considering an online analytical processing product, a close look at the results of the survey conducted last year into Olap could pay dividends, writes Nigel Pendse
It is 10 years since Ted Codd, often called the father of relational databases, redefined the world of decision support systems by publishing his 12 laws for online analytical processing and coining the term Olap. The concepts behind Olap go back three decades, so the idea obviously has sticking power.
The aim of last year's Olap Survey 2, published by Survey.com, was to find out what the benefits of Olap really are. You might even have been one of the 2,236 people from 49 countries who participated in the survey.
About 60% of the respondents to the survey were interested in Olap, but were not yet users, and another 10% were actively looking, but had not yet bought. The remaining 30% were existing Olap users, and they each answered dozens of questions about how they chose and used the products they did. They were also asked about the rate at which business goals were met and the extent to which eight possible business benefits were achieved.
Not surprisingly, "soft" benefits, such as faster, more accurate reporting were much more likely to be achieved than "hard" benefits, such as increased revenues or headcount savings, particularly in IT. Almost 70% of respondents said improved reporting was a proven benefit, while only 23% had proven revenue increases through better sales and marketing analysis; and just 19% had saved IT headcount (5% said it had actually increased).
Respondents in the UK appear more sceptical than those in other countries because they generally reported lower benefit achievement levels, while those in the US were more likely than average to be pleased.
But the benefits measurement had a more subtle use: an average benefits achievement index was calculated based on the scores for all eight benefits, and this was used to calibrate many of the other results. This was thought to be a better, if more indirect, way of finding out how success rates varied, depending on how organisations chose products, which products they chose and how they implemented them. And this was where some real surprises emerged.
This is not just academic: if you are one of those who is interested but has not yet started with an Olap project, you can take advantage of knowing what does, or does not, seem to be associated with projects that delivered the most benefits.
The first surprise was to discover that conducting a formal, multi-product evaluation really pays. Only half of the sample did this, with the rest either evaluating just a single product, or simply buying an Olap add-on from an incumbent supplier. But those who conducted a proper competitive evaluation were not only more likely to have met their business goals, but also scored higher on every one of the eight possible business benefits. Perhaps this was because they found a more appropriate product, or perhaps the process itself was educational, but either way, make sure you do not skip this important step.
Probably not coincidentally, the product whose buyers were least likely to have performed a formal evaluation on was SAP Business Information Warehouse (BW) - and this also was the product with the lowest benefits achievement index, as well as the second-lowest goals achievement rate. Paradoxically, despite achieving the worst results, SAP customers still proclaimed the highest loyalty in the survey.
At the other extreme, despite achieving the highest benefits, ungrateful customers from business performance software solutions provider Brio were relatively disloyal. Predictably, the few Computer Associates Olap customers were by far the most disloyal in their response.
The differences in benefit achievement rates are even more striking among implementers. Many sites did all the implementation work in-house, but they had a lower success rate than sites that also used some external help, even when spending only a modest £10,000 on third-party consulting. But spending more than £300,000 seemed to lead to a declining success rate.
The consultants who performed by far the best were the specialist business intelligence/Olap consulting firms; much the worst were the large, general-purpose consulting firms, including the big five. The big firms were reported to cost more, took longer and apparently delivered less successful projects.
When it comes to choosing products, buyers put the greatest emphasis on product functionality and ease of use for end-users; query performance is regarded as only half as important. But when it comes to their real-world deployments, poor query performance is by far the most commonly reported technical problem, so perhaps businesses should be paying more attention to that aspect from the start.
Because it is an area with so many user complaints, the survey analysed performance intensively, with no fewer than 24 of its 226 pages devoted to the subject. There are many ways of measuring performance, but Microsoft Analysis Services and Hyperion Essbase scored best overall, while Brio, BusinessObjects, MicroStrategy and SAP BW were much worse than average.
Curiously, though Relational Olaps (Rolaps) are normally expected to be better on build/load/calculate times than Multidimensional databases (Molaps), none of the Rolaps came close to challenging the leading Molap performer, Microsoft Analysis Services. There was no significant difference in the performance of Unix and Windows servers with comparable data volumes and product mixes, perhaps because Unix servers typically support more concurrent users.
Olap sales are continuing to do well even in today's slow market, so many new sites must be buying and deploying it. The benefits are indeed proven, but you can maximise your chances of achieving them if you learn from the experiences of the large number of people who took the trouble to participate in the Olap Survey 2.
Goals and benefits
- When it comes to choosing products, buyers put the greatest emphasis on product functionality and ease of use for end-users; query performance is regarded as only half as important
- When it comes to real-world deployments, poor query performance compared to other product-related problems is by far the most commonly reported technical problem. So perhaps you should be paying more attention to that aspect up front
- Olap sales are continuing to do well even in today's slow market, so many new sites must be buying and deploying it. The benefits are proven, but you can maximise your chances of achieving them if you learn from the experiences of the large number of people who participated in the Olap Survey 2.
Olap survey details
The Olap Survey 2 is published by Survey.com (www.survey.com/products/olap2/) and is also available in the UK from Business Intelligence (www.business-intelligence.co.uk/re/os2).
Nigel Pendse is the author of The Olap Survey and The Olap Report (www.olapreport.com). He has almost 30 years' experience of the business intelligence industry, as a user, supplier and, since 1994, as an independent industry analyst and consultant