The OLAP market grew faster than predicted in 2006

Introduction

Our projection of 12 percent growth in 2006 turned out to be somewhat pessimistic: actual growth was about 16.4 percent, slightly higher than the 13.8 percent in 2005. There was a little further market consolidation in 2006, unlike in 2004 and 2005 when the market actually fragmented slightly.

Despite the small drop in the growth rate, the OLAP market continued to grow faster than most other enterprise software sectors, though still at a lower rate than in the 1990s. However, despite the growth in OLAP sales and usage, it is becoming ever more difficult to estimate the exact size of the whole market and the individual market shares. The larger generalist vendors — Microsoft, Oracle, SAP, Business Objects — cannot even measure their OLAP business themselves, because their OLAP capabilities are often delivered as part of larger, bundled products and account for a minority of their revenues. For example, Microsoft Analysis Services and the OLAP capabilities of SAP BW are not sold separately, but are included as part of product suites. 数据挖掘研究院

Other vendors, such as Hyperion, do track but no longer publish details of their individual product sales, making it harder to isolate their OLAP business. It is therefore necessary to estimate the extent to which customers deploy the OLAP products and this is necessarily approximate. Our aim is to provide a good indication of trends, but not to claim precision to the last decimal place.

数据挖掘实验室

This market is still much more open than most enterprise software markets. Microsoft has now clearly overtaken Hyperion Solutions to become by far the largest OLAP vendor, but neither could be called dominant. Although Microsoft’s lead is likely to increase further in 2007, it will still not have anything like the dominance it enjoys in some other markets. In fact, it might have an OLAP revenue market share little higher than Oracle and the newly merged Hyperion Solutions enjoyed in the late 1990s. 数据挖掘研究院

Similarly, Oracle and SAP, which dominate in other markets, are both relatively weak in the OLAP market, and Oracle in particular has had several years of decline.

数据挖掘研究院

• Despite our expectations that Oracle might start to recover in 2005, it failed to do so in that year or 2006, and evidence from The OLAP Surveys suggests increasing disillusionment with its traditional BI offerings. It is too early to tell whether the BI tools it acquired with Siebel will do any better.
• Many SAP sites feel obliged to install and use Business Information Warehouse (BW), though there are very few successful deployments and large volumes of shelfware. But, despite this, we estimate that SAP BW deployments grew slightly in 2006.

Background to the market analysis

Measuring OLAP market size and share is a lot more difficult than a casual observer might suppose, and it is getting harder. First, one has to decide which companies actually belong in this sector. We do not accept the assertions of all those companies claiming to offer OLAP products, and we do regard some products as being in the category even if the vendors have (for marketing reasons) chosen not to use the term. For example, ‘business/corporate/enterprise performance management’ is currently a more fashionable term, though it is often used to describe exactly the same products that were called OLAP a few years ago. We do include pre-built applications, as well as the toolkit products from which such applications should be built. Our guiding principle in this is the FASMI test.

数据挖掘研究院

However, this is not the end of the process. There are many complicating factors that must not be ignored (although we suspect that many other market researchers tend to gloss over them): 数据挖掘研究院

Many vendors of OLAP products and applications are also active in other areas, so we have to estimate the proportion of their total revenues represented by OLAP. Some vendors are unable or unwilling (especially if they are not doing well) to help us make these estimates; and even those that do try to help cannot always provide reliable information (if only because their own internal records do not clearly distinguish OLAP related from non-OLAP related revenues). Some OLAP-related products, such as BusinessObjects or Hyperion Intelligence, can even be regarded as only partially OLAP, especially if the OLAP capabilities are an optional or limited extension. As we try to estimate only the OLAP related business, the revenues used to calculate market shares in such cases are much less than the total revenues of these companies. 数据挖掘研究院

Many larger, non-BI companies have acquired smaller, pure OLAP vendors, so these previously clearly-identified OLAP revenues are no longer reported separately. Some of the many examples include Oracle’s purchase of IRI Express, Seagate’s purchase of Holistic Systems (and Business Objects’ subsequent purchase of Crystal Decisions), Geac’s purchase of Comshare and Systems Union’s purchase of MIS AG, with Infor subsequently acquiring both Geac and Systems Union in 2006.

数据挖掘研究院

Some vendors bundle their OLAP and non-OLAP products. For example, until Microsoft acquired ProClarity in 2006, it had no separately saleable OLAP products apart from Data Analyzer (launched in late 2001 and now discontinued) and the Business Scorecard Manager 2005, but bundles Analysis Services with SQL Server and PivotTables with Excel. The Excel OLAP add-in is a free download. Similarly, SAP BW is bundled as part of several solutions, rather than being sold separately. In such cases, where only a subset of buyers are likely to use the ‘free’ OLAP functionality, we have to estimate the proportion who use the OLAP component and assign part of the value of the total purchase cost to OLAP. This becomes harder as time passes, as the OLAP dividing line gets more blurred. 数据挖掘研究院

A number of the smaller vendors are US private companies who are not obliged to publish financial information. Some choose to do so voluntarily, but in other cases we have had to make estimates. 数据挖掘研究院

Some vendors concentrate on direct sales, so their own revenues are equal to what final customers paid. These figures are a good guide to market size and share. But other vendors make many of their sales through resellers, and the original authors’ revenues are only a fraction of the true market value. In order to treat all vendors equally, we do not want our figures distorted by differences in their sales models, so we try to gross up these companies’ indirect revenues to what we believe to be the ‘retail’ market value. This has become easier as public companies now disclose more information about the split between their direct and indirect revenues.

数据挖掘研究院

Most OLAP products require assistance in implementation. We include these estimated implementation costs in the market totals, to enable proper comparisons between the products that are sold as complete applications and those sold as toolkits from which applications can be built. Many, but not all, of the vendors have their own revenue earning services groups to provide this assistance and the resulting fees are included in their total revenues. Most products are also implemented by independent third-party firms, and the proportion varies by supplier. Again, for consistency, we have tried to treat all vendors equally, by making suitable allowances for external consulting. We only include services directly relevant to the implementation of the OLAP product, not the total data warehouse consulting costs (which may be much larger). 数据挖掘研究院

We include some third-party revenues when calculating the OLAP server market shares. For example, tools, consulting and training can all be regarded as part of the market share for OLAP servers such as Analysis Services, Essbase and TM1. 数据挖掘研究院

Many OLAP solutions are implemented using both OLAP and non-OLAP technologies. As far as possible, we exclude the latter from the figures (for example, we exclude all relational database and ETL costs, which often exceed the OLAP costs). 数据挖掘实验室

We believe that the results of all these adjustments give the fairest possible assessment both of the size of the OLAP market and the shares of the various vendors. Of course, great precision is not possible and it would be pointless to look at shares to more than two significant figures (this is, of course, true of almost all market surveys and opinion polls, but some people ascribe a spurious degree of accuracy to the results). For each product, we describe how its share was calculated in the individual product reviews.

Readers should not attempt to view the OLAP market in isolation. It overlaps many other markets, including analytical applications, performance management, data warehousing, CRM, EIS, decision support, query & reporting, enterprise reporting and even databases, so it is misleading to try to aggregate the figures of several of these indistinct sectors to calculate a total ‘business intelligence’ market size. 数据挖掘研究院

Note that a significant proportion — around a third — of Microsoft SQL Server revenues are now attributed to OLAP, so this and the smaller amounts attributed to ETL should be deducted from the total SQL Server revenues when calculating Microsoft’s share of the RDBMS market. Neither Microsoft, nor the other analysts who calculate database market shares, usually take this into account, so they include Microsoft’s large and growing OLAP revenues in its RDBMS market share, without being as generous to Oracle, IBM and CA. This exaggerates the size of the RDBMS market, and Microsoft’s share of that market, and is an example of why it is so important to analyze and adjust the figures before calculating market sizes and shares.

Market size

We estimate that the total worldwide OLAP market, including implementation services, was about $1 billion in 1996, $1.4 billion in 1997, just over $2bn in 1998, $2.5bn in 1999, just over $3bn in 2000, $3.3bn in 2001, $3.5bn in 2002, $3.7bn in 2003, $4.3bn in 2004, $4.9bn in 2005 and $5.7bn in 2006. The 2006 market size was higher than we had predicted, helped by the fall in the US dollar against European currencies, which enhances European revenues when expressed in dollars.

We estimate that the 2006 growth rate rose slightly to 16.4 percent, which is significantly higher than the 12 percent we had predicted. Most vendors reported good growth rates towards the end of 2006, but Microsoft did particularly well, with the long-delayed SQL Server 2005 available for the whole year, and selling well. But at least some of the improved growth rate in 2006 was due to the weakening US dollar; even companies with flat revenues, expressed in local currencies, appear to be growing when expressed in dollars.

数据挖掘研究院

The return to double-digit growth in 2004 was very welcome, but growths of over 20 percent are unlikely to return; even the 16.4 percent seen in 2006 is probably above the long-term trend, and could easily fall sharply when the US dollar strengthens again. This lower trend is because of both long- and short-term factors:

数据挖掘实验室

• It is hard for multi-billion-dollar markets to grow as rapidly as they did when much smaller. 数据挖掘研究院

• There is now a degree of market saturation.

  数据挖掘研究院

• Average prices have reduced sharply now that Microsoft is the largest vendor (though volumes are growing more strongly than ever). Many Microsoft OLAP sales are being made at the expense of pricier alternatives, and there is evidence of a slowdown in several companies’ growth rates from 1999. Microsoft has sold hundreds of thousands of OLAP seats in each year from 1999, many at the expense of far costlier alternatives. Many companies who buy the low cost Microsoft OLAP server also expect to pay less for OLAP clients, which further drives down average prices. Microsoft is expected to ship the low-priced PerformancePoint in the third or fourth quarter of 2007, which will put pressure on the expensive CPM products from other vendors. This means that as Microsoft’s share increases, the volume growth may remain strong, but the revenue growth will be depressed unless the market reaches the point where Microsoft has much more than a 50 percent share, which is not likely in the near future. 数据挖掘研究院

• Microsoft Analysis Services has put real pressure on other OLAP servers, whereas the leading client tool vendors continue to thrive (especially those specializing in Analysis Services). Indeed, most of the many OLAP servers that were on sale when Microsoft entered the market in 1998 have now died or disappeared into obscurity (eg, Acuity, Acumate, Gentia, Holos, Information Advantage, Media, MetaCube, Mineshare, Pilot, Prodea Beacon, WhiteLight). The disappearance of these products means that the overall market grows more slowly than the leading vendors.

  数据挖掘实验室

• Open source OLAP tools are now available, and these join Microsoft in putting pricing pressure on the more expensive products. 数据挖掘研究院

The amount of consulting required to implement the new generation products and pre-built applications is also likely to be smaller per site (but more in total through the rise in volume). In fact, there is evidence that Microsoft OLAP sites use much less external consulting on average than users of most other OLAP servers do, which is also reducing the market growth rate. A boom in OLAP application sales has been predicted as the technology becomes more widely deployed. However, this long-predicted boom in analytical applications has been very slow to materialize; CPM vendors like Cartesis and OutlookSoft are generally growing well, but analytical applications sales from more generic vendors like Business Objects and Cognos have remained very disappointing. 数据挖掘研究院

数据挖掘实验室

Total worldwide OLAP market shares

We rank vendors, not products, because many of the vendors supply multiple OLAP products and applications. As described in the background notes above, calculating OLAP market shares is quite an involved process, and these figures include OLAP server and client software, applications and OLAP consulting, whether performed by the software author or third-parties. Our resulting estimates for the top ten OLAP vendors are shown below. Improvements in market share or position are shown in green and declines in red.

Note that, despite occasional vendor claims of a 20 percent market share for Essbase, the Hyperion figures include all of Hyperion’s many products, and not just Essbase. Only a portion of Hyperion System 9 is treated as OLAP, however, with the former Brio product being largely excluded. The Essbase market share alone is less than 50 percent of the total for Hyperion. As such, the Essbase market share in 2000 was comparable to Cognos PowerPlay and Microsoft Analysis Services, but Analysis Services clearly overtook both Essbase and PowerPlay in 2001 to become the leading OLAP server. In 2002, Microsoft moved into the lead overall, even including the then Brio’s share with Hyperion’s. The gap widened further in each subsequent year. 数据挖掘研究院

As the Oracle acquisition of Hyperion did not occur until 2007, it does not change the 2006 or earlier figures, but from 2007, Hyperion’s share will be merged with Oracle’s (which was previously rather low). The joint share will show a declining trend, as both companies separately had declining OLAP market shares, though in absolute terms, it will push Oracle into a stronger second rank behind Microsoft, from its previous weak tenth position. 数据挖掘研究院

Our figures are sometimes questioned by consultants who encounter a quite different product mix in their target areas. The reason, of course, is that these are estimated worldwide shares — the shares in particular regions, industries or organization size vary significantly.

For example, Microsoft Analysis Services is typically chosen by smaller organizations, while SAP BW and MicroStrategy are much more likely to be found in the largest organizations. Similarly, Business Objects and SAP are relatively stronger in Europe, while the MicroStrategy and Hyperion customer bases have a North American bias (this is particularly true of the former Brio customer base). The large non-specialist vendors, such as Microsoft and Oracle, are stronger in the rest of the world than the smaller BI specialists, who tend to be under-represented outside the major markets. By vertical market, MicroStrategy is particularly strong in retail, Applix in finance and insurance and Microsoft and Business Objects in the IT industry.

数据挖掘研究院

When you combine these dimensions, distinct clusters emerge, such as large North American retailers where MicroStrategy has a disproportionately large presence, while smaller organizations outside the major markets are much more likely to use Microsoft. This means, of course, that individual anecdotal evidence will rarely be consistent with the world market.

Note that the Business Objects, Cognos and Hyperion figures include historic Crystal, Adaytum and Brio data, so the market shares for each of those vendors in 1999–2002 are higher than we originally published. The Infor figures include the former Geac/Comshare and Systems Union/MIS figures.

These are provisional figures for 2006, which are subject to review. 数据挖掘研究院

Commentary

OLAP Report customers click here for a full commentary on the top ten vendors’ market share changes in 2006 and the expectations for 2007. 数据挖掘研究院