SAP: Long and Winding Road for Search

January 5, 2012

In one of the early editions of the Enterprise Search Report, that white elephant of 600 pages containing profiles of more than two dozen vendors, I described TREX, a nifty algorithm for Text Retrieval and Information Extraction. (The link is to the Wikipedia write up, however.) For those of you who are new to search, TREX is not the creature you wished you had as a pet when you were eight years old. The SAP TREX is a natural language processing search and retrieval system which was mostly home grown. Keep in mind that TREX owns the Inxight entity extraction and server technology developed by the adepts at Xerox PARC. I interviewed one of the developers, profiled the system’s approach to content processing, and pointed out that search was a killer in the SAP R/3 environment for three reasons:

  1. SAP assigns its own spiffy metadata to content objects, storing these in the wild and wonder proprietary R/3 environment
  2. SAP systems took and probably still take a long time to plan, implement, and impose on the client. My understanding is that the client does not tell SAP how the clients like to work. SAP tells the client how the client will work with the SAP system and method. Nifty for sure.
  3. SAP systems have struggled with a wide range of performance “opportunities.” The idea is that when something goes slowly, then the client has the “opportunity” to make changes which will speed up the large, IBM-inspired system.

A few years ago, before Endeca became the new billion dollar toy at Oracle, Endeca accepted cash infusions from outfits hooked up with Intel (yep, the company with the vision that its chips could crush any computational problem because they were so darned fast) and SAP’s investment unit (an outfit allegedly looking at ways to give SAP a leg up on the future). After watching Endeca do its recursive indexing and faceting processes, Intel and SAP shifted gears. Endeca, as you know, is now part of Oracle along with TripleHop (clustering and indexing), InQuira (natural language processing from two predecessor companies), and RightNow (also infused with search technology), Artificial Linguistics, PL/SQL’s wonky command driven search, and probably some technologies I either don’t know about or have forgotten due to advancing senility.

Will SAP slip and fall with its information retrieval solutions? A happy quack to the image source

When you want to run search within an SAP environment, many folks just embrace one of the SharePoint solutions, give TREX a go, or license a system which is compatible with some of the SAP processed content. In short, SAP’s approach to search is not much different from IBM’s or Microsoft’s.

The question to consider is, “What’s next for SAP?”

Several observations:

First, SAP has to pump money into TREX to keep the system in step with today’s information demands. With SAP dabbling in open source and focusing on higher margin products and services, TREX is probably not the long haul solution for SAP. Home grown search is too expensive.

Second, SAP continues to poke around open source software. At some point, SAP may follow in the footsteps of the company which inspired SAP in the first place—IBM. Lucene and Solr look like possible options. This is a trend to watch.

Third, SAP buys or ties up with one of the workman-like search vendors. SAP could either sign a deal to use a third party system on some basis or just buy one of the dozens of information retrieval vendors who are looking for a financial white knight. Despite the chatter about search, many search and retrieval companies are gasping for oxygen. SAP may have a tank and a breathing mask.

What’s my view? Well, since I am a mercenary goose, I don’t have an official opinion. I do find it fascinating that SAP has not moved aggressively to the Lucene Solr solution. So for now, I am going out of town and will wait until my Overflight service provides some solid data about SAP’s next move.

Hopefully it will be more artfully crafted than SAP’s pricing and customer service activities in the last two or three years.

Stephen E Arnold,

January 5, 2012

Sponsored by


Comments are closed.

  • Archives

  • Recent Posts

  • Meta