Autonomy - from Autocracy to Anonymity ?

For those of us who watch the industry closely, Autonomy founder Mike Lynch's imminent departure from the newly merged HP/Autonomy entity comes as no great surprise. However this lack of surprise isn't for the reasons that many people would assume: that an ambitious entrepreneur would inevitably "cash out" after a time period defined in the acquisition. I suspect at least part of the conflict relates to a clash of culture between HP and Autonomy.

Officially the reason for Lynch's departure and replacement by Bill Veghte -- HP's Chief Strategy Officer -- is that certain financial targets have not been met since the acquisition. On its face, this is a serious concern for a public company that has made an eleven-figure acquisition. Autonomy was indeed under-performing since it was absorbed. We've argued for several years it was under-performing long before the acquisition.

Although no one knows for sure, I suspect Lynch's top-down micro-management style -- where he kept a very close rein on all the activities undertaken by his employees -- lay at the root of the company's troubles. According to Autonomy customers and ex-staffers, everything from product development right through to metrics regarding the use of sales collateral at customer pitches was reported directly to Lynch. Autonomy customer demos that RSG analysts have witnessed over past few years have ranged from bemusing to disastrous. You can only say "meaning-based" XYZ so many times before people's eyes glaze over.

This places existing customers in a very difficult position regarding the future of technologies they purchased under Autonomy for which now pay support monies to HP to keep running. It will be revealing to see what elements of the large Autonomy potpourri that HP continues to maintain interest.

Excerpt from the IDOL Server Evaluation

Enterprise Search Evaluation Stream looks at... Training Autonomy IDOL

"With IDOL, the system has to be trained by iteratively exposing it to training sets. According to Autonomy, the set of exemplars needed is not large, and the system is trained out of the box. In practice however, you may find its efficacy to be disappointing if you rely on the automatic - once over, many times - too much. It is possible to over-train or under-train the system; getting it just right may take a few tries...."
(p. 92)

Get All Search Vendor Evaluations

Learn the real strengths and weaknesses of major Search vendors from around the world, in our Enterprise Search evaluation research stream.

Real Story Group. Make Better Technology Decisions.