Kellblog

My first reaction to the Boston Globe article, Endeca Founders Steering Search Firm Towards Business Intelligence, was “here they go again, yet another strategy.”

I’ve written before about Endeca’s strategic issues, which I’ve found more than a bit ironic for a company with no shortage of Harvard MBAs and 1980s strategy guru Michael Porter on its board, heading the strategy committee.

While brains are most certainly not lacking, perhaps common sense is. I think Endeca has been making a common mistake, one which I call getting bored with your market. Or, as I once quipped, “getting board” as this is often a top-down phenomenon.

A few of my favorite examples of this grass-in-greener syndrome include:

• Informatica, which decided that ETL was boring, and launched itself into the seemingly emerging but never-to-emerge analytic applications category, only to cede the ETL market to Ascential who was only too happy to serve it. This story ends happily with Informatica re-discovering its DI roots after a few years of turmoil.

• Verity, which decided that enterprise search was a dead market, repositioned itself as a intellectual capital management vendor, and thus ceded the enterprise search market to Fast Search & Transfer who was only too happy to serve it. This story ended less happily with Verity selling itself to Autonomy, a company then half its size.

• Ingres, which made the epic-fail decision that “relational databases were commodities” in the late 1980s, effectively declaring the RDBMS party over, focusing its energies and application development tools. At the time, the RDBMS market was ~$500M. Today it’s $15B oligopoly, driving 50% operating margins. Application development tools have remained a economic bloodbath for the past 20 years.

I understand why companies fall into this trap. There are several reasons:

• They imagine the market as a standing army and not as a parade. In reality, the market marches past the company as it develops over time. (Think of it as a parade with a changing width.) So what might be 7 years old to the company is brand new to folks in the market walking by for the first time.

• They are under constant pressure to change. The board, bankers, financial analysts, industry analysts all constantly ask management “what’s next?”, “what’s the roadmap?” and “what’s your vision?” For some inexplicable reason, most management teams lack the courage to answer: more of the same; we think we’ve penetrated about 10% of the total market and we intend to keep on keeping on, growing both domestic and internationally, becoming the worldwide dominant leader in what we do. Instead, they talk about new markets, new products, obsolescing and de-positioning their current offerings in the process.

• They confuse stack layer with market structure. Most people who sell platform technologies are frustrated because they can’t sell as a direct solution to a hot business problem. Imagine solution selling a business executive on electricity. You can’t. There are too many use-cases. They assume that profit corresponds with directness to business value when in reality, profit is determined by industry structure and its profit zones. Relational databases are one heck of a profitable business, but they do not map to solving one “silver bullet” business problem.

• They are afraid. My favorite example was Arbor Software, makers of Essbase, who did a panicked and subsequently disastrous merger with Hyperion Software in response to Microsoft’s acquisition of Panorama software, which provided the foundation for Microsoft Analysis Services. Yes, Microsoft’s market entry was a problem, but they was plenty of room for both high-end and low-end offerings. I always viewed the combination as the emergency evacuation of a beach in San Francisco in response to a shark sighting in Los Angeles.

Endeca’s core market is e-commerce. The product concept came from making it easier to find products on e-Bay. So you need to combine text search, search concepts like thesaurus (e.g., match “jumper” to “sweater”) and spelling correction (e.g., did you mean “paisley”) with structured data like price, size, color, and quantity. It’s also why you need the notion of faceted navigation, which I view as the search equivalent of OLAP.

Example: typing Cabernet into a wine store’s search engine should allow you to iteratively refine your query along several dimensions, such as price (<$10, $10-$25, $25+), bottle size (375 ml, 750 ml, 1500 ml), country (US, France, Australia), region (which is hierarchically dependent on country, meaning France–>Bordeaux, US–>California–>Napa), and Parker points (<80, 80-89, 90-95, 96-100).

This is what Endeca is good at. This is what they grew up doing. And this is what I think they’re bored with.

That boredom has resulted in a number of initiatives and reorganizations over the years. For an order-of-magnitude $100M company, look at everything they discuss on their website:

• Specialization in 5 industries
• Production of 5 products
• Expertise in about 20 different solution areas, total

If you take the ~$100M in revenue and divide it across the 5×5 industry-product combinations, you end up with about $4M at each intersection, or slicing across the solutions, about $5M in in each solution area. Either they’re talking about way more than they’re doing, or they’re spread pretty thin across all these areas.

The e-commerce roots were always there. They made a modest push in media/publishing. They also made a push into government/intelligence, but I think product scalability concerns hobbled the effort. The strategic push prior to BI was in manufacturing.

From the cheap seats, my advice for Endeca is simple.

• Either, return to the historical focus, a la Informatica, tripling-down on e-commerce with a goal towards owning that market, worldwide

• Or, bet all-in on a strategic evolution to BI.

They should do the first strategy is they are certain the market is big enough to reach their growth goals over the next few years. They should do the second if it’s not.

Frankly, my gut is that the first market is big enough, that they’re underestimating it, and that the best strategy would be a roots rediscovery. But there are also some good reasons to pursue the second strategy:

• BI is largely a front-end game, where DBMSs do the heavy lifting. I believe that Endeca has always been better at data presentation and user interface than building scalable processing engines, so it seems to fit with the company’s core competence.

• It’s been done before. Cognos, today a top BI vendor (and piece of IBM), did a nearly miraculous transition from application development tools (Powerhouse) to BI tools (Impromptu, Powerplay) over about a seven-year period in the 1990s. Endeca wouldn’t be the first market refugee to seek asylum in BI.

• Unification of structured and unstructured information is the next big thing in BI and, given its heritage, Endeca is well positioned to provide it. I’ve always believe that it’s easier to make systems designed for unstructured data work with structured data than the inverse.

Now, as someone who was a fairly key player in building one of the top BI companies (Business Objects), I’ve always been a bit snobbish in my response to the search vendors’ BI offerings. I always viewed them as feeble efforts by data neophytes, more driven by fear of Google in the search market than by expertise in the BI market.

But Endeca’s not your typical search vendor; they’ve paid attention to both content and data since day one. And they make pretty user interfaces, which are a critical success factor in BI. They’ve started to attack the data integration problem by licensing Informatica, so they are serious about this initiative; it’s not just marketing fluff.

All they need to do is build it atop the right DBMS platform for managing structured and unstructured information. I don’t think that’s the MDEX engine. Nor any bolting of MySQL to Lucene.

Hint, hint.