« May 2016 | Main | January 2017 »

3 posts from November 2016

November 16, 2016

What features do your search users really want?

What features and capabilities do corporate end-users need from their search platform? Here's a radical concept: ask stakeholders what they want- and what they need - and making a list. No surprise: you'll have too much to do.

Try this: meet with stakeholders from each functional area of the organization. During each interview, ask people to tell you what internet search sites they use for personal browsing, and what capabilities of those sites they like best. As they name the desired features, write them on a white board.

Repeat this with representatives from every department, whether marketing, IT, support, documentation, sales, finance, shipping or others - really every group that will use the platform for a substantial part of their days. 

Once you have the list, ask for a little more help. Tell your users they each have $100 "Dev Dollars" to invest in new features, and ask them to spend whatever portion they want to pay for each feature - but all they have is $100 DD.

Now the dynamics get interesting. The really important features get the big bucks; the outliers get a pittance -  if anything. Typically, the top two or three features requested get between 40DD and 50DD; and that quickly trails off. 

I know - it sounds odd. These Dev Dollars have no true value - but people give a great deal of thought to assigning relative value to a list of capabilities - and it gives you a feature list with real priorities.

How do you discover what users really want? 

 

 

November 14, 2016

Which search is best?

Ask that question to a number of knowledgeable enterprise search consultants, and you’ll no doubt hear a number of answers including Attivio, Elasticsearch, Google, Lucene, Lucidworks, SharePoint, Solr and many others. All are well known, and include rich capabilities and strong technology underpinnings. And the experts you spoke with will have answered honestly.

What you would experience is not unlike the parable about six blind men describing an elephant. In the John Godfrey Saxe telling of the tale, he says:

And so these men of Hindustan
Disputed loud and long,
Each in his own opinion
Exceeding stiff and strong
Though each was partly in the right
And all were in the wrong...

So, now you may be wondering which search engine really is the right one for you.

The answer is really very easy: the one that meets your needs in your environment. But knowing that may be hard, because there is a good chance it’s been a while you really looked at your current environment.

I’d suggest you break down the process into a few distinct tasks in a process we call a Search Audit.

The Audit

A search audit is very similar to the process we recommend our customers use when selecting a new search platform: you go through the same process and look at the same issues of environment, requirements and more.

Not unlike a financial audit, a search audit is an objective examination and evaluation of your enterprise search implementation. The objective is to review the important metrics to determine how well your search is performing; to identify potential weaknesses; and to come out of the audit with a plan to fix any issues found.

At a high level, the audit is a review of the current environment; repositories; access security; and user requirements. Let’s look at what each of these includes.

Operating System

The operating system you use in an organization is often determined from ‘on high’. When considering a new search platform, it’s critical to verify support for operating systems you use; in an audit, you also need to confirm that your search platform is supported on the operating system version as well as on any anticipated updates.

Systems

Whether you use physical or virtual systems is a big item to review, as well as whether your search platform is software or an appliance. It also may matter whether your servers are on-prem, in the cloud, or a hybrid of both. For example, your security review may need more attention if you use a cloud or hybrid solution; and performance should be reviewed for virtual and remote servers.

Development Tools

Very few search platforms include every feature or capability you need. The solution may be as easy as scripting some common functions to customizing or modifying front-ends.

In the audit, pay attention to the platform and scripting languages, and make sure you have those skills in-house.

Repositories

In your audit, you clearly are looking at indexing content; but it never hurts to review the repositories where your critical content lives. Confirm that any version updates have not impacted the search platform or its performance. And verify that any anticipated platform changes are supported by your search solution, and plan accordingly.

Security

As with your repositories, an audit should confirm that any changes in the security infrastructure are mapped into, and supported by, the search platform. Are there new security levels or groups? Do queries against the repositories include content with the new security mappings?

Content

Servers occasionally get major updates. Use the search audit as an opportunity to anticipate upcoming operating system changes in order to properly confirm compatibility with your existing system. A while back I spoke with a large company using Verity K2 – which has been obsolete for years. They were about to update their Windows NT servers to Server 2012 and wanted to know how they could port K2. Good thing they asked.

Users

Your search platform exists to serve your customers, whether they are internal or external. Google and various eCommerce sites on the web have defined what users expect from search. Most enterprise search software ‘out of the box’ doesn’t look, feel, or work like Google; and you’ll have a problem if you don’t solve the expectations. Ironically, even the popular Google Search Appliance doesn’t generally work like Google.

If you do not already have one, create a search center of excellence, and recruit representative users to help define how your search works.

When it boils down to it just about any search engine can work ‘like Google', but that takes time and effort. If you haven’t already done so, use the audit as the driving force to improve the search experience.

Next Steps

Once you’ve completed your audit, you may find no major problems; and decide that your current search is doing pretty well. If that’s the case, you are in good shape. Other than ongoing maintenance, your task is complete for another year or two.

More often than not, issues come up in search audits. Sometimes it involves content not being indexed or poor search result quality. It may also be that the user experience is not “just like Google”.

The good news is that a majority of these issues can usually be fixed without replacing the platform.

What is the bad news? More often than not, people are so frustrated with search that a decree has come from on high calling to replace the search platform. This usually results in great effort, significant disruption and expense, and a new platform rollout with great flare and unrealistic expectations. But unless the issues you discover in an audit are addressed, there’s a good chance that you’ll be replacing the ‘new’ platform within a few years anyway.

November 01, 2016

One search to rule them all

(Originally published on LinkedIn)

Lucene was ‘born’ in 1999, created by Doug Cutting; and in 2005, it became a top-level Apache project. That year, Gartner Group announced that the search ‘Leaders’ platforms on their Enterprise Search Magic Quadrant included Autonomy, FAST, Endeca, IBM Omnifind, and Verity. The Google Search Appliance was right on the cusp between ‘Challengers’ and ‘Leaders’. Not many people knew about Lucene; and few who did saw it as much more than a quirky little project.

Just a year later, Yonik Seeley and his employer, CNET Networks, published and donated the Solr search server to the Apache Software Foundation, where it became an incubator project in 2006; the two projects soon merged into a single top-level Apache project. That same year, Gartner narrowed the ‘Leaders’ in their 2006 Magic Quadrant for Search to Autonomy (which acquired Verity the previous year), FAST, and Endeca.

Jump forward to the present. FAST is gone, acquired by Microsoft in 2008 and morphed into SharePoint Search. Hewlett-Packard acquired Autonomy in October of 2011, followed a few weeks later by Oracle’s acquisition of Endeca. Endeca is no longer available as a search platform; and Autonomy is mostly seen as a strategy to keep a large number of HP consultants fully employed, often on compliance applications.

Only a spattering of commercial enterprise search platforms that once flooded the market just a few years back exist any more. While Gartner continues to list 14 or 15 products in their Magic Quadrant Enterprise Search grid, about the only pure commercial products we see any more are the Google Search Appliance and Recommind. And Google recently announced that the appliance is scheduled to go ‘end of life’ over the next few years. All of those bright yellow boxes become really nice Dell servers by the end of 2018.

A new crop of search platforms has grown to fill the void.

As an open source product, Solr has grown in its capabilities, and is now widely used for enterprise search and data applications in corporations and government projects. Solr Cloud extends the platform to a scalable high-availability platform for demanding enterprise and data search applications. Solr is an open source solution.

Cloudera also bundles some interesting extra tools including Solr in their HUE bundle; free to download and free to use as long as you like. Cloudera runs a slightly older but stable release, 4.10; but with a committers Yonik Seeley and Mark Miller, I suspect they’re in a good position.

Hortonworks, a Cloudera competitor, also offers Solr/Solr Cloud in their releases, in partnership with Lucidworks - a company with a large number of committers on staff.

There are also three companies that have proprietary offerings based on open source technology.

Attivio, founded in 2007, is a “Leader” in the most recent Gartner Magic Quadrant for Enterprise Search. Their product, while not open source, nonetheless thrives by combining search, BI, data automation, analytics and more.

Elasticsearch has evolved into a strong platform for search and data analytics, and a number of organizations are finding it useful in some tradition enterprise search applications as well. Elastic has also integrated Kibana, a powerful graphical presentation tool that adds value for content analytics, not just search activity reporting.

Lucidworks Fusion is a relative newcomer to enterprise search. It includes many of the rich architectural features that enterprises expect, including a powerful crawler, connectors, and reporting. With its ‘Anda’ crawler and connectors, admin UI, and reporting, some people see it as a contender to replace the Google Search Appliance.

The one thing that all of these ‘proprietary’ products have in common? They are based on Apache Lucene to deliver critical functionality. And when you consider all of the web sites that use some form of Lucene for their site search, I think you'd agree that it really is a powerful little package. It’s available for virtually any operating systems, and can be integrated using just about any programming language from C/C++ to Java to Perl to Python to .NET.

Even more amazing is that these companies with commercial products based on Lucene – and who compete in the marketplace - actually cooperate when it comes time to fix bugs or add new capabilities to Lucene. Given all of the commercial players that have closed their doors - leaving their customers to find replacement platforms – we’ve reached the point where open-source-based software really is the safe choice now. And universally, Lucene is the common element.

The quirky little search API Doug Cutting put together in 1999 has evolved to be the platform that drives the leading search platforms used in big data, NoSQL, enterprise search, and search analytics. And it doesn’t seem like it’s going to be phasing out any time soon.