Google blogsearch excludes blog entries made on Google's new blog tool

I like the new beta version of google's blogging tool. I was one of those who couldn't switch their blogs on the older product to the new one. So, I just started a new blogsite since I didn't have that many older posting.

But -- to my dismay, none of my new postings on the beta blogger were showing up when you do a blogsearch. So, I decided to do a quick test. I made the same postings from my new beta blogger account to the old bogger account and did a search immediately. And lo and behold, the entries from the old blogger turned up instantly, but not the new blogger.Try this for yourself:

1. Go to blogsearch.google.com
2. Search for "tie global poverty" -- entry of my last blog post
3. Refine search to within last hour (or day, depending upon).
4. You'll see that while my recent post on my old account shows up, my original post on the new blog doesn't.

Why is Google not indexing the beta blogger entries?

Tweet

TiE's session on The Pivotal Role of Entrepreneurship in Addressing Global Poverty

Part of my job is to leverage every networking opportunity to get the word out on the company. With that intent, and also to learn a bit more about social enterpeneurship, this afternoon I drove down to Santa Clara for TiE (The Indus Enterpreneurs) member networking session titled "The Pivotal Role of Entrepreneurship in Addressing Global Poverty".

If you are an enterpreneur living close to a TiE chapter, and haven't been to their events, it's definitely worth checking out (they always have great Indian food at each gathering). Not surprisingly, the crowd at the Silicon Valley chapter is a good mix of mostly Indian enterpreneurs and seasoned industry execs, so it's pretty decent networking. Today was a bit different, because instead of promoting our individual companies, the crowd was more interested in how enterpreneurs can play an active role in tackling global poverty.

Dr. Bill Musgrave from The Enterpreneurs Network(TEN)-Silicon Valley gave the keynote, with many anecdotes on enterpreneurs finding viable business models serving the "bottom of the pyramid". Particuarly striking was his quote of former president Clinton: "I have never seen uneven distribution of intelligence, but always seen uneven distributions of opportunity". Dr. Musgrave argued that to make the grand changes needed to address global poverty, governments and large corporations haven't proven to be much effective, but it is rather the enterpreneurs who are most effective at empowering the poor. His call of action to the roomfull of enterpreneurs was to seek out such opportunities and make a difference.

The session also gave 2-minute "open mike" pitch slots to any interested attendee to promote their company/cause. I was particualrly impressed with Vipin S (?) from Intel who's starting a program in India to enable local farmers to grow crops for biodiesel fuel. Also interesting was Navaneethan Sundaramoorthy's pitch on Association for India's Development's local chapter , which has a collection of development projects that harness the collective resources of local Indian diaspora.

I couldn't pass this "open mike" opporutnity to talk about my friend Mahabir Pun's work in Nepal. Mahabir and I went to the same college in Nebraska. Upon graduation, while I took the more common path of getting a job and later launching start-ups, Mahabir went back to Nepal, back to the same rural village where he grew up, and started a grade school for the local kids deprived of education, and also started several income-generating micro-projects. I should do a post describing his work in detail, but check out Himanchal Education Foundation's web site in the meanwhile, which is the foundation that tries to get financial help and recruit volunteers for the school. It is a very inspiring story of sheer will that is making a difference without much help from government and NGO's who are now finally warming up to this project.

Afterwards, I got a chance to chat with several other folks who are somehow involved part-time with similar social projects, big and small -- and as I write this, I feel a lot more hopeful about these enterpreneurs making a lasting social impact. I sensed a similar drive to make social changes as we have in launching new companies and seeking successful exits. Key is, how effective will we be in pooling our resources and provide helping hands (I should say, empowering hands) to organizations such as Mahabir's himanchal.org. Just like with any new venture, we don't have the answers when we start, but with the right will, we will find a way.

Tweet

Qualitative Data vs Behavioral Data: Who do You Pay Attention To?

Yesterday during a call with a potential client, this topic came up. This is a well-known desktop software company, and they have a unique challenge: Their primary measure of customer loyalty is the "Reichheld Score" aka the Net Promoter Score, which is based on customer responses to a single question -- "Would you recommend us to a friend?". Now, the interesting thing is that while this company is doing rather well in the market, their overall net promoter score is not that great. In fact, their score is lagging behind other comparable software manufacturers.

So, the obvious question is -- why doesn't their net promoter score correlate with company growth? Which metric should they rather measure as a driver of company growth?
As we talked, I also found out that this company hasn't done much in terms of evaluating the behavioral data on their customers -- you know, stuff like actual purchase transactions, new purchases versus repeat purchases, customer complaints, returns, etc. And I couldn't help but think that perhaps this is where they should start look first. A great deal of research work has shown us that past behavior is the best predicator of future behavior, and this is true when it comes to measuring customer loyalty as well. A 2002 HBR article "mismanagement of customer loyalty" describes this as:

Simply put: Not all loyal customers are profitable, and not all profitable customers are loyal. Traditional tools for segmenting customers do a poor job of identifying that latter group, causing companies to chase expensively after initially profitable customers who hold little promise of future profits. The authors suggest an alternative approach, based on well-established "event-history modeling" techniques, that more accurately predicts future buying probabilities. Armed with such a tool, marketers can correctly identify which customers belong in which category and market accordingly.

So why isn't this company looking at behavioral data on its coustomers? I didn't get a clear answer, but could it be that it is easier to conduct surveys rather than dig deep into data, specially when the data volumes are huge and the data is scattered around different corporate silos? Could this company be viewing the analysis of behavioral data as a long, complex exercise that involves getting down and dirty with data warehouses and analytical modeles, when all they needed was a nice simple metric that would measure customer loyalty and nicely correlate with company growth?

Yes, I am being a bit facetious -- but while I don't dispute the value of qualitative research, I think in this case, they are more applicable AFTER an initial study of behavioral data. This company needs to understand the "WHAT" first (i.e. what is going on with my customers? which customer attributes/behavior are best indicators of company growth?), and then they can apply qualitative reserach to understnd the "WHY", i.e. why things are happening the way they are.

Tweet

Would you consider a SaaS/on-demand solution for marketing analytics?

By normal conventions for my job, I'm supposed to answer a "hell, yes!" to this question because I work for a company that does on-demand marketing analytics. But all hype aside, let's explore this and see where it makes sense.

We all know, specially the direct marketers, that data can be leveraged to get a better understanding of customers/prospects which results to more targeted and effective marketing programs. This concept has been around for decades and generally well accepted. For a while, data used to be the challenge where an organization wouldn't have much data on its customers -- but in the age of web, RFID, CRM explosion, etc. -- people have way too much data to know what to do with it. Still then, why are we being bombarded with spam and other irrelevant marketing messages?

A lot of marketers, specially in the SMB sector would answer -- well, it's hard to do this. Analytical solutions to leverage your customer data are tough to implement. It's either some expensive software/hardware (think BI/CRM/Analytics solution providers), or some elaborate marketing service provider (agencies, mailhouses, etc.) with heavy-duty hourly rates, or else you left with your own devices to put together an analytics team who has:

1. Technology know-how to deal with large datasets and apply heavy-duty analytics
2. Business know-how to understand vital issues challenging your business and marketing
3. Strategic approach to analysis to avoid not seeing the forest for the trees

Sounds familiar?

What direct marketers, and a lot of other data-driven decision makers, seem to be lacking is an agile yet reliable set of tools that help them "see through" their data without having to own expensive hardware/software infrastructures, or having to pay for service provider hours throught the nose. And I think, a Sofware-as-a-Service (SaaS) or on-demand approach can be a very valid alternative approach to get there.

Because let's say you had an option to go with a SaaS solution provider for analytics, someone who had an easy way for you to upload all your marketing data securely and reliabley, and then let you define what exact answers you were looking for, and based on that apply the relevant analytical models and provide you the answers in a form you can actually understand. I am not talking about showing a bunch of fancy charts, graphs, and (yes) the dreaded dashboards -- but actually providing you with deliverables you can use.

Things like a downloadable list of all customers who are likely to bail out on you in the next 30 days. Plus the optimal set of retention tactics derived via a set of data mining models that analyze all past retention programs. Or, it could be a list of customers who are most likely to purchase a certain product -- or conversely, list of customers who will be really upset if you try to direct market them.

Now, I know these deliverables may sound simple, or "gee, everyone can do that" -- but at most of the companies I have worked with, I still don't see this. It baffles me as well, but most companies aren't even at this level of understanding.

Yes, I am biased because my job is to build and promote SaaS or on-demand analytics; but I also know that market will only buy a solution if it makes sense, only if it is truly a better alternative to other solutions. I don't know that. I have confidence in our approach, and with the different customers who we have been able serve via this model, but honestly, that's not enough data points. And as a solutions provider, I sure would like to make sure we are building/providing things to our customers that add a great deal of value.

So, what do you think? Do you feel SaaS/on-demand is a viable model for analytics? If you were in a situation to choose, would you go with an on-demand analytics solution (provided analytics is not the core competency of your businees)?

I think it's surely worth a shot to consider.

Tweet

Why am I doing this @#$%?

In the summer of 1990, the same day Saddam invaded Kuwait, I boarded a Northwest flight from my native Nepal and 20-odd hours later landed amongst the beautiful cornfields of Nebraska (state tag line is "the good life") to start by bachelor's degree in computer science, which later extended to a master's. After finishing my master's and a first taste of Internet startup, somehow I ended up in San Francsico, where I live today -- went through another string of startups, the latest having something to do with deriving "actionable insights" from customer databases, in other words -- business intelligence.

Why should you care? Well, I don't know, but..

I have come to believe that blogs are perhaps the best way to share your experience, both professionally and personally, with the widest possible audience who as a group are always smarter than you.

So that's why I'm here -- to share and explore conversations that haven't happened yet, but I know they will take me (and perhaps you as well) into fascinating aspects of "business intelligence" (and its repercussions on a practitioner) -- or at least get us all a good laugh.

So, come back, come often.. get my RSS feed, trackback, whatever works for you. But let me know if you find these postings useful, boring, or pointless, or what would you rather talk about.

Overall -- let's have some fun chats.

Tweet

Our baby OpenI is a Finalist on SourceForge: Please Vote!

One of my many professional roles is that of the project lead for OpenI, which is an open source BI web application that enables interactive analysis and reporting from OLAP, RDBMS, and data mining data sources. I am very proud to annouce that OpenI is now a finalist for the 2006 SourceForge Community Choice award in the Enterprise category. This is special because this puts OpenI is now in a select group of top 12 projects out of the 116,000-plus projects that exist today on SourceForge.

Now the time has come to cast the final deciding votes. Please go to the link below, select Enterprise category, and check OpenI -- it's that easy

http://sourceforge.net/awards/cca/

March 23rd is the last day of voting, so please vote now. And when you're done, please pass the word around -- we can use all the support/publicity :-)

A bit more on OpenI --at our company Loyalty Matrix, we developed OpenI out of necessity. Our main business is doing customer analytics and delivering actionable marketing insights. Microsoft SQL Server, with Analysis Services for OLAP, is our platform of choice. We needed a thin client web front end for client access. Nothing special you say.

But, the .MAC group at Apple is one of our clients. Guess what, Apple really doesn’t like to use Internet Explorer!

After spending significant time and money on comercial components, that claimed to be browser independent, we punted and decided to search out open source tools. Starting only a year ago last July, we are now about to release version 1.2 of OpenI which includes support for both SQL Server 2000 and 2005. Loyalty Matrix has been using OpenI in production for over six months to serve a number of clients including 24 Hour Fitness and Olivia in addition to Apple.

OpenI is a web-based application for business intelligence reporting, that enables users to analyze data and publish results over the web. Today, OpenI is consistently in the top 100 most active open source projects, with a growing and thriving community. More on OpenI at http://openi.sourceforge.net

Thanks for your support.

Tweet

When it comes to prediction, Machine Learning kicks CHAID and CART's Tails

Conventional wisdom for direct marketing analytics regards CHAID and CART as cutting edge analysis for discovering correlations. (see "Optimal Database Marketing" by Drozdenko/Drake, Chapter 8, 2002)

However, what's not commonly known is that these are more than 20-year old techniques, and recent academic work has shown that analyses like CHAID and CART are more applicable for hypotheses testing rather than predictive analysis. And it's the latter that's more key to direct marketers. So, what's the cutting edge technology for predictive analytics, you say? Machine Learning!

In our own experience for conducting predictive analysis for direct marketing optimization, we have found that machine learning techniques such as Random Forests(tm) are far superior to the single-tree based analyses like CHAID and CART. Machine Learning has far better accuracy for prediction, mainly because instead of trying to come up with a single best tree algorithm for prediction, it develops many different simple trees (many trees = forest!) and collectively combines them into a "forest model".

Tweet

The I in BI is about Intelligence, not Infrastructure

BI software products, open source as well as the commercial ones, have mainly been focused on the developers/SI's building BI applications, as opposed to focusing on the end-user of the BI applications. The end-user's problem is pretty straightforward: "I am sitting on a mountain of data accumulated over various data silos in my organization, and I need help making sense out of it; I need a tool that helps me derive actionable insights". The end-user doesn't want to write code or SQL/MDX queries, they don't want to mess with complex regression models or machine-learning data mining models. They want a tool that has some notion of the *business* questions they ask, and can help analyze their existing data in the context of those questions -- hence the term *business* inelligence.

The "I" in BI should be about intelligence, not infrastructure. Most of current BI technology talk solely focuses on the "infrastructure" components - database, OLAP, workflow, data mining, reporting, etc. but just by having those components in one stack, you don't become intelligent. You still need data models optimized for your specific types of analyses, you still need to find the appropriate data mining and stat models for your problem space, you still need the right visualizations both to interactive analyze your data, as well as to publish your analyses in such a way that mere mortals can understand and use those insights to make business decisions. *This* is the real "I" in BI -- Intelligence, not Infrastructre!

Open source BI platforms are going to be key in highlighting this differentiation because open source by its very nature commoditizes the infrastructure of BI -- pushinig the proverbial "value up the stack" to components that actually produce intelligence.

At my work, we bet on this very phenomena where we utilize an open source BI infrastructure, and build domain-specific BI applications on top of it. Since we started doing this a few years ago, we didn't have the advantage to choose between Pentaho, JasperSoft, or BIRT -- so we did our best by building our own platform by integrating existing open source BI components, of which some like Mondrian and JPivot are also being utilized by Pentaho, et al. For us, this was a pure R&D initiative to leverage an open source development model, and in the spirit of collabortion, we even published it into sourceforge as OpenI (http://openi.sourceforge.net). We hope that as Pentaho, JasperSoft, etc. start releasing complete version of their platforms, we should find a way to collaborate with their stack. But for what it's worth -- you can download and use OpenI today as a BI application, just like we use it in production to serve our clients. In fact, I am keenly interested in community feedback on OpenI to understand the best ways to collaborate with the other open source BI projects.

The key is that open source BI needs to first define itself as a respectable, reliable alternative to the commercial BI vendors. Then we can bicker about what's free, and what's commercial. The infrastructure should be completely open, on top of which you can have commercial, domain-specific BI applications. But first, Pentaho, JasperSoft, BIRT, and OpenI -- all of us need to collaborate and put an open source BI stack out there that can hold its own in comparison to commercial BI vendors. Open source doesn't work like enterprise software where one company/organization needs to own the entire space. And even if they wanted to, they can't, because successful open source is all about building a community that spreads far beyond the confines of your organization. We are fortunate to have several noteworthy projects and components in the open source BI space -- it is time to these projects to collaborate and present open source BI as a unified front, and start growing the community. That's how open source BI is going to become real.

Tweet