1. This nugget from Peter Relan, Crowdstar’s investor and CEO: “We larger players have to be more disciplined when acquiring users, smaller players have liberal VC money to work with.” Was it just me that jumped off the chair upon hearing that? How ironic [...]]]> Log from “Inside Social Apps, San Francisco Jan 25, 2011″

1. This nugget from Peter Relan, Crowdstar’s investor and CEO: “We larger players have to be more disciplined when acquiring users, smaller players have liberal VC money to work with.” Was it just me that jumped off the chair upon hearing that? How ironic is that: everywhere else its typically larger players who have deep spend budgets. Smaller players need to be more disciplined about spending. At least that’s how it works in real life, except for FB games (2011) and pet food (1999). Worrying sign.

2. FB Credits – the issue of 30% admission fee revenue hit was clearly on folks’ minds. Us, being a big data shop, read it to mean that micro-transaction volume is going up. Monetization data just got bigger.

3. Someone in the audience asked about analytic tools used by the big guys. NO ONE SAID A WORD. Again, no surprises there: business intelligence is just too strategic in social gaming to reveal *anything*.

4. SNS’s that are alternative to Facebook: there was talk about international, FB owing real-life-friends social graph, what about other social graphs. Yeah, OK. We agree. But we were surprised no one mentioned the expectation of higher ARPUs for alternative SNSs. The way we see it, the only way niche SNS’s will survive is if they can help monetize better. If you have fewer users, and you cannot monetize them any better, then our judgement is that the benefits of scale will crush you.

5. Evolution of a platform: initially platform expertise matters most. Later: quality rises; brand matters more. Makes sense. One of the investors in the morning plenary recommended looking at Android when asked about the next opportunity, not FB, not iOS. As much as we like Android, we happen to know that that investor has a vested stake in that statement, so would doubt its objectivity.

6. Someone made the comparison of Twitter being a social graph of things that interest you. That was a new way of looking at it (for us anyway). Another panelist: Linkedin is only a reference point to a record, not really the full promise of social app – hmmm, maybe. Although I do not suscribe to the popular Linkedin ritual of publishing one’s resume, it seems pretty sticky to me.

7. Should you translate? If Hollywood movies are dubbed in a market, you should translate in that market. Neat insight.

8. IE (Internet Explorer) – 10% lower conversion rate than other browsers! Wow. This is why data driven insights are so precious – who would have guessed?

What does this have to do with BI? Let me explain:

Imagine that you just rolled out a [...]]]> There are many management fads that come and go: Poka-Yoke, Quality Circles, Taylor Principles, Management by Walking Around…I am sure you have your favorite.
One simple one that has stuck in our minds is “Management By Exception”.

What does this have to do with BI? Let me explain:

Imagine that you just rolled out a corporate data warehouse with 500+ metrics stored in it. How do you know which 10 of these to look at on any given day? Should you look at the “strategic” ones first and then drill-down to the rest (Balanced Scorecard)? Or should you look at the ones which are trending out of bounds?

Most managers we speak to prefer the latter approach: tell me where the “fire” is. Management By Exception.

But here’s the rub: by its very nature, you do not know a-priori where the “fire” will be on any given day. So if your BI strategy is to use a row-oriented database for your corporate DWH, augmented by spot summary tables to make it perform, you may find yourself boxed in. On any given day, you may find that one of the areas on “fire” has a metric for which your IT team didn’t pre-compute an aggregate. And *just* when you need to do rapid-fire BI, it fails you – until your IT department can turn around and build a new summary table for that particular metric x dimension combination.

Avoid this you can. Use an analytic platform that scales at query performance preserving atomic grain. In this Black Swan era of increasing volatility, BI patterns are becoming increasingly unpredictable.

Think about it. Do you disagree? Email us at blog_at_metricmine_dot_com

Although there are functions available in Vertica to convert to the usual yyyy-mm-dd format, doing this at query time is generally a bad idea. It will work, but you will be grossly underutilizing one of Vertica’s key [...]]]> One common pattern we find in many gaming companies is the use of UNIX time when loading Vertica.

Although there are functions available in Vertica to convert to the usual yyyy-mm-dd format, doing this at query time is generally a bad idea. It will work, but you will be grossly underutilizing one of Vertica’s key strengths: the use of data compression. Unix timestamps compress pretty badly. Their grain is too fine, and they are almost certainly going to be useless for reporting in their base form.

If your reports need to report by day/month/year, or hour-of-day, you are much better off converting to the appropriate grain at *load* time. This will enable Run-Length-Encoding of your time dimension, resulting in much, much faster query performance. (We like to leave both unixtime and gregorian time in the super-projection, and only include gregorian in other projections to balance query performance against data storage).

Uh-oh, the time now is 1287626436, time to go!

But we also wonder if Greenplum’s ASP is about to shoot up. Once in 2006 I had a little budget left over and decided to add some storage. Included [...]]]> The first thought we had was that EMC’s acquisition of Greenplum puts the analytical database market in the limelight, good for the big data industry.

But we also wonder if Greenplum’s ASP is about to shoot up. Once in 2006 I had a little budget left over and decided to add some storage. Included in the eval were: EMC Clariion CX300i, Celerra NS350, HP MSA-1500, Netapp FAS270, Netapp 3020 and a Buffalo Terastation Pro (Buffalo was just for kicks, not a serious contender). I ended up choosing HP, and recall EMC’s high initial price, and premium support offering.

http://bit.ly/ODR5t

1. The response one of my clients gave me when I asked them why they chose the Cloud BI stack over Omniture: “For our data volumes, Omniture wouldn’t have worked – just too costly“.  (BTW, they all use Google Analytics in conjunction with their Cloud BI deployment, its an [...]]]> Some thoughts on the Adobe-Omniture acquisition -

1. The response one of my clients gave me when I asked them why they chose the Cloud BI stack over Omniture: “For our data volumes, Omniture wouldn’t have worked – just too costly“.  (BTW, they all use Google Analytics in conjunction with their Cloud BI deployment, its an “and”, not an “or”).

2 – Today’s WSJ reporting on the Omniture deal: “Web analytics generates about $600 million in world-wide annual revenue now, but the industry is expected to grow to $2.2 billion by 2011″.   The Cloud BI stack is ideally positioned to benefit, and of course, a rising tide lifts all boats.

3. I am reminded of the Forbes article “Green Tendrils” in their Sept 21 (2009) issue: “Some of the Valley’s new tech conglomerates also may be checking out software firms offering ‘business intelligence’ and related services to help companies track and analyze data used in marketing and in managing suppliers.  Jaspersoft fits the bill; so does Talend, which helps companies move and link data between various storehouses”.