Comments for Potential: Revealed

Comment on Predictive Analytics: How it Works (#2) by Randy

Randy — Mon, 26 Oct 2009 14:57:59 +0000

Manuel: you make some great points and you are spot on. Currently I’m working with a client on an analytics service (SaaS) and the notion of timeliness (e.g., real or near real time) is a differentiating factor. Also the cross-channel coordination and optimization so the right offer is made to the right customer at the right time AND place. And when any results from any of those channel interactions are available to ensure the feedback loop is correspondingly real or near real time in terms of updating across channel touchpoints.

Comment on Predictive Analytics: How it Works (#2) by Manuel

Manuel — Mon, 26 Oct 2009 13:58:28 +0000

Nice post Randy. I agree with your conclusion that results can be obtained without complexity or perfection. In my experience some of the initial benefits come from the data cleansing and data discovery exercises. Being able to visualize and profile clean data can deliver insights from the get go. To make analytics useful it is critical to be able to operationalize the results/scores in a timely fashion by making the insight available to the customer touchpoints so the right systems or individuals can take action.

Comment on Predictive Analytics: How it Works (1) by Predictive Analytics: How it Works (#2) « Potential: Revealed

Predictive Analytics: How it Works (#2) « Potential: Revealed — Mon, 26 Oct 2009 02:58:58 +0000

[...] analytics, data-driven, decision-making, experimentation, insights, marketing, Practical In the first post about predictive analtyics we learned about the essential building block of predictive analytics: the predictor. This is a [...]

Comment on Predictive Analytics: How it Works (1) by adamgrizzly

adamgrizzly — Fri, 25 Sep 2009 17:11:22 +0000

I know you love that Jim Collins book. You quote it all the time to me! I continue to like this angle you’ve been taking lately on “practical” ways to approach complex things like strategy and now analytics. I’m adding it to my vocabulary to use in our management team meetings. Grizz.

Comment on Predictive Analytics: How it Works (1) by Randy

Randy — Wed, 23 Sep 2009 13:06:50 +0000

Good question 99. I might paraphrase Jim Collins from Good to Great, "good enough is the enemy of great." But in practical terms good enough at any given point in time can be great. It is usually only in hindsight that you recognize clearly that what was great before is not only good enough or possibly less. So I'd say don't get hung up on whether current models are great but whether the current ones are better than what you were using before. And that the next day you are actively looking to make current "champion" models the loser in a contest against a "challenger" model you've developed. The simple example is where all that is being done today is targeting your customers based on a profile, say, of how what zip code they are in. If you only added predictor that targeted based on recency of last purchase you would likely get an improvement in results. That's a very simple model yet it is possibly a marked improvement for your business. Hope that helps. R

Comment on Predictive Analytics: How it Works (1) by marketingagent99

marketingagent99 — Wed, 23 Sep 2009 11:20:25 +0000

Are you saying that simpler predictive analytics are not enough? Aren’t there places where simple is good enough or even best? I like this post though because even though my company uses some strong modeling to improve our marketing we still don’t always speak the same language on analytics. The basic definitions are important for every to understand.

Comment on Happy Customer by Randy

Randy — Wed, 12 Aug 2009 16:38:19 +0000

Your concerns are common ones. The data is stored at a highly secure (and brand name) “cloud computing” service: Amazon.com’s. They provide highly secure, scalable and reliable services to thousands of companies, a service they started a few years ago to take advantage of the fact that they have world class IT infrastructure and at any given point in time excess capacity (for redundancy, DR, high volume periods) that could be use by others if a model for provisioning it could be developed. That they did and it has been pretty successful for them (and small software and solution providers who otherwise could not afford to offer low cost, SaaS solutions in the early stages). As to cost to value: I’d say that for an average customer of my client, the return on investment (calculated as incremental profits earned over the cost per month to subscribe to the analytic service) achieves payback in about 6 months or less. A modest sized company with say $10 million in revenue can easily find ways to make more profitable sales and customer decisions through smart analytics against their existing sales and CRM data that cover the $500 per month cost of the service. Think about it: it is probably less than a 1% improvement performance for a company of that size.

Comment on Happy Customer by Randy

Randy — Wed, 12 Aug 2009 16:27:56 +0000

I agree totally. The junk info out there is intimidating to the experienced and overwhelming to the uninitiated. On the community rating model that Salesforce.com uses it both helps the community better determine if a vendor’s solution really delivers as advertised. Plus Salesforce.com uses it to select which partners it might be willing to do some co marketing or promotion with and generally pick the winners from the losers in their large community of Appexchange developers. So getting and maintaining good ratings has multiple levels of value for the vendor too which then reinforces the incentives for openness and transparency any good community requires to be healthy.

Comment on Happy Customer by adamgrizzly

adamgrizzly — Wed, 12 Aug 2009 16:22:47 +0000

I know who your client is and we didn’t choose to go with total SaaS for our SFDC analytics. Some concerns about where that data is stored and performance having it be remotely stored. I’m sure companies find it compelling though. The cost is very low relative to value if you will commit to using the analytic outputs. Thanks for sharing. G

Comment on Happy Customer by tommck

tommck — Wed, 12 Aug 2009 15:13:53 +0000

Big fan of the “community reviewed” model for B2B. So much of B2B products especially software and technology are full of misinformation in their marketing (“spin”!) that is hard for smaller companies to sort through. They lack the resources larger firms have to get expert advice or have staff that can do lots of due diligence on vendors solutions. Your client is brave to go this route but probably doing a good job and being rewarded for the risk if they are getting positive reviews from customers. T