The EPA was in my neighborhood several weeks ago testing well water -- never a good sign. It was determined that the subdivision just north and upstream of us had once been the site of a farm contaminated by the dumping of fuel, pesticides, and various other chemicals.
It turned out that our water was fine, but as a precaution due to proximity, I am following the EPA's recommendation for the subdivision itself -- the installation of a carbon filter.
For me this is no big deal, as I already have a large, full house filter installed. However, I had recently purchased three of the large, standard filters to last me the next 18 months, and was not happy about the now duplicate costs to replace them.
I had ordered my standard filters from a local, online N.C. distributor that I'd been dealing with for a couple of years, FiltersFast.com, chosen originally not just because of price, but because it also carried the O-ring and specialized wrench I needed.
So I picked up the phone and called into the customer service number, expecting the worst. Now, I don't know if I simply got a very experienced customer service rep, or if they were employing something like a rules-based adaptive customer experience system (ACE), but I ended up with exactly what I wanted and needed, all with no duplicate costs.
If I'd gotten an inflexible customer relationship system, I'm certain that they'd have simply tried to follow the standard script and upsell me, when in reality what they were dealing with was not a sales growth opportunity, but a retention issue where risk management, and not marginal profit, was paramount.
In the end, I was able to order three of the carbon filters, got full credit for returning the unneeded standard filters even though I was past the 30-day return window, and the rep got authorization to give me free-shipping to compensate for my paying the return shipping. Just try that with a script-driven process run out of an off-shored call center!
The above-mentioned real-time ACE decision system is just one example of a rules-based decision support process, the topic of this past week's DM Radio show, "Rules, Rules, Rules." If you've not heard of DM Radio, Eric Kavanagh hosts a weekly (every Thursday afternoon, 50 weeks a year, at 3:00 p.m. ET) one-hour discussion on current IT and business system issues.
His Ed McMahon-like co-host, Jim Ericson, played the likeable curmudgeon, this time lamenting that he can never get a real human, no matter what number he punches, before we moved into the heart of the issue, covering ground from expert systems, to fuzzy rules, to real-time.
This week's guest analyst was Dean Abbott, founder and president of Abbott Analytics, who introduced the concept of "association rules," an important extension to the more common fixed rules of the somewhat brittle "expert systems."
Designed for a more complex environment, associative rules work like this: You walk into a grocery store in July and purchase, among 25 other things in your basket, hamburger buns and ketchup. The system recognizes the significance of these two items paired together (a backyard cookout) out of all the other possible pairs, triplets, quadruplets, etc.... and automatically creates a coupon at checkout for $1.00 off a bag of charcoal (which it recognized as NOT being in your basket). The bag is conveniently sitting near the exit -- you can pick it up on your way out, and save yourself an extra trip back to the store later. It's a win-win for both you and the retailer.
A hardcoded, binary decision system could never handle this level of complexity, even though it's only 25 items, whereas it's a relatively trivial exercise for high-performance analytics, and associative rules are smart enough to know that without the hamburger, pairing the ketchup with anything else in the basket is insignificant.
Developing an effective rule-based system involves three aspects:
- Assess the business process for the appropriate steps to consider automating with rules.
- Determine the best class of rules, and their specific parameters, to apply at each step.
- Design an effective overall system, people included, to best execute those rules.
When it comes to the first aspect -- assessing the overall business process -- I'll simply suggest these three resources: 1) My previous blog post on The Man Who Saved the World, 2) the book Normal Accidents by Charles Perrow, and 3) Kevin Slavin's TED Talk on Algorithms.
The point of this trio of resources is to educate yourself about the dangers and differences between tightly and loosely coupled systems, and the proper use of human intervention and exception handling to prevent situations like the one described by Slavin, where two computers bid each other up into the millions of dollars on eBay for a $12.95 paperback book. You don't want to be the one responsible for the next version of the Wall Street "flash crash" at your company.
The application of analytics to each of these aspects is ubiquitous, from data mining, to decision trees, to variability and confidence levels, to parameters and triggers. I started this post with the ACE real-time decisioning example, and it would be best to conclude it with additional illustrations of how analytics drives rule-based decision-making:
- At HSBC, big-data is no longer an obstacle to credit card fraud prevention. In-memory and in-database solutions mean that instead of just sampling their transactions, they can run 100 percent of them through the fraud detection software.
- The use of data integration on the front end of SAS Financial Management enables the three-day close by avoiding suspended transactions that can be resolved through a rules-based process.
- Near real-time capital market portfolio risk analysis to assess regulatory and internal policy compliance, such as liquidity or counter-party risk.
- Text analytics and social media sentiment analysis, where language and grammar are recognized as rule-based systems that can be analyzed for meaning and business intelligence just the same as the contents of a grocery cart. This is an increasingly important capability in managing your corporate or brand reputational risk, where negative news can go viral and spiral out of control before you have a chance to even recognize it as a problem.
As you can see, rules-based decision-making covers a lot of ground. As the now famous line by Warren Bennis goes: "The factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment."
Rules-based decision systems are permeating the business environment, and with the analytics available to help make those rules better, you can end up with better business decisions, faster.
This originally appeared on the SAS blog, Value Alley.