Chris Crosby

Speech Recognition is about identifying what people are speaking.

Speech Analytics is about figuring out what people are saying.

Last fall I participated again in the ICCM Canada Keynote Panel: 60 Ideas in 60 Minutes moderated by Paul Stockford from Saddletree Research. Dave Butler over at NACC recorded the session and has been distributing the ideas presented in his monthly newsletter. I keep promising him that I’ll expand one of mine into an article for him, but in the meantime here is the one he sent out today. Since it’s one of my favorite rants, I thought I would share (pardon my grammar):

Take out the garbage, I am not talking about employees or customers, I am actually talking about reports and data. One of my pet peeves, and I could go on for hours but I will go on for 45 seconds, is when you walk into a call center and you see the reports that supervisors are looking at every day and the first column you see is calls answered,  and this is for an agent. Johnny had 27 calls yesterday and was logged in for 15 hours, blah, blah, bah, blah. Step back and ask yourself what value you are getting out of this information. So take your 30 column report and pare it down to four or five columns that you can actually impact and actually take action on. If you can’t impact whether an agent is logged on for six hours or seven hours, get rid of the column. Just say, you know what, what was their schedule adherence, or what was their hold %? In other words, what are the columns that you can influence? Then write out the business value for each column on the report. Are you going to see service level on there, or outbound calls? Write down why you need to see that so you can articulate that back to the people that are managing to that data every day and why it is important.

I couldn’t have said it better myself

I came across a clever demonstration of Analytics from a Business Intelligence software provider called QlikView, and thought it worth sharing. Note- In an indirect way, Latigent competed with QlikView and I always admired their clever marketing campaigns. In this application of their analytics tool, they dumped radio airplay data from  MediaGuide.com into an OLAP cube and overlayed the QlikView front end (demo found here).

Originally I just wanted to play around with the data, so I filtered based on the greatest band of all time, Van Halen.

Unexpectedly, a couple of things struck me about the results:

1. The top two most played songs both appear on the album 1984, but 1984 is the second most played album: Seems a bit counter intuitive at first, but upon closer inspection we discover that the album Van Halen’s songs may be played fewer times, but there are more of them. The aggregate is greater by more than 300 song plays. There is a lesson to be learned here about product distribution, and a pretty good example of Chris Anderson’s Long Tail theory in action.
2. Why the big disparity between the “David Lee Roth Van Halen” and the “Sammy Hagar Van Halen”? My assumption is that this is most likely because Van Halen is on tour right now with David Lee Roth. No doubt radio stations are heavily promoting this tour by spinning the vintage favorites. My hypothesis though is one that is impossible to prove given the current data set. One would “just need to know” that these guys are on tour to draw that conclusion. Without this information, what conclusions would you draw? Would they be accurate? Is mine accurate?

Now, an interesting exercise would be to take the chart below that displays where these songs are being played and overlay the tour schedule. Also, the data is only available from Feb 24, 2007 to current. What would a wider data set show us? Is the distribution reversed when they’re on tour with Sammy Hagar? What about when they’re not on tour?

This example demonstrates that the unlocked power of analytics is not just about spotting trends that you otherwise would not have, but its often in finding and qualifying external (and sometimes non-structured) data points and quantifying their impact. It also causes you to ask questions and seek answers that you otherwise wouldn’t: What long tails are hiding in your data? How can you leverage them? What external events influence your business? How do you qualify them, and quantify their impact?

What else do you not know you need to know?

Looks like Pentaho closed a $12M Series C round of financing. This is exciting stuff. With the consolidation of large Business Intelligence players its opening the market for the already under served SMB and for Enterprise BI projects looking for a lower cost of ownership.

The only real question is if these guys will join the ranks of Zimbra and become a promising start-up swallowed by a behemoth too soon…

“The process of discovering what you don’t Know you need to Know.”

-Chris Crosby

A few weeks ago, my wife Amy and I had the pleasure of meeting the newest addition to our family. My brother-in-law Jeff and his wife Pam welcomed thier new son Braden Scott Parma-Kelly to the world in Feb of this year. We couldn’t wait to meet the little fella, so we headed south to Austin, TX.

Not having children of our own yet, this was our first real “hands-on” exposure to the process of baby rearing. It was quite a sight as we fumbled through that first diaper change and bottle warming with Amy screaming “What do I do, what do I do?”. But by the end of the weekend we were self-proclaimed professionals.

As I reflected back on our weekend in Texas, I was able to draw some interesting parallels between raising babies and Business Intelligence (yes, I can find a metaphor or analogy in just about anything).  

So, allow me to introduce a new kind of BI, “Baby Intelligence”. Enjoy! 

1. Baby Activity Monitoring (BAM) – Most of you have heard of Business Activity Monitoring, which is the real time management of your  business transactions and performance . However, Braden kindly introduced me to the “baby monitor”, which is basically a one-way walkie talkie that you carry around the house while a baby is sleeping. This allows you to listen for any disturbance and activity. I couldn’t help but think that the monitor looked just like a gauge on a BlueVue Dashboard. I quickly learned that they worked the same way. The occasional two bars on the led monitor indicated a little restlessness that is apparently a part of Braden’s normal “sleep cycle”, but three bars or more means you better get your butt in gear or the thing is about to go off the charts…
2. Integrated Solutions - I’m a big advocate of efficiency in not only business and call centers, but in life. I often use this as my reasoning to invest in cool toys like iPods and Blackberry’s that make me more “proficient” (although I think my wife has caught on to me). And thanks to my brother-in-law Jeff, I have now discovered the modern marvel in baby efficiency: “the instant stroller”. The day we took Braden to the park for a walk I observed Jeff take the car seat out of their new highly safety rated, four-door SUV and drop it right onto a stroller frame. My jaw dropped in amazement at the simplicity of the set-up as there was no longer a need to un-strap the kid from one overly sophisticated plastic bucket and transplant him into another. Brilliant!  
3. Key Performance Indicators and the Baby Scorecard- In a call center there are Key Performance Indicators (KPIs) and metrics such as First Call Resolution and Schedule Adherence that measure performance and areas that require attention. These are often accumulated into an Agent Scorecard that can be used to manage the performance of your team. However, through my Sister-in-Law Pam, I was introduced to an entirely new set of very important metrics, which I will refer to as “The Baby Scorecard”.  These include key indicators such as: “is his diaper wet?” , “does he “stink?”, “that was a quality burp”, “how long was his nap?”. I must also mention that the term “Blow Out” now has an entirely new meaning to me…    
4. Smile - Even if you have no clue what’s going on around you

-”Uncle” Chris

I was recently asked by Call Center Magazine what I thought the difference between analytics and performance management is so I thought I’d post it here as well.

Latigent defines Analytics as the process and enabling tools for root cause and trend isolation. This can range from call volume drivers, to customer segmentation, to agent performance. You hear a lot about slice and dice and drill-down when it comes to analytics. These are very important; however a well defined analytics product should also let you correlate disparate events together and analyze their impact on each other.

Performance Management ties in broad based metrics to paint a picture of individual performance against company objectives. This is where the idea of the “balanced scorecard” comes into play.

Performance Management and Analytics are often used interchangeable. However, Analytics is more about opportunity and problem identification, where Performance Management is geared towards generating the report card and faciliating the process for improvement.

-Chris

Have you seen the Verizon commercial where the two guys download the 80’s tune “Rock the Casbah”, but instead of singing along to the correct lyrics they misinterpret them into something completely different? Over the course of the commercial they botch the chorus into “lock the cashbox” and “stop the cat box”, which takes a perfectly good rock song and makes it complete gibberish.

The commercial ends with the narrator stating:

“You don’t have to understand your music to understand how to get it all from your PC to your Phone” 

Albeit, I think the commercial is funny and is reminiscent of my grade school days running around the house playing air guitar and trying to belt-out Van Halen tunes (ok, maybe that was last week); it got me thinking about something not so funny: how often I see reports and metrics that end-up getting botched because somebody misinterpreted a few lyrics along the way.

Unfortunately in the contact center, you do have to understand your data to understand how to get it all from it’s source into your reports, dashboards and scorecards.

Here’s an example of what I mean:

Average Talk Time is almost always calculated as (Total Talk Time/Calls Answered). Seems simple enough right? We’ll, not always…

Some of you may have noticed in the Avaya CMS tables there are two fields that represent “total talk time”: ACDTIME and I_ACDTIME

According to the CMS documentation:

ACDTIME = The talk time of all ACDCALLS (calls answered) for an interval.

I_ACDTIME = The length of time during the collection interval that agents were on split/skill ACD calls.

What this means is that ACDTIME represents the total talk time for calls that were physically answered in that interval vs. I_ACDTIME which tallies up all of the time for a half-hour interval that agents were actually talking.

In this example, you would use ACDTIME for an average talk time calculation and I_ACDTIME for an occupancy calculation. Interchanging those two fields incorrectly in a report changes the tune completely…

Another common misinterpretation I see is in the Cisco ICM/IPCC Enterprise world. Cisco makes available both Calls Answered and Calls Handled for reporting purposes. These terms are often interchanged in verbal context, however according to the Cisco database schema:

Calls Answered = Number of calls answered by agents associated with this skill group during the half-hour interval.

Calls Handled = The number of inbound ACD calls answered and wrap-up completed by agents associated with this skill group during the half-hour interval.

So, Calls Answered are pegged to the half hour interval when the call is physically answered, and Calls Handled pegs when the call is actually finished. The difference seems subtle, and over the course of a day the grand total should be the same (except for calls running over midnight).

But what happens if you feed your WFM application Calls Handled instead of Calls Answered? You got it, inaccurate call arrival patterns and forecasts. What would seem like a relatively minor botch will end up having a significant impact downstream on your Service Levels and staffing efficiencies.

I remember as a kid cracking open my first cassette tape that actually had the lyrics printed on the inside of the cover. This was a novel concept as I no longer had to decipher them on my own. This same concept applies to your reporting. Most vendors publish a document that explains, like the examples above, the meaning and usage behind their database elements.

When designing a datamart or report, it’s critical that you reference these documents as your lyrical road map. Also review them whenever you upgrade your ACD or WFM system, as often times fields will change with new product releases.

The best way to avoid these types of pitfalls is to consult an expert or somebody that is familiar with each vendors’ nuances and understands your desired end game requirements. (cough, Latigent)

Before you know it, you’ll be Rockin the Casbah…

-Chris