The difficult economy and customer behaviour

How is your customer behaving these days? The average customer is restless during times of economic comfort. So what does the opposite end of the economic curve mean to customers?

1. Price becomes more important among differentiators

In a mathematical model, if you attach a weight to each of the metrics that influence customers taking to your product, price would obviously be among the top few. This weight would definitely have increased, especially over the last few months. Conscious purchases would increase as opposed to periodic automatic purchases.

2. Advertisement becomes all the more important

The average customer would definitely spend more time doing research on the product. So, getting the word out that your product is more economically viable, becomes most important.

3. A field of opportunity

These times are the best times for any enterprise to show that they care about their customers. Small measures, such as delaying payment due dates, could greatly influence customer loyalty and churn.

4. It starts as slight pitter-patter on a tin roof

Customer behavior shifts ever so slightly before the deluge. The number of people opting to cook at home was at 74% last October, which increased to 76% last January, and has drastically increased to 84% by April. Identifying slight changes in choice shifts requires an analysis platform to understand what numbers could be used as touch points.
Also, buying patterns would vary drastically and would not be as obvious during better times. The nuggets of information would be in the details and might not jump out at you easily. This further highlights the need for detailed analysis.

Now what would companies need to do ensure that they stay on top of the curve?

• Create an analytics platform to identify small shifts in customer behavior.
  
• Concentrate on customer satisfaction. A dissatisfied customer is more likely to discontinue service during such times. Retention becomes more important because getting a new customer becomes all the more expensive.
  
• Pricing trends based on local competition, while customers are busy shifting to cheaper, local vendors.
  

A First Look at some Analytics Vendors

Avoke	1. Caller Experience Dashboards 2. Analysis Workbench 3. Drill Down & Listen	1. Completion rates, Transfer Rate 2. Cross mapping high level parameters without understanding pain points 3. Shortlist calls to listen and troubleshoot IVR issues
Chrysalis	1. Agent Productivity Tools 2. IVR Hardware Monitoring	1. Tools that supervisors could use to monitor individual agent performance 2. Port utilization metrics / Connection status
Clickfox	1. Traffic Analysis, IVR Path Analysis 2. Crosschannel Analytics	1. Detailed granular data of path completions, failure points and caller interactions 2. Integration of caller interactions among multiple channels
Contact solutions	1. Application Reporting 2. Analyze high level metrics	1. Lists of called numbers, calling numbers, Duration, Timings 2. To reduce live agent traffic, improve agent routing
Eloyalty	1. Cross-channel analytics 2. Trend and pattern analysis based on prior experience	1. Collects data from every possible customer interaction method 2. Provides recommendations based on reducing customer discontinuity
Metrica	1. Detailed IVR customer behavior analysis 2. Troubleshooting IVR performance	1. Analytics through multiple channels. IVR, Contact Center and Website 2. IVR path analysis, Clustering and Prediction of customer behavior
Micro Automation	1. Self Service Analytics 2. Call Control Analytics	1. Reporting tool that gives caller choices and success rates without going into whys 2. Agent availability and call duration in real time
Nuance	1. Voice input recognition programming 2. Modules that ease customer interaction	1. Facilitator for voice recognition systems used within IVRs 2. Efficiency measurement of free style customer interaction
TuVox	1. Self Service Module creation 2. Speech applications for voice recognition	1. Industry specific voice recognition systems and self service modules 2. Call routing technology to plan IVR navigation
Voice Objects	1. Real Time Analytics 2. Interface with other BI tools	1. Real-time analysis enabled without requiring additional load or transformation steps and changes automatically reflected 2. High level metrics such as Repeat calls and completion rates
Voxify	1. IVR modeling tools 2. Big picture metrics of IVR utilization	1. Individual self service modules for each application or 'menu' 2. Call completion rates, call timings and day by day call volume trends

We would like to hear about people's experiences with these products, the benefits they derived, and where the products need improvement.

The rising importance of cross channel analytics

A 2008 survey has concluded that 75% of automated customer interaction is being attempted through the IVR. This used to be 80% in 2006 and is bound to further reduce over the next few years, but only in comparison to the other automated methods such as SMS, emails and the web. Also, with the prospected increase in utilization of such channels, the volume and reach would make analysis through just one channel moot. There is hence an increasing need to track customers through multiple channels to completely understand their behaviour and make business sense out of it.

Enterprise Feedback management (EFM) is a field of research gaining in importance rapidly. This involves collecting survey / behaviour information through multiple stakeholders and multiple channels. Enterprises understand the necessity to collaborate between various methods of data collection, thus necessitating tying up disparate data feeds.

Now the issue that arises is about tying user records in disparate channels together. This is mostly performed through authentication, and there is a tie up between the customer’s Telephone Identification Number if the interaction is through the phone, the credit card number if it is through a kiosk or a retail store, and the username / password if the interaction is through the computer. But what about tying non–identified customers through channels, and maintaining their records? What about callers who directly transfer to agents? What about non authenticated using the web? How are their records tracked? Finally of-course there is the ethical issue. Do these customers really want to be tracked? Here are some of my thoughts around that idea.

•          Cookies in a computer or recording IP addresses would be able to cross map customer interactions. This would need coordination with multiple service providers which is hard to come by.
  
•          In the near future, RFID tags with customers would be able to record store movements. This of course raises issues of customer security of misuse of customer interaction data. Services could be offered to customers by choice, and data could be maintained securely. 
  
•          Bunches of data could be analyzed together by demography, which would be an incomplete way of analyzing customer trends. The idea is to store only the meta-tags of customers, as in region, income and spending patterns instead of actual customer details, to unearth insights or in providing easier self service. In other words, a master database exists that just categorizes customer details such as credit cards numbers are tied to just meta-tags and not customer name or address. Such a database would be of great value to the analytics community. 
  

Shades of Grey in calculating Completion

Completion is one of the key metrics used to monitor performance of self service systems. There are wide discrepancies in the way organizations measure and make sense of this metric.

Some organizations calculate this to be the share of callers not transferring to agents. Some take this as the share of callers reaching the Main Menu. Others take into account the callers reaching the most basic of self service choices (say, hearing Account Balance for a Financial Services firm, or Recharge Amount left for a Mobile Service Provider).

All these methods have their merits in aiding understanding of the performance. But they do not tell us the true measure of self service. Method I tells us the hang-up rate of the IVR, but does not differentiate between callers who hang-up after succeeding in their task and those who hang-up after being unable to complete their objective. Method II is useful in that it tells us the number of people who reached the Main Menu and hence its complement, the share of callers who drop even before reaching the Main Menu. Method III is one of the most popular ways, and this method gets closer to actual self service performance. But, it mostly uses a default readout such as Account Balance, which callers automatically hear after getting authenticated. This means that callers may or may not have wanted to hear their Balance, and in assuming that all callers hearing their Balance to be self served is an error.

If all these methods have some issues, what is a better method? The more appropriate method would be to understand the number of callers succeeding in serving themselves after starting a particular process.  The following are some examples:

• Callers requesting for roaming activation on the IVR, and getting it activated right there
  
• Callers recharging their prepaid mobile by keying in the requisite codes
  
• Callers requesting for a Checkbook and the system accepting their request
  
• Callers starting a funds transfer on their IVR and completing it
  
• Callers requesting for a sports package for their satellite TV, and getting it added
  

Organizations could have high hang-up rates, high Main Menu rates or high rates of reaching Account Balance. But, none of these methods take into account the intention of the caller. This method enables us to understand their self service performance better, in that it takes into account only callers who started a process, and of those callers, how many succeed. That is the true measure of Self Service Completion.

How "human" is your IVR?

Customer satisfaction levels are the lowest for the IVR amongst all customer interaction channels. This is because the channel has the largest difference between expectations and delivery. Most callers expect to speak with an agent, but are instead put through to a system that attempts at resolving their issues, sometimes inefficiently.  

Agents usually have a tight script that they stick to during interaction with customers. Hence it would seem obvious that an automated IVR which follows the same script would be just as effective. This premise has often been the cause of low customer satisfaction and IVR efficiency. The best way to increase customer satisfaction levels would be to try model IVR interaction as closely as possible to a human interaction.

When we start thinking about what the agent has that the IVR does not have, two words come to mind. The agent is more ‘Human’ and ‘Intelligent’ than the IVR. Every other factor that separates the two falls under either one of the two categories, or both. Humanness refers to the innate fluctuations in human behavior, voice and the reception to stimulus. Intelligence is the intuitive capacity of the human to arrive at the same page as the customer is, as quickly as possible. IVRs today are striving towards achieving both these qualities. The following paragraphs focuses on how IVRs could be made more intelligent in each of the three key areas: Caller intent registration, Caller intent process and Call conclusion.

Remembering previous caller choices, intuitively knowing payment times / password change times and proactively suggesting such operations could be one of the primitive ideas to make the IVR more human. Yet, most enterprises IVRs today do not have such options. Analysis software that trends caller behavior over a few months feeding into the IVR design could be implemented to further suit customer requirements. Also, a seasonal change in IVR menu structure for all callers / groups of callers would add value to the process of registering caller intent.

While the caller is being serviced, and is interacting with the IVR, present day IVRs offer failsafe states and retries where the caller re-enters the same input. What would be more intuitive is to offer ‘Is this what you mean?’ choices that customers are used to answering in web pages and in human interaction. This would greatly improve the performance of retry states, and would reduce transfers to agents. Also, while giving instructions to the customer, an IVR could introduce minor fluctuations in the voice / word patterns to imitate real agents.

In the event of repeated failure within a process, and after failure of even the failsafe methodologies, the process could be restructured in a way to suit what the customer wanted to do. For example, if a customer wants to transfer funds, and he has trouble in entering the source account number, the destination account number and the amount to transfer could be collected before transferring to an agent, thus reducing agent occupancy times.  

During the conclusion of the call, today’s IVRs offer a ‘do you want anything else’ module that routes the caller back to the primary caller intent registration module. Here, options could be provided based on customer patterns within the call and during previous calls, to quickly get to that function. Returning to the main menu would be both tedious and unnecessary. Interactive follow up through other channels regarding the status of request is also a smart way of increasing satisfaction and reducing repeat calls. 

These are some of the ways to design a more human IVR. The IVR is not going away as the most popular automated channel for customer interaction anytime soon. Enterprises must realize that increasing customer satisfaction within this channel is top priority in order to gain service advantage. 

Why sampling fails in IVR Customer Behavior Analysis

Enterprises all around the world are investing heavily on automated Customer Interaction systems. Some enterprises have millions of automated customer interactions per month and hence have the necessity to create metrics that reflect the performance of the IVR.

Metrics such as self service completion rates and the repeat call rate to the IVR provide but a peripheral and delayed view of its performance. Everything is hunky dory as long as the metrics show high self service completion rates and low repeat caller rate. It is when the home grown metrics showcase below par performance, that management eyebrows are raised.  Hence, root cause for failure is analyzed based on logged caller interactions within IVR. Statistical analysis based on call samples is an often misused methodology, and I have a feeling that it breaks down magnificently in analyzing call tree structures.

IVRs call volumes are distributed heterogeneously. There is a high level of randomness in both call origination and in caller behaviour within IVR. Chronological external factors such as time of day or month and unpredictable factors such as the state of the economy, factor in a random manner deciding caller behaviour within the IVR. Simple sampling methods hence break down magnificently.

There is no way to pre-know problem areas within an IVR. If IVR calls are sampled after identifying problem areas, we run the risk of delaying important business improvements which could very quickly snowball into loss of business. This rules out sampling ideas such as quota sampling and cluster sampling.

Shaky sampling techniques raised to complete lack of knowledge of IVR performance render it impossible for caller behaviour to be unequivocally represented in any sample set. This would hence further impact significantly in recommendations for IVR performance improvement.

Business decisions based on incorrectly represented data, directly results in lowered customer service satisfaction levels and increase maintenance costs. It is hence imperative that we utilize the raw data utilized in call the calls to aid business decisions. In these days of increased computing power, the term "it has been statistically proved" loses significance in a lot of applications. IVR interaction is one of them.

What is an intelligent IVR?

IVRs, even though have been around for over 30 years, do not have the kind of maturity that we expect of a technology that old. Here are some of the ideas that would greatly improve performance and uptake of IVRs. VEry few IVRs exhibit these characteristics, and it would be great if more designers adopted these features. 

Memory – Memory to  remember caller choices both  within the call and from  earlier calls.

Adaptability / User Friendliness – Adaptability to change parameters and menu choices per various input factors such as area of caller.

Human Resemblance – Give the caller a generic feel that he is interacting with a human. This can be achieved by randomizing certain play prompt messages to reflect human variations in speech.

Seasonal Variations in IVR choices – Callers have specific patterns in performing operations. The IVR can change itself on a periodic basis to reflect majority caller choices. For example, the IVR for the TV channel can automatically push feature addition to the top during cricket season.

Interactive Follow-up of action – The IVR can follow up through another channel to complete the operation / provide confirmation instead of making caller spend extra time on the IVR waiting for databases to be updated.