Agile Design in CX

Two comedians are interviewing for a job at a comedy club.  The first comedian tells the owner how funny he is and how often the audience laughs at his jokes.  The second comedian tells the owner a really funny joke.  Who do you think got the job?

The single most impactful way to make a meaningful and lasting impact on the customer experience is through directly impacting the customer in those critical ‘moments of truth’.   Knowing what these are and how to act on them doesn’t happen by accident, they are created through experiential design.

“But surely, one does not merely ‘design’ an experience, do they?” the meme might read.   Good question.

Historical Approaches to Doing Complex Stuff

Designing a customer experience is a complicated undertaking. In the history of humankind how we organize, conduct, and complete complex projects has a been a source of interest, study, and refinement for the better part of 4 millennia.

Surely the great pyramids were not done by shooting from the hip. Plans were drawn up. Architectural and mathematical principles were applied. Workflows and schedules were introduced. Workers were conscripted. Hookahs were smoked. One might expect a very rational linear process.

However, a review of history would argue that a pure linear process is not the source of humanity’s greatest accomplishments.  Course corrections are often made on the way.  Scrapping a large part of what was done or even “rebooting” completely happens often.

The scientists at Los Alamos in 1943 had a rough ‘plan’ to start with but was revisited when things did not work out as expected.  An intellectual punt on the concept of compression yielded the birth of nuclear fusion which in turn brought World War II to a close. Phew!

As Albert Einstein said “If we knew what we were doing, it wouldn’t be called research.”

It turns out that controlled chaos is good for humanity.

The learnings are this: if we are overly strict and structured we fail to innovate, barreling down a predefined path and not looking up.   If we have no structure and guiding princples, we just iterate aimlessly like a wind up toy bouncing across the table that eventually vibrates off the surface crashing to the floor.

These two approaches are book ends on a continuum that can be found in everything from the fine arts to the physical sciences.  Somewhat ironically, no more apparent is this creative control-chaos tension than in the world of software application development with the move from “waterfall” to “agile” approaches.


In the waterfall approach you specify the entire project, define it as tightly as possible, have it reviewed and approved, and then cut it into pieces for development.  Bing botta boom! It is speced and you can now execute like building a house.

The pros of this approach are that you have the scope of work defined and then can be very efficient in planning and executing it.  The downside is, no matter how good your plan is, you always run into the “unknown unknowns” if it is a unique (i.e., not mass produced) product.  If those exceptions are not handled right, it can deliver a death sentence to a project.  Also you run into the issue of waiting for the “grand unveiling” which can disappoint if expectations are misaligned.


Enter agile development approaches.  In this approach we have a rough master plan but iterate out chunks of observable work in the hopes of arriving at a Minimal Viable Product (MVP).  The plus side of this approach is we can see the work in progress and course correct as we go.  The challenge with agile is to not veer off in undesired directions and miss the original intent.  The ‘squirrel’ factor can be a big distractor in the agile approach if not carefully managed.

Planned Agile

Can we combine these two approaches in CX design?  I think so. We list purposely toward agile, but keep in mind our overall CX objective.  We need to be ready to go in and experiment.  We need to hash out some good enough (field) hypotheses and then implement them in a controlled environment and see what happens. We need to include the employees and the customers in the development of these ideas.  We should have good measures in place to help us determine what is working and what is not; operational, behavioral, cognitive, and emotional.  The guiding principles to your success in implementing CX lay in being simple, fast, local, and iterative.

  • Simple – overly complicated schemes die under their own weight. Make it a concise and simple idea.
  • Fast – let’s get the concept out there. Let’s test with data.  Long testing periods grow stale. Stakeholders become disinterested. Let’s fail fast.
  • Local – everything happens at the local level. When retailers see success, then we can use that grass roots success to spur deployment at a larger stage.  If it fails, the risks are relatively low.
  • Iterative –you will have complete duds. More often the execution needs to be tweaked and refined.  Don’t be afraid to kill off the duds.  But don’t confuse suboptimal execution with theory failure.

Selling More Booze

Let’s take a simple example of post purchase assistance. We might find from proper journey mapping that an area of opportunity is post purchase assistance at a beer and liquor distributor. We observe that for this particular retailer, the average purchase is not only high in dollars, but also high in weight.  We hear that customers sometimes struggle to carry their alcoholic loot to their vehicles.  We want them to carry more booze to their car. An area of opportunity?

We hypothesize by making a modest investment in assisting customers load their purchases from the store will result in better retention and increasing basket size over time. We construct a test. At this point we can either have a trial store test yoke with a similar other store or a pre-/ -post design.  I can walk you through those boring details some other time.

Next, we implement the new intervention (employees assisting customers to their vehicles) and observe the changes in the operational and financial metrics.  We might also want to ask customers directly about their perception.  Is this helpful?  How can we improve this?  Is this something we should continue? Is this something that would make you want to return.  We may also want to talk the employees about what they think.

This kind of micro-intervention can be carried out in a few weeks and then be tweaked or simply discarded.  If it works; awesome.  Let’s continue to do it, and do it better.  If it doesn’t, let’s fail fast and move on to the next thing.

The point is to be continually testing, revising, and improving using operational, behavioral, and attitudinal metrics as our guides to success.  Let’s not over complicate it.  We need to have a master plan but be willing to adapt. We also need to be agile in testing.  Find out what’s working and what’s not. Quickly, simply, and iteratively testing ideas at a local level will be the key to making a bunch of small wins…that can add to big ones very quickly.

NPS: Helped or Harmed?

It has been about 12 years since Dr. Fred Reichheld wrote his original article in the Harvard Business Review entitled The One Number You Need to Grow, which later led to an entire book on the topic.   This seminal work has had a profound and lasting impact on the CX industry.

For the uninitiated, the Net Promoter Score (NPS) is a 11 point anchored “would recommend” scale where you subtract the bottom 6 (detractors) from the top 2 (promoters) to get a “net” promoter score[1]. This measure has been enormously popular and widely adopted in many CX metric programs.  But one question that I think the industry needs to know is:

 Has it helped or hurt?


In the “helped” column it placed a clear focus on customer experience among the C suite types rather than pursuing “bad profits”; short term gains at the expense of long term enterprise health. Companies that otherwise would have not been interested or perhaps were intimated by the cacophony of opinions on the topic embraced customer experience efforts and invested heavily.

Indeed, according to Markets and Markets the total spend on CX in 2020 is projected to be roughly $10.8b up from $3.8b in 2014.  In case your math is rusty, that’s more than double the investment in 5 years.

Also, it has offered a nice shorthand to benchmark your own company, competitors, and even across industries.  The genius lays in its simplicity, it can be explained in about 2 sentences to busy executives and worker bees alike.  No complex multi-layered attribute models and weighted indices, it’s just plain old “would recommend” with a twist of lemon.  Delicious!


In the harm column, it is arguably one of the most confused concepts in Customer Experience today.  No, it is not the best predictor of business performance, and I don’t ever believe Dr. Reichheld ever exactly said that it was.  No, it is usually not the best measure to incentivize your divisions, departments, stores, or employees.  No, it is not enough to measure one question to understand the customer voice.  As one of my analysts posted, perhaps a bit harshly, on Fred’s blog “Dr. Riechheld, I give you idea an 4, based on your logic, that’s all you need to know to improve.”

However, I think the biggest unforeseen and unintended consequence was that business people started focusing on the number rather than what to do about it.  I cringe any time I hear about how some company “needs to improve NPS”.  No you don’t need to improve NPS, you need to improve the customer experience and order to do so, you have to do something.

This paucity of doing something is perhaps why we have seen almost zero improvement in the last 20 years in the American Customer Satisfaction Index.  In quarter 3 of 1994 the ACSI was 74.8 and in that same quarter of 2015 we have achieved a score 73.8.  Now that’s progress!

From Indices to Actions

To be fair, I don’t think Dr. Reichheld ever foresaw or intended the warping and perversion that we have seen with NPS in practice in some instances. So much pressure on one tiny little number. Like many inventions, it was an innocent and good idea that went rogue.

NPS has also inspired a collection of new indices such as Customer Effort,  Net Emotional Value, and others.  That’s great in that it keeps the focus on the customer. Understanding customers’ thoughts and feeling are requisite for improving the experience, but there are no magic bullets.

Let’s spend 1% of our time worrying about the numbers and 99% doing something about it. Investing in experiential design from a customer first perspective will pay much larger multiples in business outcomes than measurement.

Let’s design online and in store experiences that make customers want to come back. Let’s focus and hiring and training the best people we can and give them the right tools to create a great experience.  Finally, let’s invite customers to help co-create that experience at scale. This is where we will get the true recurring dividends out of CX, not rejiggering the metrics.

What do you think?


[1] I always felt bad about those lonely ignored “Passives”

On Death Metal, Chatbots, and CX

If someone traveled in time from 1956 to visit us today, they would be convinced the world had contracted a virulent form of contagious schizophrenia.  Everyone is talking to objects rather than people.  Today we talk with our cars, hand held devices, and even appliances. What would send our 1950s fedora festooned time traveler over the edge?  When they talk back.

Bots are everywhere. My wife and I have taken to having our in-home Amazon Echo bot named Alexa play white noise to help us sleep at night (our “dog” snores). Oftentimes, Alexa doesn’t quite get it right…

“Alexa, play rain sounds”

“Playing Reign in Blood by Slayer”

Nevertheless, the robots are clearly here to stay and will grow in prominence. In a research study conducted by Oracle of 800 Marketing and Sales Executives across the globe, 80% said that they are already using them or plan to do so by 2020. The application of “chatbots”, or partially autonomous dialoging assistants sits clearly in the crosshairs of customer service, but also has application in sales and marketing as well as other routine tasks.


Source: Chatbots Magazine

Notwithstanding their foibles, chatbots can be used to great effect to create a great customer experience. For example, I recently had to have my windshield replaced due to an errant roadway rock and so submitted a claim with Farmers Insurance. It was painless. I called the 800 number and got a bot lady. The bot lady texted me a web link and walked me through the claim all on my mobile phone while I was eating lunch, including scheduling a replacement appointment. It took 5 minutes. The next day I got a text notification that the technician was on his way a few minutes before the scheduled time. He came to my house, replaced my windshield, and it was all done painlessly in 20 minutes. In my opinion it was actually much better than dealing with a human.

“Alexa, play rain sounds”

“Playing songs by Lil’ Wayne”

AI to the Rescue

Chatbots of yore used to be rule based: that is, it would look for words and phrases and then have pre-programmed responses. Many reservation and IVR systems still function this way. If you have ever got caught in one those phone mazes of customer support, you know they can be very frustrating.

However, nowadays bots are getting much more sophisticated via artificial intelligence. IBM Watsontends to be one of the first choices in building chatbots, followed and Microsoft according to a study by Mindbowser. Cleverbot is another famous AI chatbot that you can go out and talk with right now. Created by Rollo Carpenter, Cleverbot is constantly learning with more than 4 million interactions per second. The engine behind Cleverbot and an API for accessing it is available for developers via

“Alexa, play rain sounds”

“Playing Ring of Fire by Jonny Cash”

Not only are chatbots becoming much smarter, they are also can be customized. Imagine Mickey Mouse calling to confirm your stay at Disney World or Tom Boddet following up about your stay at a Motel 6. The folks at use something called Perfect Pitch Technology to customize the outbound or inbound voice to your brand’s needs in a chatbot context. Some of these chatbots are so sophisticated they are hard to discern from real people. Eventually, I expect we will able to configure personality traits; a snippy French chatbot for an exclusive restaurant, a goofy friendly chatbot for an amusement park, a deeply empathetic and reassuring chatbot for insurance claims…you get the idea.

“Alexa, play rain sounds”

“Playing It’s Raining Men by The Weather Girls”

Reduce Need for Surveys

Increasingly those in the Customer Feedback Management (CFM) space are starting to look very closely and adopting chatbots to help supplement or even replace email, phone, or for the old school…mail as a way to get feedback. Startup AtlasRTX has been active in the homebuilders’ community where they are not only using real time dialoging to collect customer sentiment, but they are blurring the lines between marketing and retention. Starting early on in the customer journey, they interact with customers from interest to purchase to post-purchase.

It’s not research, it’s not marketing; it is engagement…with the bonus of providing useful and actionable data for marketing, sales, and insights groups.

Software provider Wizu provides a SaaS solution in using chatbots to collect customer feedback. It has a fairly simple self-serve interface and pricing model to customize and deploy your very own chatbot quickly. CFM pugilist iSky, who focuses on the automotive vertical, developed something called “ Valet” which is integrated with their text analytics engine to create real time two-way dialog with customers. Finally, if you have the time and energy you too can build a rudimentary chatbot from scratch. It takes about $1 and 10 minutes to do so.

“Alexa play rain sounds”

“Playing November Rain by Gun and Roses”

Alas, chatbots aren’t perfect. As any one screaming into Siri or attempting to get Alexa to play simple “rain sounds” can attest. Microsoft’s AI chatbot “Tay” began spewing anti-Semitic and sexist tweets in less than 24 hours after it launched to which Microsoft released a terse and brief apology and promptly unplugged Tay.

Nonetheless, with millennial preference for SMS, Snapchat and other social media messaging device coupled by decreasing attention spans, it is clear “dialoging” will supplant the “survey” in capturing customer feedback. Additionally, customers are increasingly expecting to get something in return for their time.

Combating Shiny Object Syndrome

While the bot revolution is exciting, I think we need to take measures to guard against an epidemic of SOS (Shiny Object Syndrome). We need to make sure we thoughtfully integrate these new bot technologies into an overall customer experience. For example, Alexa is great in the normal quiet home. Voice recognition systems via phone can be quiet irritating when you are in a crowded and noisy restaurant or airport. Think about the context of use when selecting preferred modalities for communication.

Finally, technology should be used to improve efficiency and create a better experience; but only if it can do both. Abuse of chatbot technology in the spirit of “cost savings” will just make your customers angry. Using expensive technology installations when a person, website, or some other form of simplified communication will do is preferred. Apply Occam’s razor to your CX problems. Simpler is always better.

If you are considering using bots in your organization their strike zone appears to be in routine high volume work where some degree of problem solving is required. Tier 1 inbound call handling is prime candidate as is using bots for a replacement for the old fashion post-call “robot lady” IVR. I am also intrigued at the possibility of its use in outbound close loop systems and gathering basic VOC feedback as part of a help system. If it can be done right, it has the potential to further reduce data collection costs, improve quality, and help customers.

“Alexa play ocean sounds”

“Playing ocean sounds, by ocean sounds”

With the advances in text analytics and AI, collecting customer feedback in the form of a dialog while helping customers only makes sense. We are going to see a lot more robots helping us in our day-to-day. I think we are also going to see email as a methodology give way to messaging in the same way that email supplanted phone (and phone supplanted mail). It won’t happen overnight, but it will happen and faster than you might think. In the interim, my wife and I have switch to listening to “ocean sounds” which Alexa apparently finds much more understandable and less disruptive than the accidental death metal tune before bedtime.