Transforming Support: A CSAT Story Part 3

This article concludes my 3-part story of CSAT Transformation.

You can read Part 1 and Part 2 here.

Path to 100

Early on in our journey, we would have open ended discussions about improvements. I would ask questions like – what will it take to get to 60 (from 30) in one quarter?

This created an anchoring effect, and the plans I would get would be for 10-20 points improvement. These were never satisfactory, because they were based on the thought ‘what can we improve’. Rather than ‘what’s the problem’.

At some point, I realised that this time-bound target approach was at fault. It was focussing on action rather than diagnosis. It needed to come after diagnosis. So I switched to a different question:

What will it take to get to CSAT 100?

(This is based on one of my core tenets of working- What will it take to get to a perfect score? Will write about this tenet separately.)

Of course, early on, this was met with smirks and laughter.

100? Are you joking? No one in the world is at 100. Amex is at 92, with way better processes, technology, and agents that are paid 3 times of what our agents are paid.

Et cetera. Et cetera.

But I persisted. The idea behind talking about the path to 100 was not that we needed to get to 100. It was to figure out why we cannot get there.

That helped change how we were looking at the problem completely. Now we had to go deep, and quantify everything.

This approach also helped change the quality of our conversations. Whatever root causes the team figured out solves for - they could own and execute. We didn’t need a large group to come together. Whatever they put under the bucket Can’t be solved – that’s where would spend time on as a group. To brainstorm and figure out if there was something we could actually do.

Here’s an example chart we drew:

This issue-level examination and the Path to 100 approach led to some interesting insights.

With clear quantification of the preventable, not resolvable category, our conversations with business and upstream product teams became more targeted. Fixing these types of issues upstream was higher priority than other issue types.

(In some of these cases, we could do a bit of mitigation, like refund for an AC not working, but that would help only a little bit. And there are situations where there's absolutely nothing you can do after the event - like the stuck UPI Payment)

Some examples of these issues:

• A UPI payment stuck with unknown status
• Hotel denying check-in to the customer
• Missed flight due to cab driver cancelling
• A car ride with the AC broken
• Dispute between merchant and customer, or between driver and passenger

Some interesting stories

During our process-focussed deep dives, specially on the 'preventable, not solvable' bucket, we came across lots of a-ha moments. Here are a couple of them.

Story 1: I need a booking

One interesting issue was at Ola – drivers would call us, requesting for bookings. These drivers would call after they had been sitting idle waiting quite a while for a booking. There was absolutely nothing the call center agents could do about it. (And the prevailing thought was – we should stop these calls at the IVR itself- tell the driver we can’t help them.)

You’d expect the CSAT score on this issue to be extremely poor. Maybe zero. But no.

CSAT score on this one was almost the same as the overall CSAT score.

This made no sense, till we spoke to the drivers, and listened to both the positive CSAT and negative CSAT calls very carefully. When we spoke to drivers, we realised that the drivers understood the demand supply mechanics pretty well.

What they were really worried about was whether their system was working well – Was their app connected to the internet? Did the ola servers know they were online? Was ola aware of the correct GPS location? Etc. And during the call, agents would sometime go check these things, which assured the drivers, and all was good.

(Agents were doing this without a defined process – they had learned from each other, and were doing this on their own! One example of empowering agents to focus on the customer). ?

So what looked like a ‘Please give me a booking’ problem, was really a ‘I just want to make sure everything’s fine’ problem. Based on this insight, we worked with the driver app product and tech teams to build a health-indicator and diagnostic solution in the driver app, which would solve the problem and reduce these calls as well.

Story 2: Cancellation fees are too high

This one was more difficult to solve. Because at the surface level, the answer seemed obvious.

On flight cancellation requests, our CSAT scores were low. The team quickly discovered that it was low because of ‘non-refundable’ fares. (On most fares, airlines applied a cancellation penalty, which was pretty high- Rs. 3500 when the average fare itself was 5000.) ?The initial conclusion was “Obviously, customers will be unhappy because we are not refunding money”.

That did not pass the smell test for me. It just didn’t seem right.

If a customer with a non-refundable booking came to us with a refund request, it would be a request, not a demand. Why would so many customers, on being told the request couldn’t be accepted, (along with a reasonable explanation for it), be angry enough to respond to the CSAT survey? With a negative response?

If you’re making a request, you’re happy if it gets through. Else, you shrug your shoulders and move on.

Right?

Right? That's why this didn't add up for me.

Hold on. What’s the key assumption here? That the customers thought this was a request. That they were aware it was non-refundable. Maybe that's where we had a problem?

Initially, I got strong pushback – Of course they are aware! We tell them while booking. And we show it in the cancellation policies section. And we send it as part of the e-ticket email.

Let’s ask them, I said. When you call these customers to ask the reason, just ask them if they were aware. Also, for good measure, ask some customers immediately after they’ve made a booking- they won’t have any reason to lie, right? And let’s take a look at the actual booking screens.

And my suspicion was confirmed- customers were not noticing the refund policy when booking!

Technically, the information was there, but it was easy to miss. Unless the customer knew from experience to look for it, they would not notice it. (It was always one click away from the page they were at, or was part of a post-booking email – which was already too late, and who looks at that anyway till the date of travel?)

Once the team (and this is the CS team!) understood this, they worked with the product team, and we started highlighting the cancellation fees on the booking journey. As a result, complaints about the cancellation fee went down. And the CSAT on cancellation requests went up significantly. (And no, booking conversion rates did not go down!)

Other CSAT Initiatives

There are quite a few other things we tried. Below are some that are in early execution phase right now.

Catch them at entry

Traditional wisdom stated that CSAT would increase with agent tenure, as agents got more used to the domain and to the resolution processes.

We rejected this outright. Our postulate was that it should actually be the opposite! Our logic was based on our core principles, again. Two primary points:

• Tenure will help improve agent productivity, yes. But why should it improve CSAT? Newer agents will take more time to resolve a ticket, yes. They might need help a lot more. Their FTR will be lower, yes. But why will they be more likely to resolve a ticket incorrectly?
• Given the typical practices at contact centers, agents are more likely to get de-sensitized over time. Newer agents are more likely to try. Older agents have already given up.

In this model, new agents should have significantly higher CSAT!

Another interesting factor about contact centers is the very high attrition rate – most agents will have a tenure of less than a year, and if it’s a growing business, average tenure will be even less!

So, if we could just focus on new agents, CSAT would go up organically anyway.

Based on this thinking, we revamped our agent onboarding and training program. It became completely outside-in and customer experience first. In the first 2 days, we didn’t even talk about the company, the tools, or the processes. First two days were only about the customer, and the customer experience. What could go wrong? What things would customers need support with? How should we resolve those tickets? The training team would only ask these questions - the new agents would come up with the answers, based on the customer journey.

We wanted the new agents to understand that all we cared about was customer experience. And we needed to show evidence that we were serious. This approach to agent onboarding would help.

Along with this, we put in strong gating criteria – new agent performance had to beat the existing median performance. Over a period of time, this would improve the overall center performance.

Strong performance Management

We also started stronger performance management –15 day rolling performance plans, to weed out the bad performers. And governance of the plan was taken out of the hands of operations team. (Their day-to-day pressures about call answer levels and response times meant they were just not able to take the right calls.) ?

In parallel, we also started lucrative incentive schemes to reduce attrition amongst the high performers. Over time, with higher attrition amongst low performers, and lower attrition amongst high performers, the curve would tilt more towards the right, and the base would become narrower too.

(The effectiveness of the incentive scheme is an open question, though – if attrition is mostly due to non-financial factors, this wouldn’t make a difference.)

Channel of interaction

Data also showed us that synchronous interaction channels (voice, chat) had much higher CSAT then asynchronous ones (email, messaging).

That makes intuitive sense – it’s easier to get to the bottom of an issue if you’re having a real-time conversation with the customer.

Based on this, we built a product on our app, that would provide the customer the choice of channel, in case the automation solutions didn’t work for them. The choices offered were based on multiple factors – primarily the suitability of the issue type to the channel, but also the supply availability, and expected response times.

But the choice of agent channels is by itself an interesting conversation – will pick it up at another time.

One Final Thought

Through this journey, there was one aspect that came out loud and clear. Our support executives are the most important people in the support ecosystem.

As you've seen, the impact we were able to create was mostly because of them. But it's not just that - when you talk to them, they are full of deep insights. They are smart, they understand the user, and they understand the user's pain points.

You just need to engage them properly. Once they believe you're serious, they will open up. I did a few focus groups with our agents across centers. And I was pleasantly surprised. Strike that. I was astonished.

Unfortunately, they are considered the least important. No one talks to them. They are treated like robots, to follow the scripts to the T. There's institutional distrust of their capability and intelligence. We've put the entire QA ecosystem to monitor them and keep them in check.

They are the face of your brand. And increasingly across sectors, they are the only human interface with your customers. Empower them, don't control them.

Place them at the center of your support ecosystem, and think how you can help them. You'll be surprised at the insights they can give about the upstream issues and about automation. But talk to them directly, not through the contact center layer.

I hope you liked this multi-part article on CSAT transformation. If you did, take a look at some of my other articles:

Product Management and Problem Solving

The Power of Service Guarantees

Thoughts on Brand Experience