This is a real account of CSAT improvement for an e-commerce company in India, from 38% to 79% in 87 days. No new technology. No new headcount. Just four fixes were applied simultaneously.
I need to tell you about a phone call I received on a Friday evening.
It was from the VP of Customer Experience at a high-volume e-commerce company. Fifteen thousand customer contacts a day. Chat, email, phone, social media. CSAT had flatlined at 38%. Trustpilot scores were dropping. App store reviews were filled with one-star rants. The best agents were leaving.
“We’ve tried everything,” she said. “New software, new scripts, new KPIs. Nothing’s working.”
I asked her one question: “When was the last time you sat next to an agent and listened to ten calls in a row?”
Silence.
So, that’s where we started. Consequently, what we found in that first week of listening is something I think most businesses would find if they stopped looking at dashboards and started listening to their own customers.
The Challenge: A Contact Centre Optimized for the Wrong Things
We spent the first week observing, sitting next to agents. Listening to calls, reading transcripts. And four problems emerged, stacked on top of each other like geological layers, each one making the others worse.
Problem 1: Speed was king, resolution was an afterthought. Average handle time was the primary metric. Agents who closed chats in three minutes got green flags. Agents who took twelve minutes because they actually helped. Flagged for review. The incentive system was actively punishing good service.
Problem 2: Furthermore, Empathy was a poster on the wall, not a skill in the room. Nobody had trained agents to read emotional cues in text, to distinguish between a process complaint and a trust violation, or to write apologies that sounded human instead of corporate.
Problem 3: Additionally, AI was fighting the customer, not helping the agent. The chatbot handled FAQs well but was also trying to handle complex, emotional complaints through the same scripted flow. It was solving problems customers didn’t have and ignoring the ones they did.
Problem 4: Moreover, Supervisors were policing, not coaching. QA meant checking whether agents followed the script. You could score 95% on compliance and still leave a furious customer on the other end. The checklist measured the process. The customer measured care.
Each of these four problems was a direct barrier to CSAT improvement. For an e-commerce operation in India handling 15,000 contacts a day, fixing one without the others would have changed nothing.
The Task: A Step Change, Not Incremental Improvement
The VP wanted CSAT doubled. From 38% to 75%+ in 90 days. Our response was honest: it would require simultaneous changes in metrics, skills, technology, and leadership. All four levers. You can’t fix a system by tweaking one part of it.
We’ve documented a similar transformation in our case study – From Chaos to Clarity: How We Boosted CSAT by 30% and Cut Costs by 23%.
The Action: Four Fixes, One Transformation
Fix 1: Redefining “Success”
We retired AHT as the primary metric. Replaced it with a composite: First Contact Resolution + Customer Effort Score + CSAT. Handle time became secondary. Within days, agent behavior shifted. When you stop penalizing people for taking time to help, they start taking time to help.
Try this: Look at your primary metric right now. If it’s AHT, ask: Are you optimizing for your operation or for your customer?
Fix 2: Teaching Empathy as a Skill
We built a 16-hour program around the ten most common complaint scenarios, using real verbatims from their own customers. No generic roles play. Real words from real unhappy people.
Before training: “We apologize for the inconvenience caused. Your request has been forwarded to the relevant team.”
After training: “I’m really sorry this happened, Meera. I can see the package has been stuck in transit for four days. I’ve flagged this as urgent, and you’ll have an update from me by 5 PM today. Here’s my direct email.”
The same empathy skills that transform escalation calls apply here. We covered the five moments that matter in Escalation Handling Training India.
Same issue. Same resolution. Completely different experience. The first is a company talking to a customer. The second is a person talking to a person.
Try this: Pull your last 50 complaint transcripts. Count how many times the agent’s first response is a process step (“Let me check…”) versus a human step (“I understand…”). That ratio is your empathy score.
This language shift is the single most underrated driver of CSAT improvement in e-commerce India contact centres, and it costs nothing except training time.
Harvard Business Review research consistently shows that reducing customer effort, not delighting them, is the strongest driver of loyalty and repeat purchase.
Fix 3: Making AI Smarter, Not Louder
We reconfigured the chatbot from “try to handle everything” to “smart triage.” Simple queries stayed with the bot. Emotional or complex queries were immediately routed to a trained human, with full context. The bot’s job changed from gatekeeper to guide. As a result, its own CSAT improved from 29% to 61% because it stopped trying to do things it shouldn’t.
Try this: Check your chatbot’s CSAT specifically for complex or emotional queries. If it’s below 30%, the bot isn’t helping your customers. It’s a wall between them and someone who could.
Fix 4: Coaching, Not Policing
We trained team leads in a coaching methodology. Weekly one-on-ones replaced monthly scorecards. The discussion question changed from “Why did that call take 8 minutes?” to “What did the customer actually need, and did we get there?”
Therefore, this shifted the culture from compliance to curiosity. Agent engagement scores rose alongside customer satisfaction. It turns out, when people feel supported rather than surveilled, they do better work.
Try this: This week, pick two calls. Sit with an agent. Don’t grade them. Ask them what they think the customer needed. That conversation is worth more than a month of QA scores.
The Result: CSAT Improvement Through System Design
CSAT went from 38% to 79% in 87 days.
This is what structured CSAT improvement looks like for e-commerce businesses in India when all four levers move together.
First Contact Resolution: 31% → 68%.
Agent attrition: Down 30%.
Lower attrition and higher CSAT are two sides of the same coin, something we explored in depth in Client Retention Is the New Acquisition.
Operational costs: Down 23% as repeat contacts dropped.
Chatbot CSAT: 29% → 61%.
The VP now presents this at conferences. She says something I think every CX leader needs to hear:
- We didn’t buy new technology.
- We didn’t hire more people.
- We just finally trained the people we had and gave them permission to be human.
CSAT isn’t a score. It’s a system. And systems can be redesigned. Start by sitting next to your team and listening. Everything else follows from there.
Struggling with CSAT? We’ve been in the trenches. Let’s talk: excellential.com/contact-us
Pooja Singh has spent over two decades in the middle of one of the most human things in business, figuring out how people and organizations can work better together.
She co-founded Excellential Consulting Services in 2015 with a straightforward belief: that good HR isn’t a department function, it’s a business strategy. Since then, she has partnered with startups, SMEs, and large enterprises across India on talent acquisition, leadership development, and talent management, often stepping in as the extended HR team that growing organizations need but don’t yet have.
Her work has taken her across industries, e-commerce, BFSI, manufacturing, quick commerce, IT, consumer durables, and FMCG, and her writing on this blog draws directly from those experiences.
No borrowed frameworks. No buzzwords. Just honest observations from the field.
She is based in Bengaluru, consults with several unicorn startups, and runs Excellential with her seasoned team. She’ll tell you, it keeps her sharp, hungry, and close to what actually matters.
