Most support dashboards measure the wrong things. After analysing 50M conversations, here are the four numbers that correlate with customer lifetime value.
When we analysed 50 million anonymised support conversations across our customer base, the metric that most strongly predicted customer retention 12 months out wasn't CSAT. It wasn't NPS. It wasn't even resolution rate.
The single biggest predictor was median first response time. Customers who got a reply within 60 seconds were 2.4× more likely to still be around 12 months later than customers who waited longer than 5 minutes — even when the resolution time was identical.
The signal customers care about isn't "you fixed my problem in 4 minutes vs 12." It's "someone is paying attention to me right now."
Once you're under a minute on first response, the next lever is one-touch resolution. Every follow-up message is a chance for the customer to lose patience or you to lose context.
AI helps two ways here: drafting a complete reply rather than a "let me check" reply, and surfacing the right knowledge-base article inline so agents don't context-switch.
This is the one most teams miss. We measured tone variance — the standard deviation of formality, warmth, and concision across the same brand's replies. Brands in the bottom quartile (very inconsistent) had churn rates 38% higher than brands in the top quartile.
Customers don't articulate this, but they feel it. The website chat that's chatty and the WhatsApp reply that's curt sound like two different companies. Tickki's voice doc anchors all channels to the same baseline.
The metric we didn't expect: workspaces where agents edit fewer than 15% of AI drafts have happier customers than workspaces where agents edit 40%+. Why? Because high edit rates usually mean the AI is wrong about voice or context, and the customer is reading the agent's frustration in the final reply.
If your edit rate is high, the fix isn't "use AI less." It's "give the AI better context — voice doc, recent CRM data, past conversation history."
If you do nothing else, swap CSAT-only dashboards for: median FRT, single-touch resolution %, tone variance, and AI edit rate. You'll have a much clearer view of whether your support is actually building loyalty or just resolving tickets.
Spin up a Tickki workspace in 5 minutes. No credit card needed.
AI drafts can sound generic and corporate. Here's the playbook our customers use to get drafts that sound like their best agent on her best day.
WhatsAppWhatsApp can be your highest-converting channel — or a black hole for support tickets. The difference comes down to five setup decisions you make on day one.
Getting StartedA walkthrough of every step a new Tickki customer takes — from sign-up to handling real customer messages — with realistic time estimates.