Metrics that Matter
When I was leading digital initiatives at Arrow Electronics, we set out to launch our first customer-facing chatbot. My instinct was to go big. I wanted to design sophisticated RMA workflows—validating product eligibility, checking component types and date codes, routing cases into external systems, and even automating approvals. I envisioned pulling in card and order history to suggest alternative components, cross-sell, and upsell.
But before building, we stopped and asked a simple question: what do customers actually need most?
We analyzed website behavior and support ticket data. What we found was eye-opening: the complex use cases I had in mind accounted for less than 3% of customer inquiries. In contrast, 65% of all requests were some version of “Where’s my order?” with the rest focused on shipping, locations, and delivery details.
By focusing on metrics that mattered, we reshaped the development roadmap. Instead of building the most technically impressive bot, we built the one customers truly needed.
The result?
60% reduction in development time by keeping solutions simple and connected directly to the ERP.
40% fewer chats handled by live agents, freeing them to focus on high-complexity issues.
Customer satisfaction soared as order status answers were delivered in seconds instead of hours or days.
125% faster resolution times for escalated cases because agents could now dedicate time where it counted most.
Ironically, the hardest part of the project wasn’t the integration—it was naming the bot. Corporate marketing weighed in on everything from tone to gender identity, proving that politics can stretch even the simplest decision into weeks of debate. A story for another time.
The real lesson? As technology leaders, it’s tempting to chase what’s possible. But “build it and they will come” is a myth. By leveraging business data and customer behavior, you can deliver solutions that drive measurable impact with less complexity, lower cost, and faster results.
