Mapping the gap between customers and self-service across the utility industry, and building the research infrastructure to close it.
Because the design fails them mid-task. Across the utilities I benchmarked, customers who want to check a bill, report an outage, or adjust a payment are starting on digital channels. They click into the portal. They don't finish. The website routes them to a PDF. The transaction times out. The outage map doesn't show their street. So they call.
The features exist at most utilities. A utility can have bill pay, outage reporting, and account management all online, and still drive customers to the phone every time. What separates a channel customers use from one they abandon is whether they can find what they need and trust that it will work. Customers who fail once don't try again. That's the part that stuck with me.
Live-agent calls cost significantly more than self-service interactions. Contact centers absorb this overflow, which leaves less capacity for the calls that actually require a human. The math is simple. The design work to shift it is not.
"If a customer can't resolve an issue via another channel, they'll likely follow up by phone. Inefficiency on one channel negatively affects CSAT across all of them."
By expanding what counts as a signal. Survey response rates have dropped but customers aren't going silent. Call recordings, app store reviews, social comments, contact center logs. Most utilities already have this data. Few treat it as VOC.
The three signal types below (unsolicited, solicited, and operational) form a listening program that doesn't depend on any single response rate.
"Customers aren't going silent. They're using other channels to share their thoughts, and utilities already have access to that data."
Four groups, and most self-service strategies are written for only one. When I mapped who's affected by a single failed self-service interaction, I found: the customer who spent ten minutes online before calling, the agent now handling a call that shouldn't exist, the utility absorbing the cost of both, and the property manager who gave up on the portal and calls for every service request. Each group has a different definition of success.
At every stage of the customer journey:
These are the core interactions, not edge cases. At each stage, most utilities have a self-service option. At each stage, most customers call anyway.
I looked for patterns in where the calls were coming from. Customers were failing in the same places across different utilities: couldn't find the right page, the mobile flow broke mid-task, no confirmation that the action went through. Click any stage below to see where friction concentrates and what utilities that got it right did differently.
With Customer Effort Score, a metric that asks how hard customers had to work, not just whether they're satisfied. A utility with a low CSAT score on billing calls can't tell from that number where the frustration came from: the bill design, the IVR, the wait time, the agent. CSAT and NPS measure the outcome. They don't locate the friction. That distinction matters when you're deciding what to fix.
Customer Effort Score changes that. CES asks customers how hard they had to work to complete a task. A high effort score on outage reporting points to a specific journey, which gives you somewhere to look. That's what makes it useful.
About 60% of utilities now measure CES. Most of them don't run driver analysis. They know effort is high. They don't know where, or for which customers, or at which stage. Measuring without analyzing is how you end up with a metric that doesn't move anything.
"Focusing on lowering effort scores helped Con Edison's digital team prioritize what to improve on their website and mobile app."
That offering a feature and designing it well are not the same thing. Across nine reports and the E Source Website Benchmark, which evaluated roughly 100 US and Canadian utility websites, the same pattern appeared every time. Utilities with high self-service adoption didn't have more features. They had features that worked when customers tried to use them, and customers could find them without calling first to ask where they were.
Three threads I'd give more time to. Each one kept surfacing as more consequential than the scope let me pursue. Nine reports, 93 utilities, the full lifecycle. Broad coverage means some findings get named but not fully resolved, the kind of thing you flag in a conclusion and wish you'd given a whole report to.