Planning a successful AMI rollout

Most AMI programs don't struggle with the technology. They struggle with the data underneath it, the customers on the other side of the door, and the hundred small decisions nobody assigned an owner. Here's what we've learned across metering deployments.

Start with the data, not the meters

The single best predictor of a smooth deployment is the quality of the starting asset record. Before the first install, reconcile the billing system, the meter inventory, and the GIS layer against each other. Expect discrepancies: meters that were replaced but never updated, accounts with no meter of record, duplicate serials, sizes that don't match what's in the pit.

Every unresolved discrepancy becomes a field exception, and field exceptions are the most expensive minutes in the program. Cleaning them at a desk beforehand is far cheaper than resolving them with a crew standing in a driveway.

Sequence the route, not just the schedule

Deployment velocity comes from routing density, not crew count. Group work by geography and access type, and separate the straightforward exchanges from the known problem sets — pit meters requiring confined-space procedures, inside sets requiring an appointment, large commercial meters requiring coordination and a scheduled shutdown. Mixing them into one queue drags the average install time toward the hardest case.

Customer communication is part of the install

Two things drive callbacks: a customer who didn't know anyone was coming, and a bill that changed after the swap. Both are solvable.

• Notify ahead by mail, then again close to the visit by the channel the customer prefers.
• Give people a way to schedule inside appointments themselves, and staff a call center that can actually book.
• Explain in advance why consumption may read differently — an aging meter under-registers, so accurate reads often look like an increase.
• Leave documentation at the property with the old and new register readings.

Validate reads before you trust them

Build a validation step into the install itself: photograph the old register and the new one, capture GPS, confirm the serial scan against the work order, and flag mismatches before the crew leaves. Then validate again at the system level — first transmission received, reads landing in the head-end, reads flowing into billing. A meter installed but not communicating is not a completed install, and the gap between those two definitions is where deployment reporting loses credibility.

Assign ownership of the exception queue

Every program generates them: no access, meter not found, unsafe condition, non-standard setting, damaged service line, customer refusal. Decide up front who works that queue daily, what the escalation path is, and what "closed" means for each type. Exceptions left to accumulate become the long tail that stretches a nine-month program into eighteen.

Measure a few things well

Installs per crew-day, first-visit completion rate, exception aging, percentage of installed meters transmitting, and callback rate per thousand installs will tell you almost everything about program health. Track them daily from the start rather than building reporting after problems appear.

How we work

VEPO Metering delivers deployments end to end — field crews, an in-house call center for appointment scheduling and customer questions, a utility portal for live progress, data validation, and billing integration. Our field teams run on VPie, so the work order, the photos, the GPS, and the meter data all move through one system instead of three.

Planning a rollout, or mid-deployment and looking for a second opinion? Reach out and we'll set up a short call.