Choosing a Commercial Energy Submeter in 2026: A Practitioner's Decision Tree

TL;DR

Start with accuracy class: ANSI C12.20 0.5 (revenue) vs 1.0 (sub-billing) vs 2.0 (informational). Most retrofit work needs class 1.0; tenant billing needs class 0.5.
Communications protocol is usually the second filter: Modbus RTU/TCP for new installs, BACnet IP for integration into existing BMS, pulse output for legacy retrofit, cellular/LoRaWAN for sites without IT cooperation.
Beyond the meter price, the installed cost (CTs, enclosure, breaker space, commissioning) is typically 2–4× the meter itself. Pick the platform that minimizes that, not the cheapest meter.
eGauge wins for distributed sites and Eenovators-style deployments with onboard analytics. DENT and Accuenergy win for tight integration with existing BMS. Schneider and Leviton win for revenue-grade tenant billing.
If you're going to feed the data to AI analytics, your point-list governance and tag quality matter more than which meter you bought.

The decision tree

We use this sequence on every commercial site:

1. What accuracy class is required?

Tenant sub-billing or regulatory: class 0.5 (ANSI C12.20 / IEC 0.5S).
Energy management, BPS benchmarking, load profiling: class 1.0.
Verification of automation actions, troubleshooting: class 2.0 OK.

This single answer eliminates roughly half the product catalog before you start.

2. What communications protocol is the data going into?

Existing BACnet BMS: BACnet IP or BACnet MS/TP.
New monitoring platform you control: Modbus TCP (cleanest), Modbus RTU (cheapest for distributed runs).
No reliable building network: cellular (per-meter or via a gateway) or LoRaWAN.
Legacy retrofit, simplest possible: pulse output to a data logger.

3. How many circuits, in what density?

One or two large circuits per panel: single-point meter.
Many circuits in the same panel (most commercial main switchboards and tenant panels): multi-channel meter (eGauge EG30xx with 30+ channels, Accuenergy AcuRev 2000 with up to 48 channels, DENT PowerScout with various channel counts).
Multiple panels per location, geographically distributed: gateway-based architecture with cellular or LoRaWAN.

4. What CT type fits the installation?

Retrofit, no shutdown allowed: split-core CTs.
New panel build, conductor can be threaded: solid-core CTs (more accurate).
Large conductors (>400A primary), VFD presence: Rogowski coils.

5. Who is going to look at the data?

An engineer with a Modbus client: any meter.
A facility manager: meter with a usable native dashboard (eGauge, DENT) or routed to a platform the FM already uses.
An AI analytics layer: any meter, but point-list governance becomes critical.

When we pick which vendor

A simplification, but useful as a starting point:

Vendor	Sweet spot
eGauge	Distributed sites, multi-channel needs, owner-controlled platform, sites where the local LAN/IT cooperation is uncertain. Onboard logging + cellular options + decent native dashboard. The default for most Eenovators deployments.
DENT (PowerScout)	Mid-size commercial with existing BMS integration via Modbus or BACnet. Good multi-channel options.
Accuenergy (AcuRev)	High channel-density panels (up to 48 channels in one enclosure). Strong choice for tenant sub-billing where many small tenants share a panel.
Setra Power Meter	Tight integration with the broader Setra environmental sensing line. Strong choice when ambient + electric metering are being deployed together.
Schneider Electric (PowerLogic)	Revenue-grade applications, large commercial/industrial, when Schneider gear is already on-site.
Leviton VerifEye	Multi-tenant residential and commercial sub-billing where the per-unit cost has to be tight and integration into Leviton infrastructure already exists.
Cellular standalone meters (various)	Single-point monitoring on sites with no network. Compromise on flexibility but unbeatable for getting data flowing without IT engagement.

The mistakes we’ve made

Buying meter on price, paying it back on commissioning. The cheap meter often needs more enclosure work, more wiring, more commissioning time. A $400 meter with a $1,400 commissioning bill is more expensive than a $700 meter with a $700 commissioning bill.

Skimping on the CT order. Backordered CTs have killed more than one schedule. Standard sizes (200A, 400A, 800A) are usually available; weird sizes (50A for branch monitoring) are six weeks out. Order the long-lead CTs first.

Underspecifying the communications cabling. A $30 cat6 run at install time is cheaper than a $400 conduit fix later. Always run two cables to each meter location.

Ignoring point-list governance until later. This is the AI-era version of the prior bullet. If your tags are inconsistent, your analytics — AI or otherwise — will be unreliable. Set the naming convention before the first meter ships. Document it. Audit it quarterly.

The Africa-specific case

Many of our deployments in Kenya, Uganda, Tanzania, and Rwanda are in industrial and agricultural facilities where the building network is patchy, IT cooperation is informal, and the project owner is the operations director who just wants data on their phone tomorrow.

For these we default to eGauge + cellular. Onboard logging covers network outages. The native dashboard is good enough for the FM. We route data to Eagles Portal for analytics, and that’s done before we leave the site.

The Colorado pattern is different — buildings have BMS, IT cares about device policies, and BACnet integration is often part of the scope — but the discipline of ‘get the data flowing reliably first, optimize later’ is the same.

What changes for AI-readiness

If you’re going to put AI on top of your metering data (and you should), three things matter more than which meter you bought:

Tag governance. Consistent naming, units, hierarchy. We document this in a per-site point-list spreadsheet, version-controlled.
Unit annotation. Every point has its unit explicit, every time. The model on top will not infer it correctly when you have liters and gallons in the same dataset.
Data quality alarms. When a meter stops reporting, the AI shouldn’t be guessing. The pipeline should flag the gap before the model has to invent through it.

We’ve written more on this in our AI in Energy bucket. Start there if you’re building this stack.

Quick picks for common situations

Small commercial site, single panel, sub-billing requirement: Accuenergy or Leviton.
Mid-sized office, no existing BMS, want AI analytics: eGauge + cloud platform.
Distributed retail or agricultural sites: eGauge cellular.
Industrial plant with existing Modbus network: DENT or Accuenergy.
Multi-tenant commercial with revenue billing: Schneider or Leviton.

Sources

ANSI C12.20-2015 ‘Electricity Meters — 0.1, 0.2, and 0.5 Accuracy Classes’
IEC 62053-22 (analogous international standard)
eGauge, DENT, Accuenergy, Setra, Schneider, Leviton current product documentation
Eenovators internal deployment notes across ~80 sites in Kenya, Uganda, Tanzania, and the US since 2018

Frequently asked questions

What accuracy class do I actually need?

ANSI C12.20 class 0.5 (or IEC 0.5S) is the revenue-grade standard required for tenant sub-billing in most US jurisdictions. Class 1.0 is sufficient for energy management, load profiling, and BPS benchmarking — the typical accuracy of utility revenue meters themselves. Class 2.0 is fine for informational metering but won't satisfy regulatory or billing use cases.

Modbus RTU vs Modbus TCP vs BACnet?

Modbus RTU runs over RS-485 serial — cheap, robust, but adds a wiring infrastructure layer. Modbus TCP runs over Ethernet — simpler for IT-friendly sites. BACnet IP is the right answer when integrating into an existing BACnet-based BMS where the building operator wants the data in their existing platform. Most modern meters support multiple protocols.

When do I need cellular or LoRaWAN?

When the building's IT department won't give you a network drop, or when the site is distributed across locations without unified networking (typical for retail chains, agricultural sites, or African industrial campuses with patchy network coverage).

Split-core CTs vs solid-core CTs vs Rogowski coils?

Split-core CTs are the standard for retrofit (clip around an existing conductor, no shutdown needed). Solid-core CTs are more accurate for new installations where the conductor can be threaded through. Rogowski coils are ideal for large conductors and variable-frequency drives because they don't saturate.

How do I pick CTs sized correctly?

Match CT primary rating to the breaker size, not the actual load. A 200A breaker should get 200A CTs (or larger if expansion is planned). Undersized CTs saturate; oversized CTs lose low-end accuracy but rarely cause measurement problems.

What's the typical installed cost per meter point?

$600–$2,500 per metered circuit, depending on accuracy class, communications, conductor size, panel access, and whether you can shut down for installation. Multi-channel meters (8–16 channels in one unit) significantly reduce per-point cost where multiple circuits live in the same panel.

#monitoring#hardware#field-note#africa#colorado

About the author

Chris Mbori

Founder of Eenovators Limited (East African ESCO), partnering with AIM Dynamics. Built Eagles and the ADM portal. AEE Energy Manager of the Year (Sub-Saharan Africa). 10 AEE certifications. Licensed Engineer. Field journal — hype-skeptical, field-tested.