On 23 September in New York, the UN Secretary-General launched the Global Grids Accelerator with a sentence nobody in this sector would argue with: "Renewables are booming, but the grids that carry clean power are not keeping pace. Grids are the arteries of the energy transition." UNDP, UNOPS and Sustainable Energy for All will convene it. Its scope runs across "policy and planning, project preparation, finance and investment, and implementation", and in Africa it will build on the Continental Power System Master Plan, the African Single Electricity Market, the Programme for Infrastructure Development in Africa, Mission 300 and the National Energy Compacts. It "will not create a new financing institution or replace existing initiatives."
The numbers in the launch are about construction. More than 2,500 GW of renewable projects are "stalled in grid-connection queues worldwide". Annual grid investment, "now approximately $450 billion", needs to rise "by around 50 percent by 2030", figures the release attributes to IRENA. The IEA estimates that "more than 80 million kilometers of grids" must be added or refurbished by 2040.
Those numbers are the right place to start. My argument is that they are the wrong place to stop, and that the Accelerator, because it works at project preparation, is unusually well placed to fix the part it has not yet named. UNDP's Administrator, Alexander De Croo, said it in the same release: "Clean power only changes lives when it reaches people." Whether power reached people is a measurement made at a site, every day, after the ribbon is cut. Whether it is recorded at all is decided before construction, when a project's metering and data arrangements are specified.
Three notes. The launch quotations come from Sustainable Energy for All's release, which I fetched on 27 September; the Nigerian rules come from NERC orders and World Bank documents, listed at the end. Where I go beyond the documents, I say so. Gen318's commercial interest is stated at the end.
What the headline numbers count
Every figure in the launch counts something that exists on the day it is finished: gigawatts in a queue, dollars invested, kilometres built. Mission 300, one of the programmes the Accelerator will build on in Africa, reports the same way. On 16 June the World Bank and the African Development Bank announced that it "has connected over 50 million people to electricity across 40 countries", against a goal of 300 million by 2030.
A connection is a real achievement. It is also a single observation. The World Bank and SE4ALL published the Multi-Tier Framework in 2015 to replace a yes-or-no count of connections. A World Bank explainer note on the framework says why: a binary count "provides a simplistic and often distorted picture because it fails to determine how this access translates into reliable, affordable and modern service." It gives the example that decides the argument: if a grid connection "does not deliver on its potential, and the household only receives a few hours of service a day, this grid connection may be classified in an even lower tier than a solar home system." The framework measures duration from "at least 4 hours a day to over 23 hours a day", plus reliability, quality and affordability. None of those can be observed on commissioning day.
Nigeria's own Mission 300 compact states the gap in plain terms. Its diagnosis says DisCos "added 4.7 million 'legal' connections" between 2015 and 2023, "mostly attributed to new metering than actual new connections", and that "even Nigerians connected to the grid face frequent outages and do not receive adequate or reliable supply." It cites estimates putting power supply from generator sets at "10 times that from the grid (~4 GW)". NERC's report for the first quarter of 2026 records a total grid collapse on 23 January and a partial one on 27 January, and a weighted average ATC&C loss across the DisCos of 37.44%, against a target of 16.92%. That loss figure mixes technical, commercial and collection losses, so it describes the system's finances as much as its supply, but it describes the same system.
A kilometre of new line in that system tells you something was built. It does not tell you whether anyone at the end of it had power last Tuesday.
Nigeria already prices hours
The encouraging part is that Nigeria has already started tying money to delivered service at the level of a single feeder or site. Two regimes do it, one in force and one designed into a grant.
Band A. Since NERC introduced its service-based tariff on 1 November 2020, a Band A feeder has been one that delivers a minimum of 20 hours of supply a day, with tariffs aligned "in proportion to the service level enjoyed by customer clusters". In April 2024 Band A customers were moved to a separate, higher tariff. NERC's order for Abuja's DisCo, effective 6 May 2024, obliges it to "publish daily on its website a rolling seven-day average daily hours of supply on each Band A feeder", says the Commission will "directly obtain data on the hours of supply on each Band A feeder from the head-end system of AEDC for near real-time monitoring of service", and that where the DisCo "fails to meet the committed service level to a Band A feeder for seven (7) consecutive days, the feeder shall be automatically downgraded to the recorded level of supply". Compliance is "measured in average hours of supply per day over a period of one month."
The monthly test has been applied. NERC's July 2025 order for Port Harcourt's DisCo was made "based on the feeder performance report for the 1st – 25th of June 2025". Feeders that fell short were to be downgraded and customers compensated; feeders averaging between 18 and 20 hours kept their band but their customers were owed compensation. One line in the appendix reads: ITAM feeder, Uyo, 17.88 hours, "Downgrade To Band B and Compensate Customers accordingly". A tariff, a downgrade and compensation, all resting on a measured number of hours at one feeder.
The rule can also be relaxed, in the open. Vanguard reported in June 2026 on a NERC directive covering February and March 2026, whose shortfalls "were largely attributed to inadequate gas supply and vandalism of critical gas and transmission infrastructure, factors beyond the direct operational control of the DisCos." For feeders below 18 hours, "Affected Band A feeders will not be downgraded during the covered period; eligible customers will receive special compensation." Feeders between 18 and 20 hours stayed under the existing compensation framework. I have not found the directive itself, so I rely on the report. That is an excused outage: declared, bounded, and with its reason on the record, which is the separation argued for below.
DARES. The World Bank's $750 million mini-grid and off-grid programme pays isolated mini-grid developers per connection (interconnected mini-grids receive a percentage of capital cost), and not in one go. The project appraisal document says the grant "will be disbursed against two pre-defined milestones: commissioning and customer connections, and upon meeting a certain capacity utilization factor one year after commissioning." The second payment depends on how much electricity customers actually use a year later. The appraisal document is a design; I have not established whether any second-milestone payment has yet been made. The same document spells out what that requires: "more extensive data sharing will be necessary than has been the practice under the NEP. Smart meter integration with Odyssey will be required from mini grid projects in order to verify electricity consumption".
Nigeria's Mission 300 compact points the same way. Its Multi-Tier Framework access surveys "will be complemented for DREs by information collected through remote verification of smart meters and integration with CRM software", and it committed to "a data platform for M&E" by the first quarter of 2025. I have not checked whether that platform exists.
So in Nigeria, the Band A tariff is set on measured hours at the feeder, and part of the mini-grid grant is designed to be released on measured use at the site. Odyssey Energy Solutions, whose platform DARES requires mini-grid smart meters to integrate with, has said the same thing about the market as a whole. In August it wrote that the question funders ask "used to be 'how many systems did you deploy?' Increasingly, it is 'are those systems still running, are users still paying, and what quality of service is being delivered?'"
Why the record has to start at commissioning
There is a practical reason this belongs in project preparation rather than in a later monitoring phase.
A delivered-hours figure for a feeder or a site is a daily observation. Missing days cannot be recreated from a site visit, an annual survey or an asset register unless a device on site logged them, and whether one does is itself a project-preparation decision. Surveys measure things a meter cannot, such as affordability, legality and safety, and the Multi-Tier Framework rightly relies on them. But a survey can ask a household how many hours it usually gets; it cannot say how many hours it got on 14 March. Odyssey made the same distinction in August: "A commissioning snapshot can be collected manually: a site visit, a photo, a signed form. Sustained performance cannot."
The sector is not starting from a good place. The World Bank's 2022 mini-grid report found that "only about a third of vertically integrated Sub-Saharan utilities reported figures for the average duration and frequency of system interruptions in 2018, and only 5 of 21 distribution companies did so." In a box on mini-grid cost data, it said data collection on mini-grids "is at an early stage" and that "more effort should be spent on standardizing data collection and integrating data collection into reporting requirements into mini grid programs." The same report cites an average uptime of 99% for AMDA members' mini-grids in 2020. That figure comes from AMDA's benchmarking of its own members, not from an outside measurement; in my experience most uptime figures in this sector are still reported that way.
This is my argument rather than any of the documents': a record that starts late is not a shorter record, it is a different one. It begins after the teething faults, after the first battery problem, after the months that decide whether a site will last. It is the easiest period to leave out and the most informative one to keep.
What a usable site record looks like
"Measure delivered service" is easy to agree with and easy to do badly. Three properties separate a record that can carry a payment from one that only fills a dashboard.
It says how much of itself is missing. A site that stops reporting is not down and not up; it is unknown. If silent hours are counted as supply, the figure flatters; if counted as outage, it punishes a communications fault as a power fault. The record has to state, beside every hours or uptime figure, how much of the expected data actually arrived and how much of what arrived passed basic checks. Those are two different numbers. A site that sends one perfect reading an hour when it should send one a minute is 100% trusted and about 2% complete.
It keeps the raw figure and the excused figure apart. Every availability regime allows some outages to be excused: a grid failure upstream, a client's own fault, planned maintenance. Excusing them is legitimate. Folding them into a single number is not, because the reader can no longer see how much was excused, or on whose word. Both figures, side by side, let a regulator or a funder argue about the excuses rather than about the arithmetic.
It starts on commissioning day and is kept by default. The record should begin when the site is energised, as a condition of the first payment, and stay with the asset when ownership, operators or monitoring vendors change. NERC's Band A regime already assumes the DisCo's head-end system is the source; DARES already assumes smart-meter integration. Both are designed so that the data flows from the site without anyone having to remember to collect it.
None of this is exotic. The first property is what Band A and DARES already rely on for a narrow slice of Nigeria's electricity. The second and third are my additions; NERC's 2026 directive, which declared its excuse and its bounds, comes closest to the third.
What the Accelerator could ask for
The Accelerator says it will turn "government and regional priorities into investable, operational projects." The word that matters there is operational. A project is prepared once, but operated for twenty years, and the evidence of whether it worked is produced during those twenty years or not at all.
In the project-preparation workstream, three conditions would, I think, cost little and change what the programme can later say:
- Delivered service per feeder or site, from energisation, alongside kilometres and connections: hours of supply against the committed level, recorded daily and kept with the asset.
- Data completeness reported beside every service figure, so a silent site is reported as unknown, never as a zero or as 100%.
- Raw and excused availability as separate numbers, with the reason for each excused outage recorded against it.
These are written for distribution and access projects. For transmission lines and interconnectors, the equivalent is line and substation availability, reported with the same completeness figure and the same raw and excused split.
Nigeria has shown that the first can be written into a tariff order and a grant agreement. The Accelerator can carry the idea into projects that have not been designed yet, which is the only point at which it is cheap.
What this piece does not claim
- Measurement does not build grids. The construction gap the launch describes is real, and no amount of record-keeping closes it. This is about knowing what the construction delivers.
- The Band A and DARES rules are not proof that measurement improves outcomes. They show that Nigeria's regulator and a major funder treat measured service as a basis for money, one in force and one by design. Whether that has raised hours of supply is a separate question I have not tested.
- The 2,500 GW figure is the launch's own. The IEA's figures are different: at least 3,000 GW in queues in its 2023 report, of which 1,500 GW were in advanced stages, and 1,650 GW of solar and wind projects in advanced stages tracked in 2024. The release does not give a source for 2,500 GW.
- Nothing here is a view on the Accelerator's funding or governance, which the launch does not describe in detail.
Gen318's interest
- Gen318 builds software that keeps site records for solar, battery and generator sites, not DisCo feeders, including data completeness reported two ways and availability kept raw and excused. We have a commercial interest in this argument.
- Our own product only partly matches what this piece argues for. Gen318's availability figure counts silent periods against the site, the "as outage" reading in the diagram, and states the missing-data minutes beside it; it does not yet report a figure over known days only.
- We are early-stage.
- The three conditions above are stated so that any operator, DisCo, regulator or funder can apply them with their own systems. None of them requires Gen318.
One request. In the Accelerator's project-preparation work, add delivered service per feeder or site, recorded from energisation with its completeness stated, to the indicators each prepared distribution and access project must report.
Corrections to anything above are welcome and will be listed.
Sources
Fetched and read on 27 September 2026. NERC's orders and reports, the World Bank documents and the SEforALL release are primary; the Vanguard report is secondary and is labelled where it is used. Odyssey Energy Solutions, quoted twice, is the platform DARES requires mini-grid smart meters to integrate with; its blog is a vendor source.
- Global Grids Accelerator launch release (SEforALL, datelined New York, 23 September 2026) — launch quotations, conveners, scope and the 2,500 GW, $450 billion and 80 million km figures
- Electricity Grids and Secure Energy Transitions, executive summary (IEA, 17 October 2023) — 80 million km by 2040; at least 3,000 GW in queues, 1,500 GW in advanced stages
- Building the Future Transmission Grid, executive summary (IEA, 25 February 2025) — 1,650 GW in advanced stages tracked in 2024
- Mission 300 (World Bank) — the 300 million goal and the 16 June 2026 announcement of 50 million people connected
- Nigeria National Energy Compact under Mission 300 (World Bank, PDF) — connections, outages, generator supply, losses, remote verification and the M&E platform
- Beyond Connections: Energy Access Redefined, Introducing Multi-Tier Approach to Measuring Energy Access (World Bank/ESMAP explainer note, PDF) — the critique of connection counts and the duration attribute
- What is Service Based Tariff (SBT)? (NERC FAQ) — introduced 1 November 2020; Band A minimum of 20 hours
- NERC order ORDER/NERC/2024/046 for AEDC, effective 6 May 2024 (PDF) — Band A tariff change, daily feeder publication, head-end monitoring and automatic downgrade
- NERC order ORDER/NERC/2025/068 for PHED, effective 1 July 2025 (PDF) — feeder downgrades and compensation, including ITAM feeder
- NERC approves compensation for eligible Band A electricity customers (Vanguard, 5 June 2026) — secondary report of NERC directive NERC/2026/002: coverage period, 18-hour threshold, stated reasons and compensation
- NERC Quarterly Report 2026/Q1 (PDF) — grid disturbances of 23 and 27 January 2026; ATC&C loss of 37.44%
- Nigeria DARES Project Appraisal Document, P179687, 21 November 2023 (World Bank, PDF) — grant milestones and smart-meter verification
- Mini Grids for Half a Billion People (World Bank/ESMAP, 2022) — utility interruption reporting, mini-grid data collection and self-reported uptime
- Why funders are shifting from deployment numbers to long-term performance (Odyssey Energy Solutions, 14 August 2026)