Field notes: the valley that lit up two years before the completion report said it would

A river winding through a small village tucked into a steep mountain valley, with power lines visible on the slopes above.

I'm not going to name the valley. The project number is still live in our system, and the implementation unit on this one has enough on its plate without a blog post attached to its file. But the pattern is worth writing down, because I've now seen a version of it on three different programmes, and I suspect anyone who reviews grid-extension portfolios for a living has a version of their own.

The appraisal document put energization of the last tranche of settlements at year four. Year four came and went. The completion report, when it finally landed, confirmed energization at year six, with the usual paragraph about contractor delays, a rainy season that ran long, and a transformer that sat in a port warehouse for longer than anyone wanted to admit in writing. Nothing unusual there. Most of my files read like that.

What was unusual was going back through a monthly tile set I'd been pulling out of habit, mostly to see if it was worth the ten minutes a month, and finding that the valley's main settlement cluster had gone from scattered pinpricks to a solid, connected glow somewhere around month 41. Not month 48, not month 72. Month 41. Almost three years ahead of the completion report's own account of when the work finished.

I checked the disbursement schedule against the light trend and the two lined up in a way the narrative report never mentioned. The contractor had finished the physical extension early and quietly, then spent the better part of two years working through acceptance testing, a dispute over a sub-contract, and a local administrative sign-off that had nothing to do with whether the wires were live. The grid was extended. The paperwork saying so was not.

This is the gap I keep running into. A completion report tells you what happened to the contract. It doesn't always tell you what happened to the valley. The two are supposed to track each other, and on well-run projects they mostly do, but the lag between "the feeder line is energized" and "the file reflects that the feeder line is energized" can run to years, and during that lag a portfolio economist is working blind unless they go looking for another signal.

The light data isn't proof of outcomes. I want to be clear about that, because it's a mistake I see analysts make once and then correct quickly. A bright cluster tells you something is drawing power and probably running at night: shops staying open, a cold-chain unit, pumps, a school with evening classes. It doesn't tell you household connection rates, tariff collection, or whether the new load is residential or a single diesel generator someone finally turned off. For that you still need the baseline survey, the household panel, the utility's own connection register. None of that changes.

What it gives you is a reason to ask a sharper question two years before the formal answer arrives. Instead of waiting for the PCR to tell you energization slipped, you can flag it in month 20, call the implementation unit, and find out whether the slip is in the wires or in the sign-off. Those are different problems with different fixes, and knowing which one you have while the project is still disbursing is worth more than knowing it after the fact.

I've started treating the monthly tile pull the way I treat a disbursement variance report: a prompt to ask, not an answer in itself. Electrification Map is built around exactly that habit, a monthly night-light read over your programme area that sits between your formal surveys and tells you where it's worth making the call early.

If you're sitting on a portfolio with the same lag between contract and file, it's worth putting one of your valleys on the map.

Start a project

Back to the blog