Why I Run Octopus Agile
I run Octopus Agile because it gives me the raw half-hourly price and manages nothing on my behalf. Prices move every 30 minutes, pegged to the wholesale spot rate, published for the next day around 4pm. Some slots are expensive, some are cheap, and a few go negative. The tariff decides nothing for me, which is what I like about it. A house full of loads scheduled against a live price is a good problem to sit with of an evening, and I like sitting with it. It saves money too, but that's not why I bother.
Why not a managed product
Intelligent Octopus Go and E.ON Next Drive, EDF FreePhase (this is the closest to Agile, but softens things too much!) are good products. For most people they are the right call. You plug in the car, the supplier picks when it charges, and you get a tidy low overnight rate without thinking about it. Someone else carries the complexity and the risk, and you get on with your life.
That works because it covers one appliance, the EV in most setups, and the supplier makes the dispatch decision. It has to be that simple. A managed tariff has to be standard enough to run across thousands of homes and aggregate into flexibility the supplier can sell back to the grid. My conservatory's thermal behaviour is not transferable to anyone else's house, so there is no version of this they could package and sell me even if they wanted to.
Most people want that. I want the raw signal, and room to point the whole house at it.
What Agile lets me do
Agile hands me one price and lets me point several loads at it, each with a different kind of constraint, all competing for the same cheap windows on rules I set.
- The EV. A load that can wait, but has a deadline. The system looks days ahead using wholesale forecasts and holds out for a cheap night rather than charging on a mediocre one. If I plug in on Sunday and Wednesday looks windy and cheap, it waits for Wednesday.
- Hot water. Immersion heater against gas boost is a fuel choice as much as a timing one. Below about 8p/kWh the immersion heater beats the boiler on cost once you account for its efficiency losses. Above that, gas wins. The automation checks the price and picks the cheaper fuel before each scheduled heat.
- Air conditioning, room by room. Each room behaves its own way, and the conservatory swings widest with almost no thermal mass to hold temperature. A small model scores how well each slot will cool each room, and a separate scheduler weighs that score against price, under a hard ceiling that keeps it out of the evening peak.
Point all three at the same price curve at once and the numbers get silly. One heavy day I pulled 174 kWh and came out at minus £8.72, paid to run the house. That is three arbitrage tricks stacked against one price signal.
The rope
The trade for that freedom is that when it breaks, it is on me. No supplier SLA sits behind a state machine I wrote. If the immersion automation misfires I get a cold shower and a debugging session. No managed fallback catches me. If the AC scheduler makes the wrong call on a hot day, a room sits at 30°C until I notice.
I build overrides for this. There is a run-now button for the AC and the car charger that ignores the price ceiling. Some days comfort is worth 80p/kWh, and I want to make that call myself. Most people are right not to want this deal. I like fixing things, so I'll take it.
What's next
I have already written up the price-aware EV charging, which I built around a PHEV. Since then I have moved to a full EV. The old automations weren't up to it, so I rebuilt the lot with multi-day scheduling and proper overrides. I will write that up as a part 2, along with the hot-water switching and the AC scheduling model, as their own posts over the coming weeks.