Generative · loading…

Make room for maintenance

Eskom fixed more of the grid by taking more of it offline.

In FY2023, South Africa’s state utility ran its coal stations during 96% of the hours they were available. Its own reports put the international norm at roughly 75%. Almost a third of total capacity was lost to breakdowns, availability fell to 56%, and the country endured 280 days of "load shedding," or planned, rotating power outages. The fleet ran flat out because it kept failing, and kept failing because maintenance, as ever, comes after dealing with emergencies.

Eskom’s board approved a Generation Recovery Plan on December 10, 2022. A month later, its chair said power stations needed "space and headroom" to do proper maintenance without constant firefighting, and offered two ways to get it: add capacity, or declare a permanent stage 2 or 3 for the next two years so the outages would at least be scheduled and the country could plan around them.

Six stations causing more than half of unplanned losses got first claim on experienced people, capital and equipment manufacturers. Each station got a recovery plan, central tracking and direct board attention.

Eskom raised planned maintenance from 10% of capacity in FY2023 to 12% in FY2024 and increased repair spending 30%. The numbers got worse in the short term, with availability falling to 55%, load shedding rising to 329 days and emergency diesel generation costing R34 billion. Planned maintenance peaked at nearly 13% in FY2025, and that is when the problem started to turn: unplanned losses fell six points and availability rose six. In FY2026 planned maintenance came back down to 11.5% while availability kept climbing to 65%; a fleet that breaks less needs less time to catch up. Through August 27 of this year, availability is averaging 68% for the financial year to date, diesel spending is down 82% against the same period last year, and South Africa has gone 469 days without load shedding.

Queueing theory predicts this: waiting times explode as utilization approaches 100%. Jay Galbraith treated slack resources as a basic response to uncertainty. Tom DeMarco called slack the degree of freedom that lets a company change. This ought to be a pattern in a future version of the book, called Protected Capacity. Teams and entire companies should intentionally reserve part of the system for maintenance, learning and the unexpected, and force themselves to not let ordinary demand claim it.

Eleven years ago, at Undercurrent, we divided the year into thirteen four-week periods. We sold twelve and reserved the thirteenth for reviews, offsites and knowledge-sharing. Each cycle, we also staffed two people below our theoretical maximum. This cost us about $1 million in foregone revenue. It also gave us room for project rescues, postmortems, vacations and the work that kept the firm working. We got a lot more profitable as a result.