Somewhere in Harare the power is off right now. Nobody reading this finds that sentence remarkable, and that is the whole point of this essay.
Most operating advice assumes the inputs. Power is on, the network is up, the bank portal loads, the fuel is there, the person who knows the system came to work. Everything in the textbook is built on top of those assumptions, and in most of the world they hold often enough that nobody notices they are assumptions.
Here they do not hold. So the question is not how to run a good operation. It is how to run one when a good part of what it stands on is missing on any given Tuesday.
The outage is the normal case
The mistake is to treat the outage as an exception. An exception is something you handle when it happens. But if the grid fails three afternoons a week, the afternoon without power is not an exception. It is a third of your operating time, and a process that does not work during it is a process that does not work.
I got this wrong for years. I designed for the day everything worked and then improvised on the others, and the improvising was where the mistakes lived. The order taken on paper and never entered. The payment confirmed by word of mouth. The reconciliation skipped because the system was down, and then skipped again because it had been skipped.
Improvisation is not a plan. It is the absence of one, wearing a plan’s clothes.
The inventory
Write down everything your venture needs in order to complete one ordinary transaction. Not the big things. All of them.
Electricity at the premises. Mobile data on the customer’s phone. The bank’s switch. The payments gateway. The one laptop with the licence on it. The supplier who delivers on Wednesdays. The person who knows the passwords. The road.
Then, for each one, three questions. What happens if it is gone for four hours. What happens if it is gone for two days. What happens if it is gone for two weeks.
Most founders have never done this and are startled by the result. The venture that looked robust turns out to rest on one person, one connection, and one supplier, any of which can disappear without notice.
Three design rules
Degrade rather than stop. When the input fails, the venture should get worse, not halt. At Kose a ride does not become impossible because a driver’s data dropped for two minutes; the trip has to be able to complete and settle afterwards, because the alternative is a rider standing on a pavement and a driver unpaid. At Cafe Oldrock, when the card machine cannot reach the switch, there is a cash float and a paper record, and the paper is reconciled that night like everything else. The worst version of a process is the one that has exactly two states, working and stopped.
Keep the record outside the system. Whatever you are running on, there should be a version of the truth that survives it going dark. It does not have to be elegant. A book, a spreadsheet on a second device, a photograph of the day’s tickets. The point is that when the system comes back, you are re-entering rather than reconstructing, and the two are separated by hours of work and a great deal of error.
Buy the redundancy you will actually maintain. Everybody owns an inverter. Fewer people know whether it would hold a charge tonight. Redundancy that is not maintained is worse than none, because you have paid for the confidence and not the protection. Buy less than you think you need and service all of it.
Redundancy you cannot afford is a promise you did not make
There is a temptation, once you have done the inventory, to try to cover everything. You cannot, and the attempt will bankrupt you slowly.
So choose. Which promises to customers do you intend to keep on the bad day, and which will you openly not make. Same day delivery that depends on a road you do not control is not a promise, it is a hope, and a customer would rather be told that than discover it.
The ventures I respect in this market are not the ones that never fail. They are the ones whose failures are the ones they told you about in advance.
The honest cost
Redundancy is dead money most days.
The generator that ran eleven hours this year cost the same as the one that ran eleven hundred. The cash float sits in a drawer earning nothing. The second supplier is more expensive than the first, and most weeks you did not need them. Your competitor who skipped all of it will look leaner, and for a while they will be.
Then the bad month arrives, and it always arrives, and the difference between the two of you is not that you were smarter. It is that you decided, on an ordinary day when everything worked, to build for the day it would not.
