Whether a solar system is producing the yield it was sold on comes down to two figures that are almost certainly already on your phone: the specific yield since commissioning, in kilowatt-hours per kilowatt-peak, and the ratio of what the system actually produced against what the quote modelled. Put those two next to the number the installer wrote down and you have an answer. The running daily kilowatt-hour total is not that answer, and treating it as one is how people end up in a standoff over a system that is working exactly as designed.
I'm on the back step with a mug of tea that has gone cold, because I've been staring at a phone. It is a Tuesday in October, the sky is that hard Highveld blue, and the reading says 41 kWh for the day. My neighbour, same size array, same street, says 46. Two numbers, both true, and completely useless on their own. I know that because I've been the person who phoned the installer about a fault that turned out to be three cloudy days and a tree.
The number you should actually be reading
Specific yield is the honest one. Take the total energy your system has produced since commissioning and divide it by the array size in kWp. A 6 kWp array that has made 6 400 kWh has a specific yield of about 1 067 kWh per kWp. That figure is directly comparable with the annual number in your quote, because the array size has already been divided out of it.
The catch is the time base. If your quote modelled, say, an annual 1 600 kWh per kWp, and you're eight months in, you cannot compare the two yet. You can pro-rate, but pro-rating assumes every month is average, and no month in this country is average.
Performance ratio is the more useful metric once you have enough data, because it measures losses rather than raw output. It is the energy that came out of the inverter divided by the energy the sun actually delivered to the roof. Panels get hot, dust sits on the glass, cable and inverter each take a slice. A well-designed system in decent condition tends to sit somewhere in the high seventies to low eighties as a percentage, lower on a hot dusty afternoon, higher on a cold clear morning. Don't chase a target number. Watch whether it drifts.
One honest limitation: performance ratio needs an irradiance figure, and unless you have a sensor up on the roof, that figure is an estimate generated on the monitoring side. It is still the best comparison you have, because it strips out array size and weather in one move.
Why the daily total lies to you
The daily figure is weather plus season plus whatever your house did that afternoon. It swings. Forty-one kilowatt-hours on a bright, cool October day. Eighteen on a wet one in June. Fifty in December when nobody's home and the pool pump is running. Watching it day to day is like weighing yourself straight after a braai and drawing conclusions about your diet.
Monthly totals are better, but they average away short interruptions, which is exactly where the interesting things hide. A forty-minute grid loss on a bright morning barely dents a monthly bar.
Where the gaps actually show up
In the event log, or whatever your system calls its list of alerts and status changes. That is where grid-loss events, inverter faults, high-temperature derating and charge limits appear, each with a time stamp. Three separate events of forty minutes each, all on sunny mornings, can quietly account for a visible slice of a month's yield. Nobody sees that on a bar chart.
Charge limits deserve their own mention, because they get misread as faults. On a hybrid system that is set not to export, once the battery is full and the house isn't drawing much, the inverter has nowhere to send the power and simply reduces what the array produces. In spring that happens by mid-morning. The system is fine. The setting and the household routine are the reason the measured yield sits below the model.
Which leads to the counterintuitive bit: putting a load in the middle of the day, like a pool pump or a geyser, can raise your measured yield. Not because you fixed anything, but because you gave the system somewhere to put the power.
Twelve months, not three
Monitoring only ever covers the period since commissioning, and it starts the day the logging started, which is sometimes a week after the system went live. So your history has a beginning, and it is not the beginning of the year.
Seasonality is large enough to make a short review meaningless. On the Highveld, winter days are short but the cold air keeps panels efficient and the skies are often clear; summer brings the long days but also heat derating and afternoon storms. In Cape Town the winter drop is bigger, because it's driven by cloud and rain rather than day length. Review a system in July against an annual average and it will look like it has failed. It hasn't. It's July.
Twelve months is the minimum before you draw a conclusion about underperformance. Twelve months also happens to contain the December shutdown, the two weeks away, and the month you forgot to wash anything.
What usually explains the difference
Ranked by how often I've seen it, not by how interesting it is.
Shading that grew. A tree that was small when the panels went up. A neighbour's second storey. A jacaranda. A new aerial. Shading rarely costs you only the shaded panel, because most residential arrays are strung in series and the weakest panel drags the string down with it. That's why a shading problem shows up in the numbers as a whole-system problem.
Soiling. Highveld dust, spring pollen, veld-fire ash, bird droppings, and a gravel road upwind. A dirty array doesn't fail. It just quietly produces a bit less every month, and the decline is slow enough that you don't notice until you clean it.
A change in how the house uses power. A new heat pump running at night. A geyser timer moved to the evening. A teenager who now showers at ten. If you don't export, less daytime load means more curtailment, which means lower measured yield. Nothing is broken. The pattern changed.
Actual equipment failure. One string down, a blown fuse, an inverter that derates hard when the ambient temperature climbs. This is the least common cause and the first one everyone assumes.
Panel degradation sits underneath all of it, at a fraction of a percent a year. A gentle multi-year slide is normal. A step change is not.
Keep a record, even a lazy one
Once a month, take a screenshot of the total generation figure and the specific yield, and note the date. When you remember, export the data. Not because a screenshot proves anything on its own, but because it's the first thing an installer or a manufacturer asks for when a performance conversation starts. It turns "I think my system is rubbish" into a dated comparison: this March, 480 kWh; last March, 530 kWh, on the same array.
Keep it in the same folder as the quote and the commissioning paperwork. And if your municipal meter records export, that bill is a second, independent measurement worth checking against the app, though most South African homes aren't exporting anything at all.
The question "is my system producing what it was sold on" is really two questions: what was I sold, and what am I measuring. Pull the specific yield, watch the ratio, read the event log, and wait for a full year. Most of the time the answer is that the system is fine and something else on the property changed. Occasionally it isn't, and on those days you'll be glad you have twelve months of dated screenshots instead of an argument.
