Your system is short on kWh. Here is what that actually means.
Six numbers look identical on a monthly bill and mean six different things on a roof. Module-level monitoring tells them apart, usually without anyone climbing a ladder.
Shortfall triage
Put in your array size and the kWh your app recorded last month, then pick the shape of the loss.
- Expected
- 1,088 kWh
- Missing
- 448 kWh
- Shortfall
- 41.2%
Large enough to have a single cause
Gaps above about 12 percent usually mean one identifiable thing rather than an accumulation of small ones. Module-level data normally names it in a couple of minutes.
Soiling
On the chart: A smooth, even curve that is the right shape but too short. Gets worse week by week through the dry season and resets after the first real rain.
Soiling. Fresno gets almost no rain between May and October, and the Valley adds orchard dust, ag burn particulates and pollen on top of it.
- Typical loss
- 3 to 7 percent by late September on a typical Fresno roof; 8 to 12 percent within 400 ft of an unpaved road or an active almond harvest
- What we do
- Deionised-water wash, no detergents and no pressure. On a care plan we schedule one wash in June and one in September.
- Who pays
- Care plan, or $9 per module as a one-off (sample)
A failed bypass diode, a cracked cell, a delaminated junction box or a loose MC4 connector
On the chart: Twenty-three modules tracking together and one drawing a lower line with the same shape. In the layout view it is a single dark tile.
A failed bypass diode, a cracked cell, a delaminated junction box or a loose MC4 connector. If the whole module reads zero, suspect the connector or the module-level electronics first, not the module.
- Typical loss
- Roughly 1/N of the array, so about 4 percent on a 24-module system. A failing diode can also cook the junction box, which is the part that matters.
- What we do
- IV-curve trace and a thermal scan at the module, then replace under the product warranty. Same-brand replacements are held in our Fresno stock.
- Who pays
- Module product warranty (12 to 25 years) plus our labour warranty
An open string: a tripped DC disconnect, a blown fuse, a rapid-shutdown device that failed closed, or a ground fault the inverter latched on (GFDI)
On the chart: A step change. Production drops by a clean fraction of the array overnight and stays there.
An open string: a tripped DC disconnect, a blown fuse, a rapid-shutdown device that failed closed, or a ground fault the inverter latched on (GFDI). The inverter almost always logs a fault code for it.
- Typical loss
- 1/2 or 1/3 of the array until it is cleared, which on a 7.8 kW DC system is 12 to 19 kWh on a summer day
- What we do
- Same-week truck roll. We read the fault log first, then test insulation resistance before anyone goes on the roof.
- Who pays
- Workmanship warranty if it is our wiring; inverter warranty if the device failed
Inverter clipping, not a fault
On the chart: The curve rises normally, then runs dead level for two or three hours around solar noon before falling normally.
Inverter clipping, not a fault. The array can push more DC power than the inverter is rated to pass as AC. A 7.8 kW DC array on a 6.0 kW AC inverter is a 1.30 DC/AC ratio and will clip on the clearest days.
- Typical loss
- Usually under 2 percent of annual kWh at a 1.25 ratio. Designed clipping is often the cheaper trade: the inverter costs less than the kWh it loses.
- What we do
- Nothing, unless the ratio was set wrong at design. We print the expected clipping hours in every proposal so it is never a surprise later.
- Who pays
- Not a fault, so nothing to claim
New shading
On the chart: A clean notch in the curve. It moves later in the day and deepens from November to January, then shallows again.
New shading. A neighbour's tree, a new satellite dish, a swamp cooler, or an added roof vent. The seasonal drift is the giveaway: faults do not track the sun.
- Typical loss
- 5 to 25 percent of the affected modules' kWh depending on how much of a substring the shadow covers
- What we do
- A fresh shade study against the original one, then either vegetation work or moving three or four modules to a clear plane.
- Who pays
- Shade study free for care-plan customers; relocation quoted per module
Normal degradation
On the chart: Nothing visible on a daily curve. It only shows when you compare weather-corrected annual kWh across several years.
Normal degradation. Expect about 2 percent light-induced loss in year one, then 0.4 to 0.55 percent a year for a good mono PERC module.
- Typical loss
- About 12 percent total at year 25, which is why warranties are written as retained output rather than as a fault
- What we do
- Nothing to fix. We weather-correct your annual figure and compare it against the module's warranted output curve at your anniversary.
- Who pays
- Module performance warranty if retained output falls below the warranted curve
A monitoring dropout, not a production loss
On the chart: A hard gap in the chart with no fault code, often starting the day a router was replaced.
A monitoring dropout, not a production loss. The gateway lost Wi-Fi, cellular or its RS485 link to the inverter. The system kept making power the whole time.
- Typical loss
- Zero kWh. You lose the record, not the energy.
- What we do
- Re-pair the gateway, usually over the phone in ten minutes. We check your true production against the utility's net meter reading to prove nothing was lost.
- Who pays
- Phone support, free for the life of the system
Expected kWh comes from a sample 1,500 kWh per kW DC per year for a fixed-tilt Fresno array, split across the year by month. Your own proposal uses a modelled figure for your roof, tilt, azimuth and shading. Sample figures, not a performance guarantee.
Seven shortfall signatures
Every one of these produces a smaller number on a bill. Only the shape of the curve tells you which one you have.
| Pattern | Cause | Typical loss | Urgency |
|---|---|---|---|
| Every module down by a similar percentage, all day | Soiling | 3 to 7 percent by late September on a typical Fresno roof; 8 to 12 percent within 400 ft of an unpaved road or an active almond harvest | Next scheduled visit |
| One module flat-lines or sits far below its neighbours | A failed bypass diode, a cracked cell, a delaminated junction box or a loose MC4 connector | Roughly 1/N of the array, so about 4 percent on a 24-module system. A failing diode can also cook the junction box, which is the part that matters. | Same week |
| A whole string or one inverter input reads zero | An open string: a tripped DC disconnect, a blown fuse, a rapid-shutdown device that failed closed, or a ground fault the inverter latched on (GFDI) | 1/2 or 1/3 of the array until it is cleared, which on a 7.8 kW DC system is 12 to 19 kWh on a summer day | Same week |
| A flat top on the production curve on clear days only | Inverter clipping, not a fault | Usually under 2 percent of annual kWh at a 1.25 ratio. Designed clipping is often the cheaper trade: the inverter costs less than the kWh it loses. | Next scheduled visit |
| A dip at the same clock time every day, drifting through the year | New shading | 5 to 25 percent of the affected modules' kWh depending on how much of a substring the shadow covers | Next scheduled visit |
| Every module slightly lower than last year, year after year | Normal degradation | About 12 percent total at year 25, which is why warranties are written as retained output rather than as a fault | Next scheduled visit |
| The app shows zero, but your meter still spins backwards | A monitoring dropout, not a production loss | Zero kWh. You lose the record, not the energy. | Next scheduled visit |
Swipe to see the full table
What your monitoring can and cannot see
The tier you buy decides whether a failed module is a phone call or a mystery that runs for two years.
Inverter-level only
One number for the whole system, or one per string
- Finds one dead module?
- No. A single dead module hides inside the total.
- Typical on
- String inverter with no optimisers
Fine on a simple, unshaded plane if you read the annual figure carefully
String-level
Each MPPT input separately, usually two or three per inverter
- Finds one dead module?
- Only if the module fails hard enough to drag the string.
- Typical on
- String inverter, multiple roof planes
Catches open strings and ground faults quickly; misses slow single-module decline
Module-level (MLPE)
Volts, amps and watts per module, every five minutes, on a roof-layout map
- Finds one dead module?
- Yes, usually before the homeowner notices anything.
- Typical on
- Microinverters, or optimisers on a string inverter
What we fit as standard, because it turns a diagnosis visit into a phone call
Monitoring questions
Module-level data arrives every five minutes, and our alert threshold is a module running more than 15 percent below its neighbours for three consecutive clear days. In practice we usually call you before you have looked at the app.
The bill. A monitoring gateway that loses Wi-Fi or cellular stops recording, but the inverter keeps working and the utility's net meter keeps counting. We reconcile the two before treating a gap as a production loss.
No, that is inverter clipping. Your array can make more DC power than the inverter is rated to pass as AC, so the curve runs level at the inverter's limit for a few hours. It is a design trade we print in your proposal, and it usually costs under 2 percent of annual kWh.
About 2 percent in year one from light-induced degradation, then roughly 0.4 to 0.55 percent a year. That is why we compare weather-corrected annual kWh against the module's warranted output curve rather than against last month.
No. If a module or inverter fails inside its warranty we handle the claim and the labour. You pay only for things outside a warranty, such as a wash after a harvest season or moving modules because a new tree has grown into the array.