Solar Production Borderline Low: Service Now vs Watch A Cycle Threshold Decision Tree

Why this matters

A homeowner calls because monitoring shows production "a little low," but no fault is logged and the array still produces meaningful power. Rolling a truck on every borderline case burns margin and trains customers that any minor swing is an emergency. Waiting on a real issue lets a recoverable loss compound into a warranty event or a missed degradation claim. This tree gives a defensible threshold: when borderline-low warrants a same-week service call, when it warrants a "watch one production cycle" callback, and when the data itself is unreliable and the right action is to fix monitoring before fixing the array.

Section 1: Validate that the data is real before acting on it

Before treating low production as a hardware symptom, confirm the data is trustworthy. Borderline cases are most often a monitoring artifact, not an array fault.

Check:

• Did the comparison period have similar irradiance, ambient temperature, and panel temperature to the reference period? A 10 percent week-over-week drop with a 15 percent drop in plane-of-array irradiance is normal.
• Is the production CT or revenue-grade meter reading consistent with the inverter's internal reported output? A spread greater than a few percent points to a CT clamp, polarity, or calibration issue, not an array issue.
• Did a microinverter or optimizer go offline silently (reporting zero but not faulted), so the gateway is showing aggregate down without an alarm?
• Was there a clipping change (battery added, export limit reduced, schedule change) that is curtailing on purpose?

If any of these is yes, fix the data layer first. Diagnosing the array against bad data wastes a visit.

Section 2: Quantify the deviation against a defensible baseline

Borderline is not a feeling, it is a number. Use the same baseline every time so the threshold is repeatable.

Recommended thresholds for a typical residential string or microinverter system, weather-normalized:

• Less than 5 percent below expected on a rolling 14-day window: monitor only, no action.
• 5 to 10 percent below expected: watch one production cycle (typically one week of good weather) and reassess. Open a ticket with a callback date.
• Greater than 10 percent below expected, or any single string or microinverter more than 15 percent below its peer group: dispatch service.
• Any zero-production module, string, or inverter for more than 48 hours of usable sun: dispatch service regardless of array total.

These are starting numbers. Tighten them for systems under a production guarantee or PPA where the contract has a tighter band.

Section 3: Pattern recognition: what shape is the loss

The shape of the loss tells you whether to dispatch now or wait.

Uniform loss across all strings, modules, and inverters:

• Likely soiling, snow residue, smoke or haze, or seasonal sun-angle shift.
• Watch one cycle after the next rain or cleaning. If it persists with clean modules and good weather, escalate.

Single string or single inverter low, peers normal:

• Likely string-level issue (shading change, one bad module pulling the string, MPPT fault).
• Dispatch within the week. Do not wait.

Single module or microinverter low or zero, peers normal:

• Likely module, optimizer, or microinverter failure, or one connector.
• Dispatch within the week. A single bad module can drag a string over time and the warranty clock is running.

Time-of-day pattern (low only morning or only afternoon):

• Likely a new shading source (tree growth, neighbor build, new vent). Confirm with site visit, no urgent hardware risk.

Sudden cliff drop on a specific date:

• Something changed that day. Cross-check firmware push, configuration change, weather event, or a homeowner action.

Section 4: When borderline becomes urgent

Some borderline cases should bypass the watch-a-cycle path and be dispatched immediately:

• Any ground fault, arc fault, or insulation resistance warning paired with low production. The production loss is a symptom; the protection event is the priority.
• Any module-level temperature alarm. Hot spots can become a fire path.
• Any DC connector temperature alarm if the system has connector telemetry.
• Borderline-low production on a system still inside the production guarantee window where a missed dispatch could affect the claim.
• Borderline-low coinciding with a battery or hybrid system not charging to expected SOC, which suggests the loss is real and affecting backup capacity.

Section 5: Communicating the decision to the customer

The customer's tolerance for "wait a week" depends on what they hear. Use plain framing:

• If watching a cycle: "Current numbers are within normal variation given recent weather. We have logged a ticket and will reassess after the next week of clear days. If you see a fault code or a further drop, call us immediately."
• If dispatching: "We see a deviation that is outside normal weather variation, so we want to look at the array before it gets worse."
• If the data is bad: "The first thing we need to fix is the monitoring so we can tell whether the array is actually losing production. We will start there."

Do not promise a number you cannot defend. Borderline production talk-tracks are where trust is lost.

Section 6: Document the threshold call

Every borderline decision should be logged with the deviation percentage, the reference period, the weather-normalization source, and the chosen action (monitor, watch a cycle, or dispatch). This protects against the homeowner saying later "you told me it was fine" when conditions changed, and it builds a per-customer history so the next borderline call is faster to resolve.

If the call was "watch a cycle," set a hard callback date. An unfollowed-up watch ticket is the worst outcome: it is neither a fix nor a confident no-action.

References

• NEC 2023, Article 690 (Solar Photovoltaic Systems).
• NEC 2023, Article 705 (Interconnected Electric Power Production Sources).
• IEEE 1547-2018, Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces.
• IEC 61724-1, Photovoltaic system performance, Part 1: Monitoring (industry reference for normalization methodology).