A solar light that dies at dusk is not a mystery. The panel and the battery were sized for a July sky that will not come back for another eight months, and your November sky never got a vote.

We build these lights and I have watched this failure happen to a hundred of them. The problem is not the LED. It is three numbers on the label that do not add up: the panel-to-battery ratio, the shade and angle you actually get, and the milliamp-hours the battery is pretending to hold. Here is the math, and the two specs that separate a working light from a garden ornament.

The ratio nobody prints

Start with what a light actually has to do in one night. Daily watt-hours = LED watts x hours of use, multiplied by a loss factor of about 1.3, then divided by your peak sun hours. A 30-watt fixture running 10 hours with 20% dimming needs 240 watt-hours a day. At four peak sun hours, that is a 78-watt panel. At the 2.5 sun hours a northern winter actually gives you, the same math asks for 125 watts.

Now look at what a retail solar light ships with. A 3 to 5-watt panel feeding a 2000 mAh cell. Someone sized that pair for four full summer sun hours, not for the 2.5 hours your November sky produces. The panel produces half of its summer current, and the battery is a few hundred charge cycles into losing capacity. The light does not break. It simply stops being able to close the gap before dark, and that is exactly the month you paid for it to work.

Shade is a multiplier, not a footnote

A panel needs three to four hours of direct sun to fill a battery. If a hedge, a fence, or a branch throws shadow across the panel for one hour each afternoon, you lose a third of your day. And solar cells do not fade gracefully. A single shaded strip on a panel can drop the whole array's output by 40 to 50 percent, because the panel is wired as one long string of cells that all pull together.

Tilt matters the same way. A panel laid flat in a planter takes a fraction of the winter sun that a panel tilted south at 30 to 40 degrees takes. I have run this on our test bench: under partial shade, the charge controller sags from 5.0 volts to under 3.8 in about an hour, and that is the moment the light dies mid-evening rather than holding to the small hours. It is not that the sun was bad. It is that the panel was doing half the work and nobody checked.

The battery is lying to you

Most cheap solar lights ship with a 1.2-volt NiMH cell. Those cells lose 20 to 30 percent of their rated capacity after a couple hundred charge cycles. In cold weather, usable capacity drops a further 15 to 25 percent, because the chemistry itself slows down.

There is a third cut you do not see. The controller only discharges the battery to 60 or 70 percent of its rating to protect the cell, so a '2000 mAh' light is giving you 1200 to 1400 mAh usable. A 2-watt LED drawing about 170 milliamps for eight hours wants 1360 mAh. You are already over budget before the cold. That is why these lights do three good nights in June, die by December, and come back to life in May. The spec sheet is a full battery. The physics is half of one.

What to actually buy

Three checks, in order. First, panel watts and battery milliamp-hours together, not separately. If a listing hides one of those two numbers, the runtime claim is a guess. A 1-watt panel with a 1000-mAh battery is a decorative object that works on sunny afternoons and stops on cloudy nights, and for a garden spike that is fine. For a path light or a candle you are meant to rely on, run the math first.

Second, battery chemistry. Lithium-ion in a 2026 light is worth more than NiMH at twice the price, and if the cell is sealed in a non-serviceable plastic housing, you are buying a one-year product, not a three-year one.

Third, IP rating at the connector. Water intrusion is the other top failure mode, and a sealed IP67 body does not help if the battery lid is the weak point. The light that fails in February is usually the one that got wet in January.

For solar candle lanterns specifically, the math is friendlier than for a security fixture. A 0.5 to 1.5-watt load is 20 to 60 percent of what a path light draws, so the same undersized panel that fails a pathway fixture can carry a candle all night. The failure physics is the same, the margin is different. Keep a solar candle where the sun angle is decent, size the panel at least double what the label claims, and it will outlive a path light with half the drama.

The lights that survive in the field are the ones where somebody did the multiplication before shipping. The ones that die are the ones where the engineer picked a battery that hit the price target.