What is inside the glass
Cell architecture is the specification that most of the others follow from. 44.0% of the 13,034 panels declaring one now ship TOPCon, which took the market from PERC in about four years and did it on cost per watt rather than on peak efficiency.
The comparisons below are averages of what each architecture is currently sold at, not of what it can do. A technology that only appears in premium product lines will average well for reasons that have nothing to do with the cell.
What the market is shipping
44.0% of declared panels. PERC is being retired catalogue by catalogue rather than all at once, which is why it still holds a large share of a database that includes everything currently listed rather than only this year's releases.
Cell type is the older classification and splits the market far less finely - almost everything is monocrystalline. It is kept because datasheets still quote it and because the exceptions, thin-film and BIPV, are genuinely different products.
What each architecture is worth
Three measurements, three axes that start at zero. The gaps are real and none of them is a chasm.
Back-contact leads because it moves the front-side busbars to the rear and stops shading its own cells. The spread from top to bottom is a few percentage points, not a different class of product.
Heterojunction leads this one by construction, and by enough that it can outproduce a nominally more efficient module over a hot summer. Closer to zero is better.
n-type architectures avoid the boron-oxygen defect that causes light-induced degradation in p-type cells, and the warranties follow: the median declared annual rate across the whole set is 0.40 %/yr, and the gap between the best and worst architecture compounds to several percentage points of output over a 25-year life. Lower is better.
Modules got bigger, not just better
The rating distribution is a wafer-format story rather than a cell one: 182 mm and 210 mm wafers moved the standard residential module from around 400 Wp to around 495 Wp. A higher rating on a bigger frame is not a better module, which is what the power density page is for.
Every average here is over the panels currently listed with that architecture, so it mixes a decade of product generations. A technology that is only just arriving will look better than it is, because the only modules carrying it are flagships.
None of these differences decides a purchase on its own. The cell architecture tells you which trade-offs a module is likely to have made; the panel's own datasheet tells you what it actually does.
How these numbers are built
Every figure comes from a manufacturer datasheet ingested into the panel database. Nothing is modelled, estimated, or supplied by a brand outside its own published specification.
Aggregates recompute as new panels arrive. A panel that does not declare a value is left out of that average rather than filled in, so every figure here is backed by modules that actually publish it.
No paid placements, and no figure here reads which shops pay us a commission. Each power rating counts as the module it is. Bifacial datasheets that quote one module at several assumed rear-side gains are counted once, on their front-side rating.