A screen printer will run twelve shirts for you. They just won't make it cheap. Every color needs its own screen, roughly $20 to $30 to burn, plus press setup and registration — the same cost for twelve pieces as for twelve hundred. It lands on every unit. That math produced the standard first launch: one idea, a few hundred units, hope. On demand the constraint is gone. The habit stayed.

One number, five explanations

Say your tee sells eleven units in three weeks. Eleven could mean the design is weak, the price is off, the garment photographs cheap, the audience wasn't listening, or you posted on a Tuesday in August.

Three pieces won't make it a clean experiment either: placement, weight, color and price all move at once, and at eleven units any gap between listings is noise. It takes months and a few hundred orders before a difference means anything. What a capsule buys before then is more ways in.

A one-product store reads as a rehearsal

A single item ends the visit where it starts: no collection, no related row, nothing to add. Per-unit cost on demand is higher than bulk and always will be, so what rescues you is the parcel: shipping is billed per shipment with a small increment for each extra item, and the acquisition cost was already spent on item one.

A minimum credible capsule

Three to five listings, and be strict: size and color are variants inside a listing; placement and garment weight are separate listings with their own photographs.

• One artwork family, two placements. A large front and a smaller chest hit — confirm your provider offers both: a second location is a second pass on the machine, often priced extra.

• Two garment weights. Airventory's 250 to 450gsm combed cotton is heavier than it sounds: a Gildan 5000 is about 180gsm, a Bella+Canvas 3001 nearer 142. The lightest here is already heavyweight, the top end fleece. Two far apart show whether people want the graphic or the garment.

• Two colors. A third colorway adds another set of mockups and teaches you almost nothing the first two didn't. Darks are pretreated and printed over a white underbase, so art approved on white can come back muddier and softer-edged on black.

• The same finishing throughout. Woven neck label, hangtag, branded mailer, zero minimums, no Airventory branding. Your label doesn't replace the legal one: fiber content, care instructions and country of origin stay on the garment — FTC care labeling and textile fiber rules in the US, Regulation 1007/2011 in the EU, its retained version in the UK — usually on a tear-away or side-seam label.

The part that's actually work

The expensive input is file prep, not capital. Build artwork at 300 PPI at final printed size — pixels in your file, not the 1200 DPI the press images at. A realistic full front is 30 to 35cm wide, about 3,500 pixels across; the 60x60cm ceiling is a platform maximum, not a placement, since an adult L tee is only 56cm chest to chest. Printable area shrinks with garment size, so one file placed identically on XS and 3XL sits wrong at one end.

Send a transparent PNG or vector in sRGB. The RIP converts to CMYK plus that white underbase, so saturated blues, oranges and neons print duller than your screen promised, and Adobe RGB or CMYK files shift again on the way in. DTG and DTF hold a line down to about half a millimeter, so a 30cm front reduced to a 9cm chest print drops hairlines and halftone detail. Small placements want redrawing, not resizing.

Mockups and a page for each is the real tax: an afternoon per piece, not a pallet in the hallway. Order the capsule as samples before you announce a date and put it on a person. You'll catch the chest print sitting two inches high.