Define every variable
List the customer-controlled fields: name, date, message, photo, pet, color, role, or number. For each, define allowed length, characters, format, fallback, and whether it changes layout. A free-form message is a different production risk from choosing one of five colors.
Plotia's structural review of 1,499 public Etsy personalization listings is used to classify names, dates, family members, pets, photos, locations, memorial events, and multi-subject combinations. It is a purposive corpus, not a random sample of the marketplace, and it does not establish sales or market share. [Plotia Research, 2026]
| Variable | Controls required |
|---|---|
| Name or message | Length, characters, spelling, wrapping, font behavior |
| Date or number | Format, locale, fact check, ordering |
| Photo | Resolution, rights, crop, subject count, retention |
| Person or pet | Identity features, count, ordering, add/remove rules |
| Color or style | Allowed choices, product base color, producibility |
Show customers exactly what they must provide and which edits are included.
Build resilient templates
Separate locked brand or decorative elements from editable fields. Test the shortest, longest, narrowest, and widest expected values. Create layout rules for line breaks, font size, photo crop, and missing data. Do not depend on AI to reproduce exact customer text.
For portraits, decide whether the promise is stylization, close likeness, or a decorative interpretation, then review accordingly.
For multi-person or multi-pet templates, define count limits, arrangement rules, hierarchy, and what happens above the limit. A mature template is defined as much by exception handling as by its default example.
Make approval explicit
Normalize the order data, check image quality, generate a proof, and display the exact spelling and crop. State whether production begins automatically or after approval, how long approval remains open, and which changes trigger a new proof.
Save the approved proof with the order and production file. Never infer a corrected name or date without confirmation.
Plan exception paths
Define what happens when a photo is too small, a face is cropped, multiple subjects overlap, prohibited content appears, or customer text exceeds the template. Use a human-review queue for identity-sensitive or ambiguous orders.
Collect only the data needed for the order, restrict access, and remove customer assets according to the stated retention policy.
Use explicit exception states—waiting for customer, waiting for review, cannot produce, source replaced, or proof renewed. Ambiguous orders should never move to production under a silent default.