For a business that ships things, the money is made after the click — in product data, per-SKU margin and the return. Sixteen prompts that know what a landed cost is.
For a physical-product business, the money is made after the click
Most conversion advice was written for software. It assumes zero marginal cost, no shipping, no warehouse, and no version of a customer who receives the thing, dislikes it, and sends it back at your expense.
Change those assumptions and the whole optimisation problem inverts. A 12% lift in add-to-cart on a SKU with a 34% return rate and $9 of outbound shipping can lose money. A product page that converts beautifully and generates size complaints is a margin leak with good analytics. The decisions that matter for a physical-product business happen in product data, per-SKU margin, and the return — all of them downstream of the click that most ecommerce content stops at.
This pack was written under one editorial rule: no prompt on this page can be restated for a SaaS product. If it could, it belongs in a different pack.
Product data became the growth asset
The clearest recent shift is in where product discovery happens. Adobe Analytics found AI-referred traffic to US retail sites up 393% year over year in Q1 2026, and — the part that matters — that traffic now converts around 42% better than non-AI traffic. A year earlier the same channel converted 38% worse. That is a reversal, not a trend line, and the mechanism behind it is unglamorous: shoppers arriving through an AI assistant have already had their questions answered, and the assistant answered them from structured product data.
Which makes the product feed a growth asset rather than a technical chore that the agency handles. Build a Product Data Sheet That Machines Can Read is the prompt that takes this seriously — attributes, units, materials, dimensions, compatibility, and the specifics that let a machine answer a shopper's question without guessing. Fix a Product Feed That Keeps Getting Disapproved handles the more immediate problem of a feed that is rejecting items and quietly removing your catalogue from circulation.
The human-facing side is Write a Product Detail Page That Removes Doubt, which is framed around doubt rather than persuasion for a specific reason — see the returns section below. Write a Marketplace Listing That Ranks and Converts covers the same job under a marketplace's rules, and Plan a Product Photo and Video Shot List covers the imagery, which for most physical products carries more of the conversion load than the copy.
Returns are a product problem wearing a logistics costume
The National Retail Federation and Happy Returns put US retail returns at $849.9 billion in 2025 — 15.8% of annual sales, with an estimated 19.3% of online sales returned. Apparel runs higher still. Processing a single return costs somewhere between $10 and $65 once you count shipping, handling, inspection, restocking, and markdown. The same research finds 9% of all returns are fraudulent, and that 82% of consumers treat free returns as a major purchase consideration — up from 76% a year earlier.
At those numbers, returns are not a customer service line item. They are one of the largest single influences on whether a SKU makes money.
They are also mostly diagnosable. Find Out Why a Product Keeps Getting Returned works from return reasons, review text, and support tickets to separate the four causes that need entirely different fixes: the product is wrong, the description is wrong, the photography is wrong, or the sizing information is wrong. Three of those four are fixed on the product page, which is why the product-page prompt is built around removing doubt. Every ambiguity you leave on the page gets resolved by the customer ordering two sizes.
Write a Returns Policy That Protects Your Margin covers the policy itself, which has to do two jobs at once — reassure enough to convert, without underwriting bracketing behaviour you cannot afford.
Know what a unit actually costs
Work Out the True Landed Cost of a Product is the prompt to run before any of the pricing or promotion ones. Unit cost, freight, duty, brokerage, inbound handling, packaging, payment processing, outbound shipping, and an expected returns allowance. Most sellers know the first item on that list and estimate the rest, which is how a product with a healthy-looking gross margin turns out to be the least profitable thing in the catalogue.
Once you have a real landed cost, the inventory and merchandising decisions become answerable rather than intuitive. Decide What to Reorder and How Much balances stockout risk against tied-up cash. Clear Overstock Without Wrecking Your Margin works through the discount ladder in the order that costs least. Plan a Peak Season Without Running Out or Over is the annual version of the same problem, where both failure modes are expensive and the lead times remove your ability to correct mid-flight.
Set Shipping Rates and Free Shipping Thresholds and Build a Bundle or Kit That Raises Order Value both work from the landed-cost figure too. A free-shipping threshold set by looking at competitors is a guess; set from your own average order value and unit economics, it is a lever.
The checkout, and the cart people abandon
Cut Friction Out of Your Checkout or Signup Flow and Recover Abandoned Carts and Checkouts cover the last stage. The pattern worth internalising is that most checkout abandonment is not indecision — it is a specific, nameable objection appearing too late: an unexpected shipping cost, a forced account creation, a form that is longer than the purchase warrants. The fix is almost always to move the information earlier rather than to make the checkout prettier.
Get More Product Reviews and Handle the Bad Ones closes the loop. Reviews are simultaneously your conversion asset and your best returns diagnostic — a run of reviews saying runs small
is the cheapest product research you will ever get.
Where this stops
These prompts work from your numbers and are written not to invent them. Duty rates, tariff classification, and cross-border tax are named as inputs you need to confirm, not figures a model should estimate — the cost of a wrong classification lands on you. Anything touching product safety, labelling requirements, or consumer-protection rules in your market belongs with someone qualified in that jurisdiction.
Sources
- Adobe Analytics, on AI-referred traffic to US retail sites and its conversion rate relative to other channels, Q1 2026
- National Retail Federation and Happy Returns, 2025 Retail Returns Landscape — $849.9bn returned, 15.8% of sales, 19.3% of online sales (press release)