# Add-to-cart experiment log instructions

This blank template contains no historical client data or assumed results.

State the hypothesis and the single change first. Define the products, market, device scope, visitor eligibility, versions and observation window. Define how views, cart additions and purchases are attributed to the same product. Track sitewide outcomes separately rather than mixing scopes.

The primary metric can be unique users who successfully add to cart divided by eligible product visitors. Define eligibility and deduplication. Keep button clicks, event counts and successful cart users separate. Record failed requests, missing variant selections and unavailable stock separately.

Subsequent purchasers must come from the successful-cart-user group and be followed within the agreed window. Record paths that bypass the cart separately; an order count is not a purchaser count. Do not remove genuine direct purchases to make the funnel look tidy.

For a randomized experiment, fix the randomization unit and keep each user's assigned version stable. Set the sample, smallest effect worth detecting and analysis time beforehand. Do not declare success from a short-term rise. Monitor purchases, cancellations, refunds, loading and errors alongside the primary metric.

Absolute rate difference in percentage points = test rate minus original rate. Relative change = test rate divided by original rate, minus 1. If the original rate is zero, leave relative change blank and explain. Record uncertainty with the numbers. Do not enter zero for missing data or an inadequate sample.

Label task observations and before-and-after comparisons accurately; they are not randomized experiments. Never enter invented results for an unimplemented change. Limit conclusions to the products and users actually observed.