Design Product Q&A

Product Q&A on PDP: problem framing

Product Q&A lives on the product detail page, where purchase intent is highest — it is often the last objection before add-to-cart. Shoppers ask compatibility, sizing, and warranty questions; sellers, brand reps, or verified owners answer. Unlike reviews (a star rating about a past purchase), Q&A is conversational, searchable, and pre-purchase — a shopper asking "does this fit a 2019 Civic?" is deciding whether to buy right now.

The structural complexity is multi-tenant routing on a marketplace. An Amazon-style listing has a buy-box winner and possibly many sellers on one ASIN — a question must route to the winning seller for that offer, not blast every merchant. Home Depot may blend manufacturer tech reps with store associates; Best Buy may prioritize Geek Squad answers for electronics. So the system separates catalog identity (the ASIN/SKU — what the question is about) from listing/seller context (which offer the shopper is viewing — who should answer), capturing the offer context at question-creation time (the buy-box can rotate later). At scale — 200M MAU, 80M Q&A-enabled product pages, a ~12% PDP attach where the panel loads — it is read-heavy (everyone reads, few ask) with write bursts when a SKU goes viral, and the read path must be edge-fast (p95 ~120 ms).

The interviewer probes three things: separating catalog from listing/seller, preventing duplicate questions (a viral SKU gets the same question hundreds of times), and surfacing answers on SEO snippets without stale claims (a cached answer must not assert "in stock" after the SKU is discontinued). The failure to avoid is treating Q&A as a generic comment thread — the marketplace routing, the dedup, and the moderation are what make it hard. The principle: Q&A is pre-purchase, conversational, read-heavy commerce content where the hard parts are multi-tenant routing (catalog vs listing), duplicate suppression, and moderation — not the CRUD.