Single Request Architecture (SRA) vs Per-Slot Calls

Single Request Architecture (SRA) vs Per-Slot Calls

If you're on GPT and making a separate ad request per slot, you're leaving roadblocks and competitive-exclusion money on the table. SRA bundles all slots into one call.

What it does — sends every page slot to the ad server in a single request.
Setup effort (3/5) — enableSingleRequest() plus defining all slots before display.
RPM lift seen — +5-9% from better line-item competition and proper roadblock/exclusion fills that per-slot calls can't coordinate.

Pros — faster (one round-trip), enables roadblocks, better demand competition.
Cons — incompatible with naive lazy-loading (all slots must be defined upfront); needs careful pairing with deferred display.

Lift: ★★★★☆ — Risk: ★★☆☆☆
Best for: GAM publishers with 4+ slots above the lazy boundary. Skip if your whole page is lazy-loaded individually — the models conflict.