CONCEPT Cited by 1 source
Placement-theme cohesion¶
Definition¶
Placement-theme cohesion is the quality property of a multi-section recommendation page that every section (placement) is intentionally grouped, ordered, and aware of the other sections around it — so the overall page reads as a coherent narrative rather than a chaotic pile of independent themed cards.
The term is Instacart's — "every placement should be intentionally grouped, ordered, and aware of others around it. We want the discovery journey to feel seamless."
Why it's a named tenet¶
The failure mode cohesion targets is explicit in the 2026-02-26 Instacart post:
"Placements are often created by different siloed teams with divergent focus areas and goals. As a result, the series of placements can result in a chaotic surface presentation. Users are required to scroll without the ability to easily navigate the page to solve their needs."
The organisational root cause is legacy content-ops workflows: multiple teams authoring placements independently, each optimising its own section without page-level awareness. Cohesion is the design tenet that flips this — the generation system sees the whole page at once and produces a page-level narrative, not a union of section-level decisions.
Architectural consequences¶
Three design choices implied by taking cohesion as a tenet:
- Top-down generation, not bottoms-up. See concepts/top-down-vs-bottoms-up-generation. Top-down generates the page structure first, so cross-section cohesion is a Phase-1 optimisation target rather than emerging accidentally from a clustering step.
- Single generative agent, not per-section generators. The page-design agent sees user context + business objectives and emits the full ordered section list jointly; different agents per section can't reason about cross-section interaction.
- Cross-section evaluation as a first-class rubric. See patterns/llm-as-judge-multi-level-rubric. Page-level cohesion + diversity is a dedicated evaluation criterion at the page-level rubric, distinct from per-section quality.
Canonical wiki instance — Instacart Shopping Hub (2026-02-26)¶
Source: sources/2026-02-26-instacart-our-early-journey-to-transform-discovery-recommendations-with-llms
Cohesion is one of three named tenets for the Shopping Hub rebuild (alongside delightful personalization and adaptability). The post's framing:
"Cohesion: The system should enable full cohesion across the page — every placement should be intentionally grouped, ordered, and aware of others around it. We want the discovery journey to feel seamless."
Implementation in Phase 1: the page-design agent emits ordered themed placements representing "discrete and coherent shopping intents". Ordering is part of the output — not just which themes appear but in what sequence. Phase 3's LLM-as-judge includes "does the page feel cohesive enough?" as a page-level quality dimension.
What cohesion trades against¶
- Diversity. Over-indexing on cohesion can produce themed-but-monotone pages (five placements all about breakfast variants). The Phase-3 LLM-as-judge rubric pairs cohesion with diversity to balance.
- Per-section personalization. If the page structure is optimised jointly for cohesion, individual section content may be less personalised than a per-section generator could achieve. Instacart's cascade softens this by having Phase 2 personalise within the cohesion-optimised section structure.
- Cross-team autonomy. Cohesion requires a single generative pipeline owning the full page — individual teams lose their section-level freedom in exchange for page-level coherence.
Relation to sibling concepts¶
- concepts/generative-recommendations — cohesion is one of the tenets that makes generative-recommendations an upgrade over human-authored-section-libraries.
- concepts/top-down-vs-bottoms-up-generation — cohesion is the primary reason top-down wins over bottoms-up for hierarchical content surfaces.
- concepts/cascaded-llm-generation — Phase 1's cohesion- first ordering enables the rest of the cascade.
Seen in¶
- sources/2026-02-26-instacart-our-early-journey-to-transform-discovery-recommendations-with-llms — canonical wiki instance. One of three named tenets for the Shopping Hub rebuild. Phase 1 emits ordered themed placements; Phase 3 scores page-level cohesion as a dedicated rubric axis.
Related¶
- concepts/generative-recommendations — the parent concept.
- concepts/top-down-vs-bottoms-up-generation — the design- choice axis that cohesion drives.
- concepts/cascaded-llm-generation — the pipeline shape.
- patterns/top-down-cascaded-page-generation — the canonical production pattern that realises cohesion.
- patterns/llm-as-judge-multi-level-rubric — cohesion is scored at the page level.
- systems/instacart-shopping-hub — the surface where cohesion matters.
- systems/instacart-generative-recommendations-platform — the platform that delivers cohesion.
- companies/instacart