Explain grocery-specific product strategy and scrappy XP

## 1) What makes grocery different from restaurant delivery (and why it matters) Below are nine grocery-specific aspects, each with product and experimentation implications. 1) Substitution behavior (item-level fulfillment) - Why different: Customers order many SKUs; stockouts are common. Shoppers/pickers often substitute. - Product: Upfront substitution preferences ("same brand vs lowest price"), approved alternatives, price-parity rules, picker suggestion UX. - Experimentation: Item-level outcomes (fill rate, substitution accept rate) matter; cluster at store or order level to avoid interference. 2) Perishability and cold-chain - Why different: Frozen/chilled items require temperature control; long waits degrade quality. - Product: Cold-chain badges, courier thermal bag compliance, sequence chilled items last in route. - Experimentation: Guardrails on temperature-sensitive late arrivals; may need store-hour switchbacks to respect capacity. 3) Basket size and variety - Why different: More items per order, heavier/voluminous; mix of planned and top-up missions. - Product: Aisle navigation, bundles, free delivery thresholds, scheduled delivery. - Experimentation: Heavier tails in AOV; power analyses need variance-aware designs and longer durations. 4) Inventory accuracy and stockouts - Why different: Catalog volatility; imperfect APIs/feeds; on-shelf ≠ feed. - Product: Real-time stock badges, OOS prediction, hide/penalize risky SKUs, proactive subs. - Experimentation: Store-level heterogeneity; use stratification by partner/integration type; monitor non-compliance when feeds fail. 5) Picking/packing latency - Why different: In-store picking adds prep time; shopper efficiency varies. - Product: Aisle-aware pick lists, batching, image search, barcode scan verification. - Experimentation: Staff behavior changes; prefer switchbacks with picker training held constant. 6) Delivery windows (scheduled vs ASAP) - Why different: Customers plan baskets; stores have hours/cutoffs; staging enables batching. - Product: 1–2 hour windows, dynamic pricing by window, capacity gating. - Experimentation: Window availability must respect capacity; randomize at store-hour (switchback) rather than user-level. 7) Fees, margins, and contribution economics - Why different: Lower margins on CPG items; picking cost significant; ad/retail media revenue possible. - Product: Smart fees, minimums, threshold-based promos, retail media placements. - Experimentation: Optimize contribution margin/order, not just conversion; include refunds and shopper time. 8) Refund/replace policies - Why different: Higher return/refund volume; partial refunds common. - Product: Self-serve refunds, price-adjusted subs, post-order make-goods. - Experimentation: Guardrails to detect moral hazard/fraud; outcome metrics net of refunds. 9) Compliance and restricted items (ID, EBT/SNAP, alcohol) - Why different: Legal/ID checks; tender types; age-restricted flows. - Product: ID-at-door flows, tender selection, restricted-item gating. - Experimentation: Segment experiments to exclude flows with strict compliance or monitor separately. --- ## 2) Two focused bets ### A) Growth bet — Scheduled delivery windows to unlock planned baskets - Hypothesis: Offering 1–2 hour scheduled delivery windows (with smart capacity gating) increases conversion and AOV for planned baskets without increasing lateness beyond guardrails. - Target segment: Zip codes within 2 miles of participating stores; users browsing >5 minutes or >6 product views (planning signals); store-hours with available picking capacity. Metrics - Primary: - Order conversion rate (sessions → orders). - Average order value (AOV) and units per order (UPO). - Secondary: - New buyer rate for grocery (first grocery order). - 28-day repeat rate for grocery. - Guardrails: - On-time-in-window rate (OTW) ≥ baseline − 1 pp. - CS contacts per 1k orders ≤ baseline + 5%. - Courier/picker utilization within safe bounds. - Leading indicators: - Window selection CTR, distribution across windows. - Add-to-cart after window selection. Scrappy, low-risk experiment - Treatment: Show 1–2 hour scheduled windows (including off-peak incentives like $0.99 delivery for low-demand windows) on eligible store-hour inventory. - Control: ASAP only (status quo). - Randomization unit: Store-hour switchback (e.g., each participating store toggles treatment/control by 2-hour blocks). This respects capacity and avoids cross-user interference. - Assignment: 50/50 blocks, stratified by day-of-week and peak/off-peak. Capacity gate prevents offering windows when pickers/couriers are constrained. - Duration: 2–3 weeks to cover weekly cycles. - Power note (example): If baseline conversion is 7% and we expect +0.7 pp (10% relative), with pooled SD ≈ 0.25 at session level, we’ll need ~60–80k sessions per arm; feasible across multiple stores. Tiny numeric example - Baseline: 7% conversion, AOV $48, OTW 92%. - Target lift: conversion 7.7%, AOV $52, keep OTW ≥ 91%. - Contribution margin: if margin/order is $6 baseline, +$4 AOV at 25% gross margin adds ~$1, easily offsets window incentive of $0.50 when used <30% of orders. ### B) Efficiency bet — Pre-cart OOS prediction with proactive substitutes - Hypothesis: Predicting likely OOS items and preemptively recommending in-stock substitutes reduces pick/pack time, cancellations, and refunds while preserving conversion. - Target segment: Items with high OOS probability in participating stores; stores with historical OOS labels or API feeds. Metrics - Primary: - Item fill rate = delivered items / requested items. - Order-level cancellation rate. - Secondary: - Substitution acceptance rate (customer- or picker-approved). - Picker time per item (mins/item) and total pick time/order. - Refund $ per order. - Guardrails: - Session-to-order conversion (no worse than −0.3 pp from baseline). - AOV (no worse than −2%). - Leading indicators: - Click-through on suggested substitutes. - % of high-risk items avoided (grayed/hidden) before add-to-cart. Scrappy, low-risk experiment - Treatment: For items with predicted OOS ≥ τ (e.g., 0.8 precision threshold per store-SKU-hour), gray out or de-rank the item and show 1–2 in-stock substitutes with clear labels ("In stock; similar brand"). - Control: Current experience (no pre-cart OOS treatment). - Randomization unit: Store-switchback by day (whole catalog treated vs not), to avoid item-level interference and simplify ops. Alternatively, item-level within store for only the top 100 high-risk SKUs if traffic is limited. - Assignment: 50/50 days per store; stratify by weekday/weekend. - Duration: 2–4 weeks. - Safety valve: Start with a small % of traffic (e.g., 20%) and high-precision threshold to minimize false positives; ramp after 3 days if guardrails hold. Tiny numeric example - Baseline item fill rate: 92%, cancellations 3.0%. - Target impact: +2 pp fill rate (94%), cancellations −0.5 pp. - Economics: If refunds average $1.20/order and picker time is 18 min/order at $0.35/min, a 1 min saving/order plus $0.30 fewer refunds yields ~$0.65/order margin improvement. --- ## 3) Top risks and mitigations 1) Prediction errors harm conversion (false OOS) or trust - Mitigations: Use high-precision thresholds; limit to SKUs with strong signal; prominent “see alternatives” and easy override; continuous monitoring with fast rollback. 2) Capacity and SLA breaches from scheduled windows - Mitigations: Capacity gating by store-hour; conservative initial quotas; dynamic throttling; guardrail monitoring and auto-disable. 3) Partner/store pushback (perceived cannibalization or catalog suppression) - Mitigations: Share store-level results and net sales; opt-in pilot; exclude key SKUs from treatment initially; co-design substitutes with merchants. --- ## 4) How grocery experimentation differs from restaurants - Randomization unit: Grocery often needs store-level or store-hour switchbacks (capacity, picker ops), vs user-level for restaurants. - Catalog volatility: SKUs and availability change intra-day; define intent-to-treat and track non-compliance when items go OOS. - Multi-stage fulfillment: Shopper picking introduces additional latency and variability; measure pick-time KPIs and control for picker shifts. - Interference risk: Couriers and pickers shared across arms can cause spillovers; cluster by store and time blocks. - Longer decision cycles: Planned baskets span days; need longer experiments to capture repeat behavior and weekend effects. - Heavier tails: AOV and order times have high variance; consider nonparametric estimators, CUPED, or hierarchical models. - Compliance constraints: Age-restricted/EBT flows limit randomization; segment or exclude from tests. --- ## 5) Product and operational knowledge needed up front - Inventory integration: Feed latency/coverage, API reliability, which stores have real-time vs batch updates, historical OOS/error taxonomies. - Picking model: Who picks (in-store staff vs platform shoppers), training, typical pick times, barcode scanning coverage, aisle maps. - Capacity model: Picker and courier availability by hour and zone; batching rules; SLA definitions (promised window, handoff times). - Catalog and merchandising: Top SKUs by store, substitution maps, private label constraints, restricted items. - Financials: Fee structure, commission/markup, incentives, retail media revenue, per-minute picking cost, refund liability. - Compliance and policy: ID checks, alcohol, EBT/SNAP eligibility, building access norms in NYC (doorman, walk-ups), cold-chain requirements. - Experimentation platform: Ability to randomize at store/store-hour, support switchbacks, log item-level events (add-to-cart, OOS, subs), guardrail alerting and kill switches. - Baselines: Current conversion, AOV, fill rate, cancellations, OTW, picker times, refund rates by store and time of day. --- ## Appendix: Key formulas and definitions - Fill rate = delivered items / ordered items. - Substitution rate = items substituted / items ordered; acceptance rate = customer-accepted subs / substitutions proposed. - On-time-in-window (OTW) = orders delivered within promised window / orders with a scheduled window. - Contribution margin/order = revenues (delivery fees + markups + ads) − variable costs (courier pay + picker pay + refunds + promos + support). - Picker time per item = total pick minutes / items picked. Together, these bets and designs let you validate demand for planned grocery missions while improving fulfillment efficiency, with low operational risk and clear guardrails.