CPG Beverage Personas

What this is

Built from real baskets

Loyalty and switching in beverages are usually measured by asking people, or modelled from retail share. This dataset is derived from actual retail purchase activity by consented participants: Amazon in the US, Amazon and Tesco in the UK, Amazon and Flipkart in India. Every purchase is classified to brand and category, then resolved to one profile per person.

Why it is different

What the dataset gives you

What is in each profile

Up to 88 features across six blocks

Volume and cadence

volume_purchase_count, units, spend, cadence_mean_interpurchase_days, plus calendar-month and weekday seasonality (cadence_spend_jan through cadence_spend_dec).

Brand

brand_primary, brand_concentration_hhi, per-brand purchase counts, and portfolio share fields.

Category

category_primary as a need-state signal, category diversity, and per-category share of spend.

Loyalty

loyalty_repeat_purchase_rate, loyalty_brand_switching, repertoire depth.

Momentum

momentum_spend_ratio, momentum_trend, momentum_at_risk_flag, momentum_reactivation_flag.

Coverage

Per-retailer coverage flags and coverage_cohort_tier, which buckets people as light, regular or heavy.

How teams use it

Five starting points

Sample query

At-risk portfolio loyalists by market

SELECT
    b."identity_market",
    b."brand_primary",
    COUNT(DISTINCT b."user_id")                            AS unique_personas,
    ROUND(AVG(b."brand_coke_portfolio_share"), 3)          AS avg_portfolio_share,
    ROUND(AVG(b."identity_days_since_last_purchase"), 0)   AS avg_days_since_last
FROM MEASURE_EVENT_METRICS_PROD.ACCESS.CPG_BEVERAGE_PERSONAS b
WHERE b."brand_coke_portfolio_share" >= 0.8
  AND b."momentum_at_risk_flag" = 1
GROUP BY b."identity_market", b."brand_primary"
HAVING COUNT(DISTINCT b."user_id") >= 20
ORDER BY unique_personas DESC;

The HAVING clause in this example limits results to groups of at least 20 people.

Specifications

What ships

Privacy & provenance