Weather is the most talked-about external signal in CPG demand planning and also the most operationally underused. Demand planners intuitively know that weather affects beverage sales — everyone knows ice cream sells more in summer. But "weather affects demand" is not a forecast model. A forecast model needs the correlation coefficient, the relevant temperature range, the lag time between forecast and demand response, and the regional distribution weights. This article is about those specifics.
We'll go through five beverage sub-categories, detail the weather signal that matters for each, and describe what a usable correlation looks like in practice. None of these are universal laws — they're starting hypotheses that should be validated against your own brand's historical data before being embedded in a forecast model. But they're directionally right for most North American CPG beverage operators.
Category 1: Sparkling Water and Still Water
Primary weather driver: average daily high temperature. Secondary driver: heat index (temperature + humidity combination).
Sparkling water demand has a moderate-to-strong positive correlation with temperature above the 70°F threshold. Below 65°F, temperature has minimal predictive power — demand is driven more by habit and usage occasion than by temperature. The relationship activates meaningfully around 72–75°F and strengthens through 85°F. Above 88–90°F, the incremental marginal effect flattens.
Correlation coefficient in the temperature-sensitive range (70–90°F) vs weekly sparkling water velocity: approximately r = 0.62–0.71 at the national aggregate level. Regional calibration matters: this correlation is stronger in Sun Belt markets (Texas, Florida, Arizona) and weaker in the Pacific Northwest, where consumption patterns are less temperature-sensitive.
Lag time: essentially zero for temperature-driven demand. Consumers respond within 1–2 days of a temperature event. This makes the 7–10 day weather forecast the relevant planning window for sparkling water — not the 8–12 week S&OP horizon. For S&OP purposes, the relevant input is the seasonal temperature outlook: "will the upcoming 8-week period be warmer or cooler than seasonal average?"
Category 2: RTD Iced Coffee and Cold Brew
Primary weather driver: temperature, but with a notable nuance — iced coffee has a habit usage component that moderates the weather signal compared to sparkling water.
Habitual iced coffee drinkers consume cold coffee year-round regardless of temperature (ask any barista in Minnesota in January). The weather signal affects the marginal consumer — the person who switches from hot coffee to cold coffee when temperatures rise, or who increases purchase frequency in summer. This dual demand structure means the aggregate weather correlation is weaker than for sparkling water: r ≈ 0.45–0.55 for RTD iced coffee vs temperature.
Where weather becomes important for RTD cold brew is the seasonal transition timing — the weeks in March–April when temperatures cross the 55–60°F threshold in northern markets and habitual hot coffee drinkers begin transitioning to cold formats. Catching this transition 2–3 weeks early allows production adjustments before the POS signal reflects it. NOAA's 8–14 day temperature forecast is reliable enough for this purpose.
Planning implication: for RTD iced coffee, weather signal is most valuable for transition-period forecasting (spring and fall) rather than peak-summer forecasting. During July and August, temperature variability has modest marginal impact on an already-elevated demand baseline. During April and October, a 10°F temperature deviation from seasonal average can shift weekly velocity by 15–20%.
Category 3: Hot Tea and Instant Hot Beverages
Primary weather driver: cold temperature and precipitation, not heat. This is the inverse case.
Hot tea velocity correlates negatively with temperature above 55°F (r ≈ -0.58 to -0.65 at weekly aggregation) and positively with precipitation on the same day or 1–2 days following a cold front. The precipitation correlation is intuitive — people associate a rainy, gray afternoon with hot tea consumption in a way they don't with a cold but sunny winter day.
The planning challenge for hot tea is the fall demand ramp-up. Hot tea brands typically see demand accelerate in early-to-mid October in northern US markets, but the exact timing varies 2–4 weeks year-to-year based on when the first sustained cold period arrives. An early October cold snap can pull the demand ramp forward by 3 weeks compared to a warm fall. A POS-only model using prior-year October scan data misses this entirely — it can only tell you what demand looked like last October, not whether this October will come early or late.
NOAA's 3–4 week temperature outlook for September–October is the relevant forecast input here. Even with its 55–60% verification rate, a "warmer than average" vs "cooler than average" October outlook materially shifts the probability distribution of when the demand ramp-up begins. That conditional probability is more useful than assuming mean-reversion to prior-year timing.
Category 4: Sports and Performance Drinks
Primary weather driver: outdoor activity weather, not temperature directly. Secondary driver: temperature during summer months.
Sports drink demand is driven by physical exertion, and physical exertion outdoors is influenced by weather. But the mechanism is indirect: people exercise outdoors when weather is suitable for their activity, and the weather thresholds for "suitable" vary by region and activity type. Running shoes are laced up in conditions that would ground cyclists. Summer soccer leagues drive category velocity in a different temperature range than summer hiking.
For national-level planning, the most reliable weather signal for sports drinks is outdoor temperature above 75°F combined with low precipitation probability — "good outdoor activity weather." The correlation between weekly activity-weather days (≥75°F + <30% precipitation probability) and sports drink velocity is approximately r = 0.53–0.61. Weaker than the sparkling water correlation but still operationally meaningful.
Worth noting: the sports drink category has a significant social and identity component that moderates weather sensitivity compared to pure hydration beverages. A consumer who identifies as an athlete may purchase sports drinks at consistent rates year-round as part of their routine, regardless of whether they ran outside that week. The weather signal captures the incremental category expansion consumer, not the core user.
Category 5: Functional/Better-For-You Beverages
Primary driver: social trends, not weather. Weather signal is weak (r < 0.3) for most functional beverage sub-categories.
Adaptogen beverages, probiotic drinks, electrolyte waters, and similar functional formats are purchased based on perceived health benefit and social influence — not temperature. The category grew substantially regardless of the weather in any given year. Trying to apply weather-signal forecasting to these SKUs will produce noisy, unhelpful results.
This is the important boundary case: not every beverage SKU benefits from weather-signal enrichment. Applying weather signals uniformly across a beverage portfolio is a mistake that dilutes the signal-to-noise ratio for the entire model. Category-specific signal selection — using weather where the correlation is real, using social signals where they're the primary driver — produces better results than treating all external signals as universally applicable.
Building Your Own Correlation Profile
The correlations above are starting hypotheses, not your model parameters. The right approach is to run your own correlation analysis against 2–3 years of weekly scan data for each SKU or category, correlating against the weather variable most likely to be relevant (temperature, precipitation, heat index, or activity-weather composite depending on category).
A few specifics to check in your own data:
- Threshold effects: does the correlation hold across all temperatures, or only above/below a threshold? Plot demand vs temperature in a scatter chart before computing a linear correlation — a non-linear relationship won't show up clearly in r alone.
- Lag structure: test correlation at 0-day lag, 3-day lag, and 7-day lag. For some categories (hot beverages, weather-occasion foods) demand leads or lags the weather event by a few days in a consistent pattern.
- Regional variation: aggregate national correlation may be moderate even when regional correlations are strong, because markets in different climates have opposing signals in the same week. Segment by Nielsen market or custom region before drawing conclusions.
- Seasonality interaction: weather signal is stronger in transition seasons (spring and fall) than in the middle of summer or winter, where demand is already at seasonal extremes and additional temperature variance has smaller marginal effect.
Once you have category-specific weather correlation profiles, you can build those relationships into forecast deviation factors: when the 10-day forecast calls for temperatures X degrees above seasonal average, apply a Y% upward adjustment to the affected SKU baseline for the corresponding weeks. That's a practical, defensible approach that a demand planner can explain to the commercial team without a statistics lecture.