Roofing spikes the same week. HVAC lags a quarter.
A single storm front moves two trades on two different clocks. Treat them the same in your media plan and you will overspend into a roofing trough and miss the HVAC peak entirely. The fix is to read the lag, not the headline.
Most demand planning treats weather as a single switch: a heatwave or a hailstorm arrives, and every weather-exposed business should lean in at once. That instinct is half right and expensively wrong. Weather does drive demand, but different services respond on different timelines. The same week that fills a roofer's calendar can be a quarter too early for an HVAC installer.
The cleanest evidence comes from L.E.K. Consulting's Weather Index for Service Essentials (WISE), a consulting-grade framework that regresses building-services demand against two weather inputs: degree days measured from a temperature baseline, and counts of extreme weather events such as severe storms, floods and high winds. L.E.K. is a US study run on US roofing and HVAC shipments, but the structure it found is a property of buyer behaviour, not geography, and it maps cleanly onto Australian trades. It found two statistically significant patterns that point in opposite directions in time.
One unit note before the findings. L.E.K. and the US Energy Information Administration measure degree days from a 65°F base. The Australian equivalent is 18°C, the metric degree-day base the Bureau of Meteorology publishes against (BOM uses 18°C, with 12°C and 24°C variants for milder and harsher thresholds). Heating degree days accumulate below 18°C, cooling above it. We use 18°C throughout.
The two clocks, in L.E.K.'s own words
On roofing, L.E.K. found that the number of weather events in a given quarter leads to an increase in roofing shipments in the same quarter, then a decline of larger magnitude two quarters later. Storms break roofs, homeowners replace them immediately, and that pull-forward empties the near-term pipeline. The same-quarter spike is real, but it borrows from the future.
On HVAC, the timing inverts. L.E.K. found that the number of degree days in a given quarter leads to an increase in HVAC shipments in the following quarter. Extreme temperature first shows up as repair calls and strained systems; the replacement decision, a larger and more considered purchase, lands a quarter later. Demand does not vanish, it queues. In Australia this is overwhelmingly a cooling signal: cooling degree days dominate most of the country and peak in summer, December to February, while heating degree days concentrate in the southern and temperate states (Victoria, the ACT, Tasmania and alpine NSW) and peak in winter, June to August.
Source: L.E.K. Consulting, "A WISE Approach to Weather-Related Services Demand" (Bromfield, McGrath, Highfield & Moore, 2024), a US study; both regressions reported as statistically significant. Degree-day framing for Australia: Bureau of Meteorology heating and cooling degree days, 18°C base (bom.gov.au). The original L.E.K. and US Energy Information Administration base is 65°F, which equals 18°C.
Why the same storm runs on different clocks
The split is about decision latency, not weather. A failed roof after a hailstorm is an emergency with no substitute and no deferral, so the job books that week. A failed or strained air conditioner often gets a repair first, and the full replacement is a planned, financed, multi-thousand-dollar decision the household sits on for weeks. The weather signal is identical; the purchase behaviour is not. In Australia the event clock for roofing is the spring and early-summer hailstorm in the eastern capitals (Brisbane and Sydney severe-storm season runs roughly September to December) and the November-to-April tropical cyclone and storm season across the north (Queensland, the Northern Territory and northern WA), per the Bureau of Meteorology.
Search behaviour shows the same asymmetry at higher frequency. US analysis of seasonal home-services search volumes (WebFX, 2024) finds heating and cooling queries swing far harder than roofing: heating-system repair searches lift roughly 594% from trough to peak and frozen-pipe repair about 609%, while emergency roof repair peaks near 70% and roof-replacement-cost queries around 55%. These are US figures; we use them for the shape of the curve, not the level. Roofing intent is event-driven and spiky; HVAC intent is seasonal and sustained, which is why it rewards a forward-loaded plan. The phase inverts in Australia: cooling and emergency air-conditioning intent peaks in the southern-hemisphere summer (December to February), and the smaller heating and frozen-pipe spike lands in southern-state winter (June to August), not the US January peak.
The math: setting the lag instead of guessing it
You do not need L.E.K.'s budget to find your own lag. The standard method is a lagged Pearson correlation between a weather feature and demand, scanned across candidate lags, with an autoregressive control so you do not mistake plain seasonality for a weather effect.
One caution the academic literature is firm on: control for the baseline climate first. A ToolsGroup study of forecast accuracy found that when normal seasonality is not netted out, categories get falsely flagged as weather-sensitive. The defensible approach is multiple regression with autoregressive terms (MLR-AR), where last period's demand absorbs the seasonal cycle and the weather coefficient captures only the genuine, exogenous shock. In HVAC load studies the autoregressive terms dominate the fit while weather enters as a real but secondary driver.
Method references: lagged Pearson correlation with autoregressive controls is standard in weather-demand work. See ToolsGroup, "Using Weather and Climate Data to Improve Demand Forecasting"; and the peer-reviewed heating and cooling degree-hour study (Kajewska-Szkudlarek, Nature Scientific Reports, 2023; PMC10576096), which finds recent temperature lags the strongest predictors of building thermal-load demand.
A worked example: timing one roofing media budget
Consider a regional roofer in south-east Queensland who treats a major hailstorm as a steady-state lift and holds spend flat for two quarters after the event. Walk the demand index through the WISE-shaped pattern and the waste is plain.
What to actually do with the lag
The operating rule is simple to state and uncomfortable to follow: spend where demand is going to be, not where the weather just was.
For roofing and other event-driven trades
Treat storms as a short, sharp capture window, then expect a trough. Lean in hard the same week with responsive search and Local Services Ads while intent is live, harvest the same-quarter spike, and pull paid spend back two quarters out rather than defending a market that has already bought. Reserve the recovered budget for the next event.
For HVAC and other lagging trades
Read this quarter's degree days as next quarter's pipeline. Launch 30 to 45 days before peak, and in Australia the peaks invert from the northern hemisphere: cooling demand peaks in summer (December to February), so cooling campaigns should be live by October, and heating demand peaks in southern-state winter (June to August), so heating campaigns should be live by April. The horizon logic holds either way: temperature forecasts are only reliable 7 to 10 days out, but seasonal demand patterns are predictable 3 to 6 months ahead. The window to commit budget opens long before the thermometer moves.
| Trade | Weather signal | Demand lag | Media posture |
|---|---|---|---|
| Roofing / siding | Extreme-event count | k = 0 (same quarter), trough at k = 2 | Surge at impact, pull back 2 quarters out |
| HVAC install | Degree days | k = 1 (next quarter) | Pre-load 30-45 days before peak |
Lag structure from L.E.K. WISE (2024, US). 30-45 day pre-launch guidance from WebFX seasonal home-services analysis (2024, US). Australian seasonal timing (cooling peaks Dec-Feb, heating Jun-Aug, hail Sep-Dec in the eastern capitals, cyclone season Nov-Apr in the north) per the Bureau of Meteorology.
Australian Air Conditioning & Electrical
A seasonal HVAC and electrical business where reading demand a quarter ahead, not a week behind, let the account scale into peak windows instead of reacting to them.
The takeaway
Weather is not one signal, it is a clock with two hands. Roofing demand spikes the week the storm hits and dips two quarters later as the pipeline empties; HVAC demand barely moves at impact and peaks the quarter after. The operators who lose money on weather read the headline and spend on the spot. The ones who win read the lag, set it with a lagged correlation against their own data, and put the next dollar where demand is heading. That timing discipline, demand routed to its highest-margin moment, is what we mean by raising profit velocity: the same budget, placed on the right clock.