The new agency started in May. The heat arrived in June. Who gets the credit?

TL;DR

• HVAC marketing results should be reviewed beside weather, advertising spend, available technician hours and customer mix. A before-and-after job increase does not isolate the agency's effect.
• Cooling and heating degree days summarize how much daily temperature differs from a chosen base. They give the review a weather record; they do not translate directly into repair demand.
• Use one consistent weather station, equal follow-up windows and the same definition of a completed job. Keep missing records visible.
• The worksheet below is a check on a performance claim, not a statistical method that proves marketing caused the change.

What changed when the agency changed?

The new agency takes over in May. In June the phones ring more, the dispatcher books more work, and a report announces a 50% increase in jobs.

There may be good work behind that number. A repaired phone-routing problem or a smaller service area can improve the account. But June may also be hotter, advertising spend may be higher, and another technician may have joined the crew. A number that mixes all four changes cannot assign the credit among them.

Ask for a comparison that names the changes. The first version can be a one-page review built from your advertising bill, CRM, dispatch board and weather history. No new reporting software is required.

The budget and season hub covers planning the calendar. This article covers evaluating a result after it happens, including a result presented by an agency you like.

Which weather measure belongs beside HVAC marketing results?

The National Weather Service's degree-day explanation, checked October 8, 2026, uses a 65°F base. A day whose mean temperature is 80°F has 15 cooling degree days. A day whose mean is 50°F has 15 heating degree days. For each measure, a day on the other side of the base contributes zero.

Add the daily values across the review period. Keep heating and cooling separate; adding them together hides whether the weather favored furnace calls or cooling calls. Use the same station and temperature method throughout, and record the station, dates, base and source in the review.

This is a description of outdoor temperature relative to a base, not a prediction that fifteen degree days produce fifteen calls. Humidity, building conditions, equipment age, extreme heat and local behavior can change the work that follows. A heat wave can also leave a backlog after the temperature drops.

If records are missing for several days, do not treat them as zero. Either resolve the missing days or mark the period as incomplete. A comparison with incomplete weather can be useful, but it should not be labeled weather-adjusted.

What does the before-and-after table actually show?

Take an invented two-period comparison. Each period has the same number of days and the same cutoff for following inquiries into completed, collected jobs.

Jobs rose 50%. So did advertising spend. The cost per collected job did not improve. The weather measure doubled and available technician hours rose 25%. New-customer jobs increased by five.

None of those observations proves the agency did nothing. Higher spending might have bought work the shop would otherwise have missed. Better scheduling might have let the crew complete more jobs. Equally, stronger weather demand might have improved the result without an advertising change.

The defensible conclusion is narrower: the headline job increase alone does not establish improved acquisition efficiency or an agency effect. Ask which settings changed, when they changed, and which outcomes moved after each change.

Do not divide jobs by degree days and call the result adjusted performance. In this example, that ratio falls from about 0.44 to 0.33. It assumes a proportional relationship between weather and jobs that the table has not established. The ratio is descriptive arithmetic; it cannot settle the agency's contribution.

Which operational changes can imitate better advertising?

More answered hours. A shop that starts answering until ten at night can book calls it previously lost. Keep staffed phone hours in the review. A better ad account and a better call-handling process can improve together, but the result should say so.

More truck capacity. Record available technician hours, not just headcount. Vacations, overtime, training and installation assignments can change how much work the shop can finish. A full board caps completed jobs even when inquiries improve.

A different job mix. Ten additional repairs and ten additional replacements are different operational outcomes. Compare repair, replacement and maintenance separately before using average revenue per job.

A different customer mix. Returning customers can arrive through paid searches. Use the brand and customer-history reporting worksheet to keep name searches, customer history and completed work distinct.

A new measurement rule. Google's conversion measurement guide, checked October 8, 2026, describes conversions as configured actions. Replacing a call conversion with a completed-job import changes what is counted. Preserve the old and new definitions; a change in the column cannot serve as proof that the business changed.

These are review variables, not excuses. If the agency says an operational improvement was part of its work, ask for the change record. If it says weather explains a poor result, ask for the same weather comparison it would supply for a good one.

What comparison is fairer than May versus June?

Start with several earlier periods and the same season in the prior year where records exist. Keep spending, area, job type, customer history and capacity beside each. More context can reveal that the later period is an ordinary peak rather than an unusual improvement.

A prior-year comparison still has limits. The auction, prices, equipment base and crew can change across a year. A neighboring location may have a different climate or service mix. Select the comparison for similarity, and write down the differences before looking at which one makes the result flattering.

If the business has enough volume and separable markets, a planned test may be stronger. Define the treatment, comparison, outcome and stop rules before starting. Consider whether customers, technicians or campaigns can cross between the groups. A comparison location affected by the same changes is not an untouched control.

For a small shop, the useful next step is often to repair the records and collect more comparable weeks. Do not attach “statistically significant” to a spreadsheet with no uncertainty calculation, and do not run a spend experiment solely because an article suggested it. The worksheet is designed to review a claim before commissioning a test.

What should the agency put beside the result?

Ask for a dated list of changes: phone routing repaired, exclusions added, landing page changed, budget increased, conversion action replaced. Put the relevant approval or change-log reference beside each entry.

Then write one sentence describing what the data supports. In the invented comparison above: completed ad-attributed jobs increased with equal ad cost per job, during a hotter period with more spend and more capacity; the agency's independent contribution has not been isolated.

That sentence is less exciting than “50% growth.” It is more useful when deciding next month's budget. It leaves room to recognize real improvements without giving a seasonal change a permanent marketing budget.

For the spending decision, see HVAC slow-season planning and the HVAC marketing page. For what a report should disclose, see marketing agency report red flags.

What this doesn't cover

• A causal estimate. The worksheet identifies alternative explanations. It does not calculate the agency's incremental effect.
• A universal weather model. No fixed number of jobs per degree day is assumed.
• A test plan for every shop. Low volume, shared crews and overlapping areas can make a clean comparison difficult.
• Client performance claims. The numbers above are invented. No client result or market benchmark is implied.

Do this in your account

1. Copy the weather and operations review worksheet. It is open text with no signup.
2. Freeze the claim being reviewed, the date range, the outcome definition and the equal follow-up cutoff.
3. Add spending, heating or cooling degree days, weather source and station, available technician hours and staffed phone hours.
4. Split completed, collected jobs by job type and new, returning or unknown customer history.
5. List dated account and operational changes. Mark any tracking-definition change as a break in comparability.
6. Write the supported conclusion and the unresolved explanations. Keep both beside the headline result.

FAQ

Can weather explain all of an HVAC advertising improvement? It can be an alternative explanation, but this worksheet cannot measure how much. Weather, advertising and capacity can move together. A before-and-after result does not isolate their separate effects.

Should I compare the same month last year? Use it as one reference when available, with weather and operational differences stated. The same month can contain different weather, auction conditions, staffing and pricing.

Are jobs per cooling degree day a useful KPI? They can describe two recorded totals, but they are not automatically a weather adjustment. That interpretation would require evidence that the relationship is appropriate for your market and job type.

What if the shop has very few jobs? Report counts alongside percentages and preserve several comparable periods. If the records are sparse, the honest result may be that the agency effect cannot yet be separated from other changes.

Sources

Checked October 8, 2026. The review method is our framework. All comparison figures are invented, and neither source endorses a jobs-per-degree-day model.

• National Weather Service: What Are Heating and Cooling Degree Days, for the 65°F base and daily calculation.
• Google Ads Help: About conversion measurement, for the definition of configured conversion actions.