HVAC leads fell after hot weather, but Google Ads spend did not. Is demand shifting or is the campaign failing?

Start with the service people wanted, not the account total. Hot weather ending may reduce cooling searches, but it cannot by itself explain every lead that went missing. A lower bid can make an ad appear less often even if search demand holds steady. And if people still click a cooling ad but stop calling, the question has moved from reach to what happened after the click.

Keep an efficient Search presence for services that still produce worthwhile jobs. Move Search budget toward heating or maintenance only when those searches and the business's capacity justify it. Consider an outside-Search option only if it is available and has a credible route to customers. Money without a good route can wait. It does not have to attend every auction just because it attended the last one.

The same $2,000 can buy a different week

Suppose an HVAC shop compares two illustrative, equally long, fully matured periods in the same service area. The later period follows hot weather and a cooling bid reduction. These are invented numbers to show the diagnosis, not a report from an account:

Scroll sideways to read every column. With a keyboard, focus the table and use the arrow keys.

Spend held steady because fewer clicks cost more on average and the mix changed. A lower cooling bid can limit reach without making the clicks still won cheaper; their later $14 average CPC does not tell us why it rose. For cooling, qualified calls per click fell from 16/160, or 10%, to 5/100, or 5%. Cooling cost per qualified call rose from $100 to $280. Heating and maintenance remained at 10%, or $100 per qualified call. Those are two different problems hidden in one account total: fewer cooling clicks and fewer calls from each cooling click.

The table does not prove why cooling clicks fell. The bid reduction is one candidate. Fewer local cooling searches is another. More competitive auctions, a changed query mix, or a changed ad position can contribute. It also does not prove the five cooling calls were poor jobs or that six maintenance calls were good ones. The dispatch book, not the column named “Conversions,” decides that.

Separate demand, reach and response

Use the same geography, weekdays, hours, campaign type and service intent for both periods. Write down the bid and budget change dates first. Compare cooling repair with cooling repair, heating with heating, and maintenance with maintenance; split brand from nonbrand when the account permits. A branded call from a returning customer should not quietly become evidence that nonbrand cooling demand survived. If heating and cooling share a campaign, classify the visible search terms and call outcomes before relying on its total. Google’s search terms report shows searches that triggered ads, but omits some low-volume queries, so a visible-query sample is not the whole market.

Then ask three different questions:

1. Did local interest in this service change? Compare independent evidence for the same service and place: local intake or organic enquiries, appropriately filtered Google Trends interest, Keyword Planner historical search estimates, and actual local weather if it helps explain the service. Trends uses a sampled, 0–100 relative scale normalized by time and place; it is not a count of searches or jobs. Planner’s historical view is generally monthly and depends on location and network settings. Neither diagnoses a campaign alone. A warmer September in one market need not resemble a colder one elsewhere.
2. Did the campaign still reach those searchers? Read impressions, clicks, average CPC, bid/budget changes and Search impression share at a comparable scope, when reported. Impression share divides your impressions by estimated eligible impressions. Google says bids can change the auctions used in that estimate. Therefore, after cutting bids, fewer impressions—even a changed share—cannot be turned directly into “searches disappeared.” Auction Insights, when it has enough activity, can reveal a changed competitive picture, but it covers auctions in which you participated, not every local search. Check the actual bid strategy before using a loss column: Google specifically says Search lost impression share (budget) is incompatible with Maximize conversions. A blank or unsuitable metric is not zero lost opportunity.
3. What happened to reachable enquiries? Compare qualified calls or booked jobs per click for each intent only after the outcomes have had time to arrive. Google Ads conversion lag can make recent clicks look worse before later actions are reported. Reconcile Ads conversions against call logs and dispatch/CRM outcomes. Google’s call details can show missed versus received calls where its call reporting is in use; a duration-based call conversion is not the same as a qualified HVAC job. If real calls held steady but tracked conversions fell, inspect measurement. If cooling clicks held steady but real calls fell, inspect the search terms, offer, landing page, phone path and answer rate. Seasonality can coexist with a broken phone path; it does not excuse one.

The two products are simple—spend = clicks × average CPC; qualified calls = clicks × qualified-call rate—but the factors are not interchangeable. Lower search interest can reduce available clicks. A bid cut can reduce reached clicks. Competition can change their price. A changed query mix or enquiry path can change their usefulness. The account total reports the result of all four, then cheerfully labels it “Performance.”

Some accounts cannot produce a reliable impression-share column or a clean service split. In that case, use the change log, the search terms you can see, separate phone/dispatch records, auction evidence where available and local interest signals. You can still decide whether the currently bought calls are worth their cost. You cannot claim to have isolated market demand from lost reach, or to know cooling's lead rate, if the relevant denominator or service label is missing. Start labeling new enquiries by requested service and source; that improves the next comparison, but does not recover the unlabeled history.

What the example would let you decide

In the illustrative shop, a fall in independent cooling interest would support part of the seasonal explanation. The cooling bid cut means its 160-to-100 click decline still cannot be assigned wholly to demand. The 10%-to-5% fall in qualified-call rate is a separate reason to examine query mix and the path from ad to answered, bookable call. If cooling search interest stays level while reach drops after the bid change, restore only the exposure you can justify economically before calling the season over. If ad calls remain steady in the phone system while Ads reports fewer conversions, repair the measurement before re-budgeting from that column.

Meanwhile, the later six heating and maintenance calls cost $100 each in the example. That is a reason to examine those intents, not an automatic $1,400 transfer from cooling. Compare their booked-job rate, contribution after service costs, available crews and likely next-week demand. A maintenance tune-up and an emergency cooling repair may have different immediate value and different follow-on work. If heating searches are rising locally, the shop can match an actual heating offer to them and give that Search intent enough budget to serve. If heating demand is not yet present, writing “furnace” into the ad is not a substitute for it.

The remaining spend has three honest destinations:

• Reduced Search: retain cooling, brand or other service queries that still bring valuable jobs at an acceptable cost, with budget and bids matched to current volume. Review again when weather, query demand or job value changes.
• An eligible outside-Search route: if the shop has an owned customer list and appropriate permission, maintenance or plan-renewal outreach can address customers who need service before they search. If Local Services Ads serve this category and place and the business passes their verification, they are another option to evaluate against qualified jobs, not merely cheaper leads. They may compete for some of the same search demand; their existence does not prove extra demand.
• Reserve: if Search is not producing acceptable work and neither alternative has access or a credible audience, hold the unused amount for a later weather change or a better-supported opportunity. Reserving is an allocation decision, not a campaign failure.

This is also why a routine cooling-to-heating transition is a poor reason to reach for a Smart Bidding seasonality adjustment. Google describes that control as an expected conversion-rate change for a short exceptional event, typically one to seven days; it says Smart Bidding already handles ordinary seasonal events. A different mix of services over weeks is a different problem. Google’s seasonality-adjustment guidance

The practical verdict is deliberately conditional. If local cooling interest and valuable jobs both persist, keep buying the reachable portion efficiently. If cooling interest fades while heating or maintenance demand and capacity appear, move toward the service people now want. If the evidence is thin, maintain only the spend you can defend and keep the rest available. There is no need to declare that pausing erases Quality Score or that September has the same thermostat setting in every town.