The phone rang until ten. The ad budget was gone by five. Nobody had put those two facts next to each other

Who this is

Barbosa has been doing plumbing and air conditioning in Dallas since 1979. Two licences, TACLA002374 and M-37241. The first of those is a four-figure number in a state that now issues six, which is the shortest way to say how long they have been at it.

Forty-six years in one metro produces something no amount of budget buys: a very large number of households who already have the number written down somewhere. That is the asset. It is also what made the problem invisible.

Two logs that nobody had put side by side

The call log said the busiest stretch of the day started around half five and ran to about nine. That is not surprising. People discover a water heater has failed when they get home, not at eleven in the morning.

The spend log said the daily budget was substantially gone by mid-afternoon. Not because anybody chose that — because a daily budget spends itself in the order the auctions arrive, and the auctions arrive all day.

Both logs had been sitting in the same account for years. Neither is wrong. They had just never been read on the same screen.

The grey bar is money and the coloured bar is demand. When they point in opposite directions, the account is buying the quiet hours at full price and rationing the busy ones.

Why an account runs out of money before the rush

A daily budget has no opinion about what time it is. It answers auctions as they come, and the morning has plenty of them: people at a desk, mildly worried, typing the same words as the person who will call at seven with a flooded utility room.

Those two searches look identical in a keyword report. One of them ends in a booked visit at about twice the rate of the other. Nothing in the platform separates them, because the platform cannot see the water on the floor.

What we got wrong first

Our first read was that this was a bidding problem, so the first version of the fix was a bid schedule: less in the morning, more from five. It helped, and less than it should have.

The reason took another three weeks to find. Some of the evening calls were being bought and then not answered — rolled to an after-hours service, or picked up by somebody who could take a message but not book a visit. We were paying a premium for the busiest hour of the day and handing a share of it to whoever the caller rang second.

So the second version of the fix was not in the ad account at all. It was agreeing which hours were genuinely covered, and only then buying them.

What happened to the evening

Fewer calls in the morning, more in the evening, and a higher share of them turning into a visit on the calendar. The interesting row is the middle one: the booking rate moved without anybody changing the script.

That is what re-pacing does. It is not persuasion. It is arriving at the moment the person is already convinced.

The order things happened in

The sequence is the part worth copying. Every step before week six was measurement, and the money did not move until the hours were confirmed by a person rather than inferred from a chart.

What this costs a company that is already busy

Almost nothing, and that is the uncomfortable part. There was no new budget in this. The same money arrived at a different hour.

Which means the previous arrangement had been costing this business real work for a long time, and nothing in any dashboard was ever going to flag it, because no dashboard knows what time the office stops booking.

What this does not prove

• Four months either side includes a Texas summer on one end and not the other. Evening demand is not flat across the year and some of this movement is the calendar.
• 'Calls that rang out' is measured from the tracking line, so a caller who hung up and immediately rang the office direct is counted as a miss here and as a customer in the business. The real figure is better than ours.
• The re-pacing and the Local Services Ads schedule shipped three weeks apart and we did not hold one back. They are reported together because we cannot honestly separate them.
• We capped morning spend rather than removing it, so this does not prove morning traffic is unprofitable. It proves it was over-funded relative to the evening at this budget.
• The biggest single change was staffing coverage, which is the client's doing and not ours. We found it; they fixed it.

What to take from this

1. Put the call log and the spend log on one screenBoth exist in most accounts already, and the gap between them is usually the cheapest improvement available — it costs no extra budget, only a different hour.
2. Buy the hours somebody books visits in, not the hours somebody answersA message taken at eight is not a booked job, and paying peak prices to generate messages is the most expensive way to be polite.
3. Cap the morning rather than cutting itMorning searches produce the planned, higher-margin replacement work. They are worth buying calmly, which is a different instruction from worth buying first.
4. Make every channel agree about your opening hoursWhen search and Local Services Ads run different schedules the account contradicts itself, and the caller experiences that contradiction as a phone nobody picks up.

Questions people ask about this one

Can you not just set an ad schedule and be done with it?

You can, and it will get you part of the way. A schedule changes when the money is spent; it does not change whether a person is there to book the visit when the call lands. In this account the schedule alone produced a modest improvement, and the rest only arrived once the covered hours had been confirmed with the office rather than assumed from the chart.

Why is the evening cheaper when it is the busiest time?

Cost per click was not cheaper. Cost per booked call was, because evening callers book at a much higher rate — they are ringing about something that has already happened rather than researching something that might. Paying more per click for traffic that books twice as often is a lower cost per job, which is the only one of those numbers a contractor banks.

Does this apply to a company that is not open in the evening?

It applies more. If the office closes at five, the correct response is to stop buying the six o'clock auction, not to buy it and hope. The method is the same either way: find out which hours actually convert into booked visits, and then decide whether to buy them or to extend the hours. Both are valid. Buying hours you do not cover is not.

How do you know a call was missed rather than just short?

Each tracked call carries a disposition from the recording and the dispatch record — answered and booked, answered and not booked, rolled to the after-hours line, or unanswered. That is more work than reading a call count, and it is the only way to tell an expensive success from an expensive failure.