Ad copy that filters out the wrong customers

Nearly every ad in a local auction is written to attract. Free estimates, family owned, licensed and insured, serving the area since some year. All true, all generic, all inviting everybody.

Inviting everybody is affordable when clicks are free. In paid search you are buying each one, which makes an unqualified click a purchase you did not want.

Why would I deliberately reduce my clicks?

Because you are not buying clicks. You are buying customers, and the two have almost nothing to do with each other.

Start with what a click costs. LocaliQ, the agency arm of Gannett, publishes annual search advertising benchmarks under the WordStream name. Its 2026 edition, updated 1 June 2026, puts the average cost per click at $8.33 for Home and Home Improvement, $8.00 for Dentists and Dental Services, $9.87 for Attorneys and Legal Services, and $2.05 for Restaurants and Food. Across all industries the figure is $5.42. Two methodology notes travel with those numbers. The published averages are technically medians, and the disclosed sample for the 2025 edition of the same series was 16,446 US search campaigns with a minimum of 64 active campaigns per category. LocaliQ does not publish the exact sample size for the 2026 edition, and LocaliQ sells advertising management, so this is vendor research.

Now apply it. A click from someone who wants a job half your minimum costs you the same $8.33 as a click from your ideal customer. You pay identically and get nothing. Do that a hundred times a month and you have funded a competitor's education about your market.

The pushback is always the same. Would I not rather have the call and talk them into something bigger. Sometimes. Mostly you have spent twenty minutes on the phone and an afternoon driving to an estimate you were never winning.

Every unqualified click has two costs. The click, and the hour somebody spends on the lead it produced. The second is larger and it never appears in the ads report.

The arithmetic behind this is worth doing on paper before you write a word, since the conversion math decides whether ads make sense at all for your job values and close rate.

What does filtering actually do to the numbers?

Here is the worked version. Use your own click cost and close rate and the shape holds.

Before filtering. You spend $8,330 in a month on Home and Home Improvement clicks at the benchmark $8.33. That buys 1,000 clicks. At the LocaliQ conversion rate for that category, 8.05%, those clicks produce roughly 80 leads. Your cost per lead is $104 derived, which is above the published $90.92 benchmark and close enough not to worry about. Your team closes 25% of those leads, because a quarter of the callers wanted something you do. That is 20 customers, at a cost per customer of $417 derived.

After filtering. You add a minimum job size and a named service area to the ad. Click volume falls 30%derived to 700 clicks, so spend falls to $5,831. Leads fall further in absolute terms, to 50, because the people you filtered were disproportionately the ones who inquire and vanish. Cost per lead rises to $117 derived. Your close rate rises to 40%, because the people still calling wanted what you sell. That is 20 customers again, at a cost per customer of $292 derived.

Same twenty customers. $2,499 less spent. Cost per lead up 13%derived, cost per customer down 30%derived.

Run it with your own numbers. Spend divided by clicks, clicks times conversion rate, leads times close rate, spend divided by customers. If your cost per customer falls while your cost per lead rises, the filter is working exactly as designed, and any report that stops at cost per lead will show it as a failure.

What can I put in an ad that does the filtering?

Five things, in rough order of how sharply they cut.

Price or minimum. The bluntest and most effective. Naming a starting price, a minimum project size, or a typical range removes bargain hunters before they cost you anything. Most owners will not do it. The ones who do get quieter phones and better weeks.

Specialty. Not "plumbing services" but "repipes and slab leaks." A specific service repels the person with a different problem and attracts the one who has yours.

Geography. Naming the neighborhoods or cities you serve does what location targeting cannot, because targeting handles where the searcher is and text handles where the job is.

Customer type. Commercial only. Residential only. Property managers. Two words that eliminate an entire category of wasted conversation.

Constraint. Booking three weeks out. By appointment. Not taking emergency calls. Disappointing information delivered before the click is a favor to both of you.

The exception worth naming is emergency work, where filtering by urgency is the whole strategy rather than a limitation. How home services businesses capture emergency searches is a different discipline, and there the right filter is availability, not price.

Why do owners believe every local searcher is a buyer?

Because a statistic that says so has been in circulation for years, and it is not real.

You will see "88% of searches for a local business result in a call or visit" quoted in agency decks, and it appears on the LocaliQ benchmarks page itself, the same page that supplies the click costs above. No primary source supports it in that form. It is a downstream corruption of Google's local search behavior research, and the actual research says something much narrower.

The primary document is a May 2014 deck from Google with Ipsos MediaCT and Purchased. Its own methodology slide describes two studies: an Ipsos online survey of 4,500 respondents across nine vertical surveys fielded 10 to 22 January 2014, and a Purchased mobile diary run over seven days between 18 December 2013 and 30 January 2014 in which 3,431 smartphone searches and 2,262 store visits were logged. The figure it actually reports, verbatim, is that 50% of consumers who conducted a local search on their smartphone visited a store within a day, and 34% who searched on computer or tablet did the same. It also reports that 18% of local smartphone searches lead to a purchase within a day, against 7% for non-local searches.

Half of local smartphone searchers visited some store within a day. Fewer than one in five bought anything. That is the honest version, and it is twelve-year-old fieldwork on a search results page that no longer exists. Say both things when you use it.

That gap between searching and buying is the entire case for filtering. If 88% of the people who searched were about to hire somebody, writing a wide ad would be rational. They are not.

Does this hurt my quality score or my rankings?

Not in the way people fear.

Quality score is Google's diagnostic estimate of how relevant your keyword, ad and landing page are to the search. A specific ad that matches a specific search is more relevant, not less. What falls is click-through rate on searches you should not have been in, and that is the intended outcome.

For context on what normal looks like, LocaliQ's 2026 figures put average click-through rate at 6.47% for Home and Home Improvement, 5.66% for Dentists, 6.83% for Restaurants and Food, and 6.64% across all industries. If your rate drops below the benchmark after you add a filter, that is not automatically a problem. It is the measurement of people deciding, correctly, not to call you.

You may see impressions stay flat while clicks drop. That is the filter working. The searches you excluded still saw you, and some of them will remember the name when their job gets bigger.

What genuinely changes is your reporting. Cost per lead goes up and stays up. If you or your manager are judged on cost per lead, this will look like a failure for a full quarter. That is why you set expectations before you make the change, and why you need conversion tracking that reaches all the way to the outcome rather than stopping at the form. Tracking installed before the first dollar of spend is what makes this defensible later.

How do I write one without guessing?

Take the language from your own customers, not from your competitors' ads.

Open your last twenty won jobs and read how those customers described the problem when they first contacted you. Their words are your headlines. Then open your last twenty lost or declined jobs and find what they had in common. That commonality is your filter.

If most of your bad leads were under a certain size, name the minimum. If most were outside a radius, name the area. If most wanted a service you stopped offering, say what you do offer.

Then write the ad in that order. What you do, who for, where, and what disqualifies. Four short lines and no adjectives doing the work.

Do not expect to test your way to this, because your account cannot carry a test. Sample size follows the metric's variance and the effect you want to catch, not a traffic number. At 95% confidence and 80% power the constant is 2 × (1.96 + 0.84)², about 16, so conversions per variant is 16 divided by the square of the relative lift. Catching 20% takes 400 per variant (derived). You will not get there. Write the specific version and watch the leads.

Where does this break down? Two places. If your filter is wrong, you will not find out quickly, because the same low volume that prevents testing also prevents fast correction. And a filter that is too narrow can starve a campaign of the conversion data automated bidding needs, which makes the account worse in a way that looks like the copy failing. Cut one dimension at a time and give it a full month.

The people you filtered out are not gone forever. Some return when the job grows. That is the honest case for keeping a small presence in front of them, though remarketing works differently when your traffic is low and it is not a rescue for a badly targeted account.

What to do this week

Read your last twenty leads and mark each one qualified or not. Do it fast, by instinct. Then write down the single trait most of the unqualified ones shared.

Put that trait into your ad copy as a boundary. One line. Publish it this week.

Then watch three numbers for thirty days. Clicks, leads, and jobs booked. Expect the first two to fall. Judge the change entirely on the third.

If you are not running ads yet, do this exercise anyway. It tells you who you should be advertising to, and whether you should be running ads at all is a question worth settling before you write copy.

Questions about what your ads are inviting?

Email me at eric@seod.com with your current ad headlines and descriptions, pasted in plain text, plus one sentence on the job you actually want. I will mark which lines are inviting the wrong caller and rewrite two of them for you.

Takes me ten minutes and you keep the rewrites regardless. If the copy is already tight, I will say so rather than inventing work.

Or keep reading more on paid search for local businesses.