The Accounts Birmingham Agencies Turn Down: A Guide to Urgent Local Service Clients

You are three weeks into a locksmith account covering Birmingham and the Black Country. On your screen the picture is defensible: click-through rate above vertical average, impression share climbing, average position strong, cost per click roughly where you forecast it.

Then the call comes.

➔ "I've had two jobs this month. Two."

➔ "The bloke in Kings Heath doesn't spend half what I spend and he's flat out."

➔ "What exactly am I paying you for?"

You have the report open in front of you and every number on it is good. You know the campaign is built properly. And you also know that none of that is going to matter in this conversation, because you are being measured against a phone that did not ring.

The structural problem: in most verticals you and the client are looking at roughly the same reality through different lenses. In urgent home services you are often looking at two genuinely different realities — and the gap between them usually has a name.

There is a pattern in this vertical that rarely gets said out loud: a lot of agencies have stopped accepting these clients altogether.

It is not that the work is uninteresting, or that the clients are difficult people. It is the sheer volume of click attacks these accounts absorb. They take more hostile traffic than anything else on a typical roster — and past a certain level, the account simply stops responding to management.

You can structure the campaign properly, write strong ads, build tight keyword lists, set sensible bids. The budget still drains before the leads arrive, because a meaningful part of it is going somewhere you never sent it. Do everything right and the account still underperforms.

That is an unwinnable brief, and it turns the work into a permanent firefight: pulling reports, updating exclusion lists, rebuilding budget pacing, explaining another bad month. Hours every week that are not strategy, not creative, and not billable in any way the client recognises — spent on the one account most likely to leave anyway.

Faced with that, agencies do the rational thing. The category quietly moves onto the "we don't take those" list.

Why these categories attract the most attacks

➔ The auction is tiny and local. A handful of operators competing over the same postcode, all budget-limited, all aware of each other.

➔ Winner takes the whole job. There is no second place in an urgent call — the customer rings the first name that answers and never returns to the results page.

➔ Exhausting a rival's budget is trivially easy. No software, no skill, no cost. Just clicks.

➔ Budgets are small enough to exhaust. In a category where a click can cost as much as a small job's margin, a handful of clicks removes a competitor from the auction for the rest of the day.

What that refusal actually costs the agency

Turning down the category is a rational response to an unsolvable problem — but the problem is only unsolvable without measurement, and the category is a genuinely good one to serve:

➔ Steady monthly retainers with clear, immediate value when it works.

➔ Clients who need the channel permanently, not seasonally.

➔ Falling agency competition, precisely because others have withdrawn.

➔ Owners who talk to each other constantly — in these trades, a referral network moves faster than any outbound campaign you can build.

The attack volume is what drives agencies away — and it is also the only variable here that can actually be removed. Everything else about these accounts is structural: the budgets stay small, the CPCs stay high, the client keeps judging in days rather than quarters. The hostile traffic is the one thing that does not have to stay. Take it out and what is left is a demanding account, not an unmanageable one.

Which is why campaign management alone is no longer the whole job

This is the conclusion most agencies arrive at eventually, usually after losing two or three of these clients: in this vertical, running the ads is only half of the service.

Every lever you have assumes the traffic is real. Bid adjustments assume the clicks represent demand. Match types assume the search terms came from buyers. Landing page work assumes the visitor intended to arrive. Smart Bidding assumes the engagement signals mean something. When a substantial share of the traffic is hostile, every one of those levers is being pulled against a distorted picture — and skill makes no difference, because the input is wrong.

So click protection stops being an upsell and becomes part of the deliverable, in the same way call tracking is not optional on an account judged by phone calls. You are not adding a product to the retainer; you are adding the instrument that makes the rest of the retainer work.

➔ Scope it in from day one, alongside call tracking and conversion tracking — not after the first bad month.

➔ Decide the commercial model deliberately (absorbed into the fee, or a separate line), but never leave it out of the proposal.

➔ Say what it changes. The honest promise is not "we will make this cheaper." It is "we will make sure your budget reaches real customers, and we will show you the proof each month."

That single addition is what turns a category most agencies have written off into one they can serve profitably — and it is why agencies that do serve it tend to keep these clients far longer than the ninety-day average.

It is worth being explicit about why these accounts behave differently, because it is not the client being unreasonable. The economics really are harsher.

The compounding effect

When an account only receives a handful of clicks per day, statistical noise and wasted spend stop being rounding errors. Lose three clicks out of twelve and you have not lost 25% of a budget — you have lost a quarter of that day's chances to win a job, on a day the client will remember.

The tracking gap you inherit

These clients usually arrive without the infrastructure the rest of your roster has:

➔ No call tracking, so the calls you do generate are invisible in your reporting.

➔ No conversion tracking, which means bidding algorithms are optimising on nothing.

➔ A website built by someone's cousin, with the phone number as an image rather than a tappable link.

➔ No CRM, so "did that lead become a job?" is answered from memory.

You are being held to an outcome standard while working without the instruments that measure the outcome. That is the actual job, and it is worth pricing accordingly.

Here is the conversation as it usually plays out. You report 240 clicks last month. The client says they got maybe eleven calls, four of which were wrong numbers or price-shoppers.

Both of you are telling the truth. So where did the rest go? The same gap, seen from the other side of the table, is what a repair business owner experiences as a budget that empties before lunch while the phone stays quiet — worth reading if you want the client's version of this conversation in their own terms.

The legitimate explanations

➔ Research intent. People comparing prices before deciding, or checking whether a repair is worth it at all.

➔ Landing page failure. A slow page, or a number they cannot tap, loses people who genuinely intended to call.

➔ Out-of-area traffic. Default location settings reach people interested in the area, not only people in it.

➔ Answering rate. In this trade, an unanswered call is a lost job — the customer rings the next name in the results within seconds. Many clients do not realise how many they miss.

The explanation nobody puts in the report

Then there is the portion that was never a prospect at all. Independent measurement has been sizing this for years: a study published in January 2026 by Lunio, based on 2.7 billion clicks across six ad platforms, estimated that 8.51% of paid ad traffic is invalid, amounting to roughly $63 billion of wasted spend in the study period. The same research found lead-generation accounts see materially higher invalid rates than ecommerce — and urgent local services sit at the sharp end of lead generation.

Other Google Ads-specific datasets report average invalid click rates in the region of 11.5%. Methodologies differ; the direction does not.

Why this matters commercially, not just technically: when you cannot quantify this portion, it silently becomes your fault. Every click you cannot account for looks, from the client's chair, like an agency that spent their money badly.

These categories are geographically dense in a way few others are. In a conurbation like the West Midlands, forty locksmiths or thirty recovery operators inside a fifteen-mile radius are bidding on near-identical keywords, and every one of them is budget-limited. The auction for locksmith Birmingham is effectively a closed room in which everyone knows roughly who else is in it.

Add the seasonal spikes and it sharpens further. September brings the student intake around Selly Oak and Edgbaston, lockout calls climb, and every operator raises bids in the same fortnight. The demand is real — but so is the incentive to make a rival's budget disappear before the peak hours arrive. That specific behaviour has a name and a signature: competitor click fraud.

That density produces a specific, well-known dynamic. Because these accounts run out of budget mid-day, a competitor does not need to outbid your client to remove them from the auction — they only need to help the budget run out sooner. It requires no technical skill and no software.

What it looks like in the data

➔ A single IP address clicking repeatedly across a week, always in working hours, never converting.

➔ Sessions lasting two to four seconds with no scroll and no tap.

➔ Clusters of clicks at hours when nobody in that trade has an emergency.

➔ Repeat activity that resumes on a new address shortly after you exclude the old one.

To be fair about it: not all of this is competitors. Bots, scrapers, accidental in-app taps and click farms make up a large share of invalid volume. But in dense, high-cost local categories, competitor-driven clicking is the part of the picture that every practitioner recognises and almost nobody documents.

Most agencies discover the problem eventually and start handling it by hand. It works for one account, for a while, and then the model breaks.

The mechanics of what you are defending against — bot traffic, click farms, rotating addresses, competitor activity — are covered in more depth in the click fraud protection guide. What follows here is why the manual version of that defence collapses at agency scale.

The per-account ceiling

➔ Exclusion lists are capped per campaign, so you cannot simply keep adding.

➔ Exclusions apply going forward only — they do not recover what was already spent.

➔ Mobile and VPN traffic rotates addresses within minutes, so yesterday's list defends against yesterday.

The portfolio ceiling

Now multiply the manual work by thirty accounts. Pulling reports, spotting patterns, updating lists and documenting it for the client is hours of skilled time per week that is not strategy, not creative, and not billable in any way the client understands. It is also the first thing dropped when the month gets busy — which is exactly when it is needed.

The optics problem nobody enjoys

There is an uncomfortable structural detail in the percentage-of-spend model: waste technically inflates the number the fee is calculated on. Honest agencies find this genuinely awkward, because it makes a fair conversation about wasted budget look self-serving.

The way out is not to avoid the subject. It is to measure it, report it, and be the party that raised it first.

The practical shift is smaller than it sounds. You add one section to the monthly report — clicks identified as invalid, blocked, and the spend that would otherwise have gone to them.

1. The blame conversation becomes a shared problem

"Here is where the missing traffic went, and here is what we stopped" is a fundamentally different meeting from "the metrics look good, I'm not sure why the phone isn't ringing." You move from defending to diagnosing, alongside the client rather than opposite them.

2. Your own optimisation gets sharper

Cleaner conversion data means bidding algorithms stop learning from noise. As click-through and conversion rates recover on genuine traffic, Quality Score improves — and a higher Quality Score reduces what the account pays per click on the same keywords. The drop is indirect, but it is real, and in this vertical it is material.

3. Retention improves where it hurts most

These clients churn fast because they judge fast. Buying yourself an extra sixty days of credibility during the period when the account is still maturing is often the entire difference between a client who stays two years and one who leaves in the first quarter.

4. It becomes a pitch differentiator

Very few agencies pitching a locksmith walk in with a click quality audit of the prospect's existing account, and protection built for service-led businesses makes that audit something you can produce before the first meeting rather than after the third month.

Almost every competing agency walks in with the same slide about keyword research. Turning up instead with evidence about where the prospect's current budget is actually going is a materially different conversation.

Worth saying plainly: none of this fixes a client who does not answer the phone, or a service area that is too small to support the spend. It removes one specific, measurable category of loss — and it lets you prove you removed it.

If you take on this kind of account, front-load the following. Every item below is cheaper to do in week one than to retrofit in month four.

➔ Call tracking before launch. Without it you will spend the engagement arguing about a number neither party can see. Where the job is booked by phone rather than on the site, feeding those calls back as offline conversions is what lets bidding optimise toward booked work instead of raw clicks.

➔ Answer-rate baseline. Ask what happens to a call at 9pm on a Sunday. Set the expectation early that missed calls are lost jobs, and that this is inside their control, not yours.

➔ Location targeting set to presence, not presence-and-interest.

➔ Network performance separated from day one. Report Search, Search Partners and Display independently so budget decisions can be made on evidence rather than instinct.

➔ Gradual channel scaling. Where invalid rates are high and returns are weak, reduce budget first and reallocate — then reopen incrementally as efficiency improves. A channel that earns its way up is more durable than one switched on at full budget and switched off in frustration.

➔ Click quality monitoring from week one, so you have a baseline before the first difficult conversation rather than after it.

➔ Reporting framed in jobs booked, with clicks and CTR as supporting detail rather than the headline.

➔ A written expectation about ramp time. Say out loud, in the proposal, that the first six weeks are for gathering data.

Everything above points to the same practical conclusion: at agency scale, this has to be automated. That is why a whole product category exists for it — ClickCease, ClickSambo, Lunio, TrafficGuard and Clixtell among them. We build one of these tools, so treat what follows as a description of the category rather than a neutral verdict on it; the criteria matter more than the badge.

Why it belongs under optimisation, not security

Agencies tend to file click protection under risk or compliance, which undersells it. In practice it behaves like an optimisation tool, because it improves the quality of the inputs every other lever depends on:

➔ Bidding gets better data. Smart Bidding stops learning from engagement that was never human.

➔ Reports become decisions. Network, hour-of-day and geography data mean something once the noise is filtered out of them.

➔ Budget pacing stabilises. Accounts stop emptying at unpredictable times, which makes daily budgets behave the way you planned them.

➔ Quality Score recovers. Genuine click-through and conversion rates lift the score, and the score lowers what the account pays per click.

None of that is defensive. It is the same work you already do to make an account perform, applied to the one input you previously could not clean.

What to evaluate when choosing one

For a roster of small, high-CPC local accounts, these are the questions that separate the tools:

➔ Multi-account management. Can you run thirty client accounts from one panel, with per-client rules — or are you logging in and out all day? For an agency this is usually the deciding factor.

➔ Detection depth. IP threshold rules are the baseline. Device fingerprinting that survives address rotation, plus behavioural and VPN/proxy detection, is what separates a tool from a list.

➔ Blocking speed. Exclusion at click time, not in a nightly batch. On a twelve-click day, a batch delay costs the whole day.

➔ Client-ready reporting. Something you can put in front of the client, ideally exportable or white-labelled. A tool whose output never leaves your screen does not solve the credibility half of the problem.

➔ Evidence for refund claims. Timestamped click logs with geography and behaviour, so an invalid-click claim is documented rather than asserted.

➔ Platform coverage. Search is the priority here, but check whether Display and social are included if your clients run them.

➔ Pricing against the account. On a small local budget, a tool that costs more than the waste it removes is not worth running. Match the plan to the click volume rather than the client's ambition.

The honest test: does the tool tell you something you could not have found yourself, fast enough to matter, in a format you can show someone else? If yes, it earns its place in the retainer. If it is only a list of blocked IPs, you have bought a slower version of what you were already doing by hand.