Because leads are being counted, not qualified. Most poor-quality-lead problems come from one of three places: the ads answer a different question from the one you want to be paid for, the form asks nothing that filters, or nobody feeds the outcome of each enquiry back into the account. Fix the feedback loop first.
The reason that ordering matters is mechanical. An automated system optimises toward whatever it is told to count, and if every form fill counts equally then every form fill is what it will go and find.
Here is how to define the problem precisely, then fix it in the order that actually changes the outcome.
Define a bad lead before you try to fix it
Poor quality is a feeling until it is a category. Ask five people in the business what a bad lead is and you will get five answers, which is why the fix so often misses.
| Type | What it actually is | Where the fix lives |
|---|---|---|
| Out of area | Real job, wrong place | Where the ads are shown |
| Wrong service | Real customer, work you do not do | What the ad promises |
| No budget | Wants the job, cannot pay for it | The page, and what it says about price |
| Not the decision maker | Tenant, agent, or somebody gathering options | The form, and the follow-up |
| Never answered | Possibly your best lead | Your phone, not your advertising |
Tag one week of enquiries against those five and the shape of the problem appears immediately. The last row is the one worth checking first, because a business that cannot answer the phone will describe every channel it ever tries as producing bad leads.
The account optimises toward whatever you told it to count
This is the mechanical heart of the problem. Automated bidding works toward the outcome it has been given. If every form submission is recorded as an identical success, then the system is being told, accurately and repeatedly, that a fifty-dollar job and a five-thousand-dollar job are the same thing.
An automated system will get you more of exactly what you told it to count. If you counted form fills, you will get form fills.
It will then do its job extremely well: it will find the cheapest possible form fills. Those come from the least committed people, because they are the easiest to reach. Nothing has malfunctioned. The target was wrong.
The fix is to change what is counted so that the outcomes differ — either by recording only enquiries that pass a qualifying step, or by attaching a value to each so that better outcomes weigh more. Both require the business to know which enquiries became work, which is the loop the next section is about.
Your form is doing none of the filtering
A form with a name, an email and a message box asks nothing that separates a serious enquiry from a passing one. Every enquiry it collects arrives identical, so nobody downstream can prioritise and no system upstream can learn.
One well-chosen question changes that. Not five — one. The right question is whichever one predicts a real job in your business.
- Where the job is, when your problem is enquiries from outside the area you travel to
- Which service, when your problem is people asking for work you do not do
- When it needs doing, when your problem is people who are researching rather than buying
- Owner or renting, when your problem is people who cannot authorise the work
Yes, this reduces the number of enquiries. That is the intended effect, and it is only a loss if you were measuring the wrong thing. Collect only what you genuinely need — adding qualifying questions means you are collecting personal information, and the sensible standard is to ask for the minimum that does the job. The wider version of this is the field-level hygiene work worth doing before any spend starts.
An automated system will get you more of exactly what you told it to count. If you counted form fills, you will get form fills.
Wrong-fit enquiries are usually a message problem
If the enquiries are real people wanting real work that you do not want, the ad and the page are answering a different question from the one you want to be paid for.
The search terms report shows what people actually typed before they clicked, which is regularly not what anyone assumed. Read a fortnight of it. The mismatch between the wording that brings people in and the wording of what you sell is usually visible within a page.
The second half is what the page says when they arrive. A page that leads with speed attracts emergencies. A page that leads with price attracts price shoppers. A page that leads with warranties, licensing and finished work attracts people planning a considered job. None of those is wrong; they just have to match the work you want. Saying plainly who you are for — and, more usefully, who you are not for — filters more enquiries than any setting.
This is where the paid channel and the site have to agree with each other. When they promise different things, the enquiry arrives confused and is recorded as poor quality.
Close the loop: send the outcome back
This is the fix that actually holds, and it is the one most businesses never do.
At the moment, the account almost certainly knows that an enquiry happened. It does not know whether that enquiry became a booked job, a quote nobody accepted, or a wrong number. Google Ads supports importing outcomes recorded outside the website precisely so the system can learn from what became work rather than from what became a form fill.
Nobody ever told the account which enquiries were worth having. It has been guessing, accurately, at the wrong target.
The requirements are modest: every enquiry needs an identifier that survives into wherever you record jobs, somebody has to mark the outcome, and the outcomes need to travel back on a regular schedule. The hard part is never technical. It is the habit of marking outcomes, every week, without being chased.
Once that loop exists, everything else gets easier, because the system is finally optimising toward money instead of activity. Setting it up is measurement work, and it belongs to whoever owns the numbers rather than whoever owns the advertising.
Nobody ever told the account which enquiries were worth having. It has been guessing, accurately, at the wrong target.
How to tell a lead-quality problem from a sales problem
Worth checking before spending anything, because the two produce identical complaints.
Take twenty recent enquiries you would describe as poor. For each one, answer two questions: was this person a plausible customer for work you actually do, and did anybody reach them within an hour.
If most were plausible and most were not reached quickly, this is not a lead-quality problem. It is a response problem, and no change to the advertising will touch it. Fixing what happens after the enquiry arrives will improve results faster and more cheaply than any account work.
If most were genuinely implausible — wrong state, wrong trade, wrong universe — then the problem is upstream and the three fixes above apply, in the order they are written.
Either way you now have a diagnosis rather than a complaint, which is the difference between a productive conversation with whoever runs your acquisition programme and another round of changing things and hoping.
Common questions
What counts as a bad lead?
Write the definition down before you use the phrase. Out of area, wrong service, no budget, not the decision maker, and never answered are five different problems with five different fixes. Grouping them guarantees the wrong one gets addressed.
Will adding questions to my form reduce enquiries?
Yes, and that is often the point. Fewer, better-qualified enquiries can be worth more than a larger number you cannot service. Add one qualifying field, not five, and only collect what you genuinely need.
Is it the agency's fault or mine?
Usually shared. An agency cannot optimise toward job quality it has never been shown, and the business is normally the only party who knows which enquiries became work. That information has to travel back before anything can improve.
How long before better data improves the leads?
Once outcomes are flowing back, expect several weeks before there is enough signal for anything to shift. It is slower than changing an ad, and it is the change that lasts.
Should I just pause and start again?
Almost never. Starting again discards whatever history exists and rebuilds the same problem with a clean interface. The target being wrong is not fixed by a fresh account; it is fixed by counting something different.
Questions and answers
Should sales feedback be included in the campaign review?
Yes, using clear categories and relevant context. The team handling requests can explain mismatches that a click or form total cannot show.
Can a longer form improve quality without helping customers?
It can add friction without resolving the actual mismatch. Keep questions that support service selection or response and remove those the team does not use.