Most PPC accounts I look at don't have a traffic problem. They have a waste problem. Spend is spread across hundreds of loosely matched keywords, ad groups nobody has opened in a year, and campaigns built for a version of the business that no longer exists. When I overhauled the Google and Microsoft Ads accounts at Marshall Minibus, the gains didn't come from a clever bidding trick. They came from an audit, a new structure and a lot of stripping back. The account ended up lean and built around long-tail search, and cost per lead came down by about 35-40%. This is the approach, and how you can apply it to your own account this week.
Why cost per lead creeps up
CPL rarely jumps overnight. It drifts. A few broad keywords get added to chase volume. Google's match types get looser every year, so a keyword you added for one intent starts triggering on five others. Someone splits campaigns by region or device for a reporting reason, and suddenly each campaign has a thin trickle of conversions that smart bidding can't learn from.
The other quiet cost is generic head terms. A two-word search like "accountant London" could come from a business owner ready to switch, a student researching careers, or someone looking for a free tax calculator. You pay the same for each click. My rule of thumb: if a search term could plausibly be typed by someone who will never buy, you are paying for curiosity, and curiosity is expensive.
So the job is twofold. Spend less on searches with mixed intent, and concentrate your data so the algorithm has something useful to work with.
Start with an audit, not a rebuild
It's tempting to rip everything out and start fresh. Don't, at least not yet. The existing account holds months or years of data on what actually converts, and you want that before you design anything new.
Here's the order I work in:
- Check conversion tracking first. This is broken more often than anything else. Make sure a lead is counted once, that phone calls are tracked if calls matter, and that you aren't counting page views or newsletter sign-ups as leads. If tracking is wrong, every number after this is fiction.
- Pull the search terms report for the last 90 days. Sort by spend. Tag each term as converting, relevant but not converting, or irrelevant.
- Find the concentration. You will almost always see that a small group of search terms drives most of your leads, and a long list of terms eats budget with nothing to show for it.
- Look at the edges. Location, ad schedule and device. Leads at 2am from a region you don't serve are a sign of settings, not demand.
By the end you should know which intents produce leads, which ones just produce clicks, and roughly how much money is leaking. That leak is your opportunity.
What lean and long-tail actually means
These two words can sound contradictory. Long-tail suggests lots of specific keywords. Lean suggests fewer moving parts. The trick is that you use long-tail keywords inside a small number of well-defined campaigns.
Long-tail means specific, multi-word searches with clear intent. Compare "boiler" with "emergency boiler repair same day". The second has far less volume, but the person typing it has a problem and wants it fixed now. These searches tend to have less competition, so clicks are often cheaper, and the conversion rate is usually much higher. That combination is exactly what pulls CPL down.
Lean means you consolidate. Instead of forty campaigns with a handful of conversions each, you might have five or six, each built around a genuine buying intent and each with enough conversion volume for bidding to work properly. Fewer campaigns also means you can actually manage them. An account you can review properly in an hour a week will beat a sprawling one you only half understand.
Building the new structure
Once the audit tells you which intents convert, the build is fairly mechanical.
Group by intent, not by keyword. Each campaign should map to something a customer is trying to do. Service type, urgency or customer type are usually better dividing lines than geography or keyword theme.
Size campaigns to your budget. If a campaign can't realistically get a couple of dozen conversions a month, smart bidding is mostly guessing. Merge it with something related until it can.
Use tight match types for proven terms. The long-tail searches that converted in your audit go in as exact and phrase match. Broad match has its place, but treat it as a controlled experiment rather than the default.
Build negative keyword lists early. Shared lists for the obvious waste: jobs, careers, free, DIY, training, courses, reviews, plus anything specific to your sector that showed up in the search terms report. Add to them every week.
Match ads and landing pages to the intent. If someone searches for a same-day service, the ad should say same day and the landing page should open with it. A generic homepage throws away much of the advantage a specific search gives you.
Be prepared for total traffic to fall. That's fine. You are swapping cheap-looking volume for clicks that turn into leads, and the metric that matters is cost per lead, not cost per click or impressions.
Test the automation on your terms
Google will push you towards automated formats, and some of them are worth trying. At Marshall Minibus, testing dynamic search ads and Performance Max was part of the overhaul. The point was to see what each one added on top of a clean, lean core, rather than handing the account over to them.
That's how I'd approach it on any account. A few rules I stick to:
- Run automated campaigns as tests with their own budget, alongside your core search campaigns, so you have a clear comparison.
- Exclude your brand terms where you can. Performance Max in particular is very good at picking up people who were going to find you anyway and claiming the credit.
- Check what you can see of the search terms and placements. If the format is finding the same leads your search campaigns already get, it isn't adding anything.
- Judge on lead quality as well as lead volume. Automated formats can find a lot of cheap form fills. Cheap leads that never answer the phone are not cheap.
Give tests long enough to get past the learning period, then be honest. If a test isn't beating your core campaigns on cost per qualified lead, turn it off and move the budget back.
Keep it lean, and cut what doesn't earn its place
Structure decays if you leave it alone. The same forces that bloated the original account will start working on the new one the day you launch.
That applies to channels too. The Marshall Minibus overhaul covered Google and Microsoft Ads, and we later cut Bing. As a general rule, a second search network has to justify the management time as well as the media spend. If it delivers a small share of leads and needs its own tracking checks, negatives and testing, that time is often better spent sharpening the main account.
A simple weekly routine keeps things tight:
- Review the search terms report and add negatives.
- Pause keywords that have spent meaningfully without converting, using your target CPL as the yardstick. A keyword that has spent two or three times your target CPL with no leads is usually ready to go.
- Move budget towards the campaigns with the lowest cost per qualified lead.
- Check in with whoever handles the leads. Sales or customer service will tell you which enquiries are real long before the dashboard does.
If you can, feed lead quality back into the platforms through offline conversion imports. Bidding towards leads that actually became customers is one of the most effective ways to reduce cost per lead in PPC over time, because you stop optimising for form fills and start optimising for revenue.
FAQ
How long does it take to see a lower cost per lead after restructuring?
Expect a few weeks of noise while bidding relearns on the new campaigns. Most accounts start to settle within a month or two, assuming conversion volume is reasonable. Don't judge the restructure on the first fortnight, and avoid making big changes during the learning period.
Won't long-tail keywords limit my lead volume?
Sometimes total volume dips at first, but the leads you lose are usually the poor-quality ones. Long-tail searches add up, and because they convert better you can often afford to bid more competitively on them. If volume is genuinely too low, widen carefully with phrase match and controlled broad match tests rather than going back to generic head terms.
Should I use Performance Max for lead generation?
Test it, but don't build your account around it. Run it alongside a clean search structure with its own budget, exclude brand where possible and judge it on qualified leads. If it can't beat your core campaigns on cost per qualified lead, it doesn't deserve the spend.
Cutting CPL is rarely about finding a secret setting. It's about knowing which searches make you money, building a small, tidy account around them, and having the discipline to keep it that way. If you'd like help doing that, I take on consulting work in PPC and wider growth marketing, and I'm also open to Head of Marketing and Head of Growth roles, particularly with DTC and consumer brands. You can get in touch through the contact link on tomosborne.me.