Have you ever asked yourself why spending more on a converting keyword causes Google Ads performance to drop? Or, why didn’t it result in additional conversions?
When I find converting search terms and add them as positive keywords, why do they not perform anymore? They don’t convert, or they might convert but at sky-high prices. Then when your client or boss asks you for forecasts, you sit there wondering what you’re meant to do.
If you’ve ever had trouble scaling Google Ads, there’s probably a good reason. And it’s usually one of a few things. Let’s work through them and understand how Google’s auction and associated machine learning works.
How the Google Ads auction really works
With the situation of spending more, let’s consider Smart Bidding. When we’re looking at an account, particularly over a period of time, what we see is the smooth average cost per conversion. But if you look at actual individual conversion costs, they’re up, down, and zigzagging all over the place. This is where some of the problems start.
This zigzagging is a product of the search market. How many people are out there at any given moment physically looking for the thing that you’re advertising? And of course, there are your competitors and how they impact the auction. You can’t really control any of this, and it’s part one of the problem. It’s why sometimes, bidding or spending more doesn’t have the impact that you desire. And that’s why things are very chaotic.
You might even have situations where you up the bid and budget, and you don’t get more impressions. Or you get more clicks, but you don’t see the conversions come through. It can be confusing if you double the budget but impressions do nothing.
That’s because the auction isn’t linear. There is a sweet spot—a moving target, just to make things harder. And this moving target is dramatically different depending on the market. If you’re operating in a local area, like for plumbing services, the target is less dynamic and the auction remains quite stable. For wider markets—something rather large like finance—the auction can be very volatile.
But none of this is linear, so when you bump your budget 10% while getting 100 conversions, don’t expect 10 more. It doesn’t always work that way, unless you’re in the sweet spot or “easy street” of the auction. This is where you’re nibbling around the lower regions of impressions and, therefore, cost.
The cost is lower down there because the bid engine can be selective. So if there are a million impressions and you only need 10,000 to hit your target (budget divided by conversions), Google can be very selective with their automated bidding. It can drop you in and out as needed, and essentially cherry-pick leads for you. When you’re at the bottom 20% of the market in terms of impressions available, it’s very easy to have success in advertising.
At the top 20%, it’s much harder because there is no cherry-picking. When you’re at the maximum of the available impressions, the maximum of the available market, you are literally saying to Google through the bidding system, “I want to be there every single time. Doesn’t matter if it’s inefficient or if that person is not right for me.”
And this is where you can find that inflection point for many accounts where you might want 100% impression share—you might even have the budget for it—but you start to get over the 70-90% range and cost per conversion goes through the roof or ROAS tanks. There might be a couple of more conversions up there, but they simply aren’t efficient or worth the spend—unless you’re a big advertiser where volume is the concern and it’s an aggressive approach.
But most advertisers aren’t able to pay anything for a conversion when dealing with a very chaotic piece of technology—and human beings, who do whatever they like.
How to curry auction-time favour with Google
So how do you fix this? What do you need to do to get past the sweet spot and spend more, or get more conversions?
From Google’s perspective, they want more click-through rate. It’s as simple as that, and it’s never going to change. If they put an ad in front of someone, they want the click because that’s how they make their money. So all this fancy stuff about ad rank, quality score, audiences, all this other good stuff—it comes down to click-through rate.
It’s business sense for Google, so increasing click-through rate means better ad copy, which will show favourability in the auction. And negative keyword management simply cannot be skipped. Even with today’s automations, it’s not optional.
The harder way, of course, is to improve your conversion rate, but that’s rarely done inside a Google Ads account. Your website’s job is to convert people. But an incremental gain in conversion rate is another option.
This free tool will scrape the search results for a specific keyword or location, and then it’ll output that to a sheet for you with the 12 factors that you can use to . In addition, there is free training on the back of this—probably the best 25 minutes you will spend without paying.
How Google has made Broad Match viable
What about search terms that don’t convert as well after you add them as keywords? They spend more money, but the cost conversion goes up because the conversion volume isn’t there. This is what I call the Broad Match Paradox. It explains why Broad Match keywords outperform Exact Match in many cases.
What’s happening (particularly with Smart Bidding) is that Google has way more data and way more data points in the auction, which is why it’s better than manual bidding. Manual bidding is dying at death, and we have to trust that this machine learning is amazing and does the job for us. But this is actually a glimpse into it doing its job.
So you’ve got your Exact Matches, your Phrases and your Broad. This is more typical behaviour of a Broad. You’re seeing all these search terms come in, some of them converting, and you shift to an Exact Match keyword in its own ad group. But it doesn’t work, or it doesn’t work as well, or the cost per click is way higher.
This is the automation dipping you into the auction at the right moment. Google has got all the data. They say one person is more likely to convert against that search term at this moment in time, but the next person to search for that might not convert.
And the problem is there are 10,000 people behind them, and Google knows right now that 17 of them are going to convert against these data points. It’s not exactly like that, I’m being explicit for the sake of the example, but you completely ruin that when you add the keyword as Exact Match i.e. wanting to target the keyword regardless of Google’s data points.
And this is why I have been favourable towards Broad Match teamed with Smart Bidding i.e. Target ROAS and CPA. It works; it uses data that we don’t have, signals that we don’t have access to. It once again cherry-picks for us, even in a busy auction where you might be halfway to the top. You can’t really do that on your own.
You could start layering audiences and doing some fancy stuff. But quite frankly, are you really going to beat the machine? I don’t think you are.
Navigating campaign cost reduction
If you can’t afford to pay for more clicks with a higher propensity to not convert, then you need to reduce costs. And you can only do that with a higher click-through rate, a higher converting landing page, and negative keywords—not to mention optimisations like location bidding, segmenting by location, devices, audiences, and so on.
But it’s not an option for everyone because if your market’s not sufficiently big enough, all you’re going to do is limit the data in the account by over-segmenting, which would then hinder the Smart Bidding and you’re in a catch-22.
Don’t assume because Google is giving you a certain number of impressions that there are a million people available for you to get in front of, at a cost per conversion that suits your or your client’s business. Understand how Broad Match works now, especially with Smart Bidding, and use them together to maximise the cherry-picking.
A caveat about keyword match types
Don’t misconstrue me: I’m not saying you should stop using Exact Match keywords. Those that convert will typically outperform their Broad counterparts; the same goes for Phrase. The issue with Exact is that you are limited by impressions. For a lot of people, this can become a problem.
If you’ve got a tiny account, then you can live quite happily on Exact Match and manual bidding alone. Rule of thumb for me is less than $1,000 spend a month or 10-20 conversions at most. You may find small accounts spending a few thousand a month can live on Exact, too. But typically, they start to see restrictions in the volume.
If you need more volume, this is where you have to start leaning on Broad. And this is where the Broad + Smart Bidding pairing gives flexibility in the form of cherry-picking that additional volume.
How to forecast Google Ads growth
I want to leave you with some context around why forecasting is a fool’s errand, so you can explain it to your clients or your boss.
Can we do it? Of course we can. But we simply cannot second guess how a budget or bid increase will impact impression share, or how we fare in the auction. This all goes out the window if impressions drop as well as that makes things more competitive e.g. fewer people looking, supply and demand, or if a new competitor turns up. All of these put more pressure on the auction.
Projections are very difficult, indeed. Predicting the future is impossible. Just look at the financial markets. The professionals get it wrong for a living and they get paid to do it. So when your client and your boss ask you for projections, the safest thing to do is to exaggerate the cost impact.
By that I mean assume that your cost per click is going to go up by factor of two for every increment in the auction. If you’re at 50% impression share and you want to get 10% more increment, project the CPCs for that increment at +20%.
To make the maths really simple, let’s say you’re currently getting 1,000 impressions a week and 10% over that is 100 more impressions. You might get two conversions, and your CPC might be $5. Don’t assume it stays $5; assume it will go to $6. That sort of approach—a scaling exponential scale of cost to take part in the auction to gain those impressions—has served me well for many, many years.
At the end of the day, you want to either deliver what you promised, or under-promise and over-deliver. You might scare some people off from spending more money, but it’s better to play it safe.
That’s particularly true if it’s not your money.
