So states Goodhart’s Law. A wise warning about unintended consequences and the dangers of a narrow focus on optimising for a particular metric.
In our metric-rich world of PPC, there are plenty of these dangers lying in wait.
Here are four examples – from the most straightforward, to the least intuitive…
To the uninitiated (and this includes plenty of PPC clients) CPC looks like a reasonable KPI.
- Pay less per click…
- Get more clicks for the same cost…
- Get more conversions for the same cost…
- Bob looks a lot like your Uncle.
All things being equal, that makes perfect sense.
Trouble is – all things aren’t equal.
After all, there is a reason why some auctions are more expensive than others
… It’s because higher-cost clicks are often higher-value clicks.
Low-quality inventory, such as Search Partners traffic and much of the Display Network, is often the cheapest.
Even within a particular keyword, the lowest-cost clicks don’t tend to be the most desirable (a serious issue for the Maximise Clicks Strategy, by the way…) because some of the eligible search terms will be more valuable (hence more hotly-contested, hence more expensive) than others…
Then within a given search term, some devices / audiences / locations etc are more valuable than others… and so it goes on.
The quality / cost balance can be favourable at any price point, depending on where the balance lies… Sometimes it’s easier to find your way into the black with high CPCs – sometimes with low.
So you see the problem with a focus on keeping CPC low. It stems from a desire for cost-effectiveness, but limits the scope for it.
In this chart, if we stipulate that CPCs must stay under $2, we are no longer able to access much of the profitable zone
Impression Share gives a unique window onto potential. This is highly valuable when you want to know how much more you could do either by raising budgets, or by bidding more aggressively.
But what exactly is being evaluated with Impression Share?
The answer to that question is too easily overlooked. I.S. is only interested in your precise, current setup.
It does a great job of taking into account your keyword selection (including negatives)… Audience and location targeting / device exclusions…
And in doing so, all the results it shows must come with the big, unspoken caveat, “This is how you’re doing vs your potential, assuming you don’t make any changes to your targeting / add any keywords / remove any exclusions…”
Another way of putting it is that Impression Share is one of those fractional metrics that is apt to give misleading results stemming from changes to the denominator…
Impression Share = Impressions/Potential Impressions
Both ‘impressions’ and ‘potential impressions’ are apt to change, and both have an equal say in the output metric.
• Campaign A has an unnecessary exclusion on all mobile traffic, and achieves 40% of potential impressions.
• I re-enable mobile traffic for campaign A. Its impression count rises… but its potential impression count rises by a higher proportion.
Campaign A now achieves 30% Impression Share, because I’ve expanded its scope and successfully increased its impression count.
This makes perfectly good sense within the logic of the metric, but not within the use of Impression Share as a target…. Because it’s a moving target.
This is a hard one to swallow, because a high CTR has real benefits, not just from additional traffic but also from Ad Rank, and therefore cost.
So we generally want our ads to have a high CTR, for good reasons…
But the fact is that making an ad more widely appealing tends to improve its CTR, and in attracting those additional (less ‘filtered’ clicks) decreases CVR.
Hence you’ll often see lower CVR with higher CTR. Not always of course… but it’s a common pattern – e.g.
One way in which this dilemma crops up is with the decision over whether to include the price in an ad.
By showing the price, you (disproportionately) reduce clicks from people who a) don’t want to spend as much as you charge, and b) are looking for informational content rather than a commercial site.
So a lower proportion of people will remain willing to click on your ad (CTR down) but a higher proportion of those who are still willing, should be convertible (conversion rate up).
This may be a particular pitfall for those more familiar with Facebook Ads.
With FB Ads’ CPM model, a higher CTR is pretty much unequivocally good news: More opportunity to convert, with no direct impact on cost.
In Google Ads, more clicks means more cost…
And while more clicks still means more chances to convert, the kind of targeting and messaging that brings a higher CTR often also brings less considered, lower-intent clicks (hence it doesn’t necessarily give you a higher chance of converting).
Hang on a minute, didn’t I just use Conversion Rate as the obvious ‘good’ that can get lost in pursuit of CTR?
Well yes – and it usually is…
But there are times when a higher conversion rate not only stops being a worthy KPI but, conversely, starts pointing towards wasted spend.
These cases are important to spot, and to act on.
This example is from a commercial cleaning client…
Along with high-quality leads, the client was seeing quite a few enquiries from prospective cleaners.
We were able to do a certain amount of filtering with keyword selection, negatives, ad text… but some traffic from the ‘wrong’ kinds of user will always slip through. Too much, in this case.
By paying attention to the data over time, we learnt that a much higher proportion of the irrelevant enquiries are coming from mobile than desktop… so the higher conversion rate we were seeing on mobile was not the usual indicator of better performance.
(In practice, B2B campaigns often see lower-quality leads – with a higher proportion of employment seekers – among the 18-24 band, so this could be a good first place to look for ‘dud conversions’ in your campaigns.)
Ideally what you count as a conversion really is the desired outcome… but with lead generation campaigns that’s not always the case.
So be prepared to let go of conversions, conversion rates, and CPAs when you can pinpoint segments of traffic that tend towards driving those conversions that you don’t want.
And remember that we now have Conversion Value Rules for cases like the one above, which allow us to downweight (or upweight) the value of conversions coming from a particular device category, audience or location.