Impression share as a KPI is not just a flawed approach to your Paid Search/Search Engine Marketing (SEM) strategy, it's actually disorientating.
If you preach it, follow its commandments and aim towards achieving a perfect impression share, you'll end up hurting your own business performance.
The core of how it's calculated is the bane of its existence and despite how well it might sound as a KPI, it's very easy to manipulate, extremely inconsistent and steers you towards your inevitable demise.
If you're on the business/client end, the focus on impression share as a KPI moves you further away from the ROI of paid search, causing you unable to answer the question "how does paid search contribute to the bottom line".
Which is why I find it puzzling why "Impression Share" gets so much support and buzz to be used as a KPI.
I'm very glad that I'm seeing more coverage online on why Impression Share is a bad metric and this post is inspired by one of them.
As a paid search practitioner previously and the paid search lead on the client side now, I believe I should contribute to the good of the SEM community by voicing my thoughts on why you shouldn't, ever, look at impression share and if your client is addicted to it, take this material and shove it to their face.
This post is based on the content of the help article present at the publishing date of this post. We'll be talking about Google search engine in this post.
First, a crash course. Here's the formula for Impression share:
Impression share = impressions / total eligible impressions
Take a look at that and see what's inherently wrong with the metric.
While you think, let me go off-tangent.
For the purpose of this post, I'll use the equivalent metric of "Share of Voice" in the offline advertising business to illustrate the idea. I'll use the billboard advertising industry as that's more easily understood.
Anything that is a "share" metric has what you have obtained/achieved as its numerator. If it's the Share of Voice in billboard advertising, then this would be the number of billboards I have my advertisement on.
In the case of billboards, I believe the denominator should be the total number of available billboards in the city I'm targeting. That number should be a constant number throughout the period of my campaign. The number would not change drastically unless a disaster strikes or gentrification happened quickly.
Now coming back to Impression Share. The numerator looks good. It is what we have obtained/achieved.
The denominator says total ELIGIBLE impressions. What does that mean?
Let me tell you what it does not mean: The total available impressions for the keyword you're targeting in the geography you're targeting in Google Search.
It is not representative of the total market. Then what does it represent?
It represents the auctions you're eligible or qualified to participate based on the level of bid you set and your quality score, regardless of whether you lose or win.
Here's Google's explanation:
And here's a diagram that illustrates it:
Let's use an analogy to explain.
Let's say you're looking to buy a used car. Depending on the budget you set, you're only eligible for a set of cars in the used car showroom.
You're able to negotiate with the used car salesman, and depending on the quality and skill of your negotiation, you might get a steal, something that is a great value for money for your budget.
Or you could flop and get conned by the used car salesman to pay an extravagant amount.
The point here is, if you have $40,000 as your budget, you won't go for a Lamborghini or Ferrari, used or not used. You won't even waste your time cause regardless of how good your negotiation skill is, you can't bring that price down to your budget.
That is the auction/market that you're ineligible for.
If you're bidding $2 CPC for the term "car insurance", you will be in the low-value auction, where you can win some and your eligible impression is probably low.
Both numerator and denominator being low don't mean that your Impression Share is low, it could be high. That's Mathematics.
While you're bidding $5, there are other auctions invisible to you that are bidding for $25-100 for the term "car insurance" per click. Despite how bad they are in QS, you can never be better overall, even with QS of 10. You're ineligible for that auction.
But that's not all.
Google, with their AI/machine learning/neural networks optimising your keyword targeting, makes the whole situation more complex.
All keyword match types, from Exact to Broad, has grown fuzzier over the years as Google claims they're able to identify searches of similar intent using their AI system, hence making exact not very exact anymore.
In light of that, your "Total Eligible Impressions" change as Google's AI get smarter at identifying intent.
Slang and the way we used words do change with each generation and Google's AI can possibly capture that too.
Why would you use a goal where:
It should be very clear why you shouldn't use Impression Share, but if you're not convinced, let's move on.
Taking Crazy Egg's definition of a vanity metric, a vanity metric is:
I've covered the first point in the previous section, so I'll move to the second point.
I'll skip the third point as I don't have a publicly-sharable data on hand. However, I've seen many cases where an increase in impression share does not correlate to an increase in leads/sales nor an improvement in CPA.
A KPI or goal changes your behaviour, attitude and psychology towards your work and life.
It's like your university life, where you have to choose between working hard to be a Valedictorian, excelling in leadership positions or making a lot of long-lasting friends. Depending on your goal, your behaviour towards your university life changes.
When setting Impression Share as a KPI, you indirectly encourage an environment of conservatism, diminution and contraction.
Take note that in almost every business in the world, the main goals are growth and expansion.
Why does impression share promote conservatism, diminution and contraction?
Let's say you have 50 keywords and currently you have an impression share of 70%. Your KPI is at 90%.
What will you do to reach 90%?
Bid higher? If you're following this post very well, you will know that this is a gamble. Referring to the formula again:
Impression share = impressions / total eligible impressions
While your impression will definitely increase when you bid higher, your eligible impressions will increase too. Dependent on the rate of increase of both, your impression share could go either way.
Improve your quality score? While this is possible and less of a gamble, improving QS requires time and you don't always have the freedom of time when trying to hit a KPI.
A quick fix would be: Kill off the keywords with lowest impression shares
Not just any keyword. If you want the highest impact and fastest impact, kill those with high impressions but low impression share. Let's see how it simulates in this example:
See how easy it's to manipulate your impression share by pausing the right keyword? (Option 2)
Though pausing keywords with high impressions and low impression share feels like a sound strategy especially in terms of cost efficiency, it could be a bad idea to kill those keywords because:
Points 1 and 2 describe my point earlier where almost all businesses' main goals are growth and expansion.
If you eliminate those keywords with high impressions but low impression share, you will enter contraction mode - your campaign will grow smaller and smaller in volume.
You lose potentially valuable conversions and lose the ability to expand your campaign and conversion volume in the future as you've lost the valuable search query report of those highly-searched keywords.
Hence, if you implement this KPI for your agency or an internal team, they'll eventually come to a situation where they will be forced to take diminutive, contractive actions to meet your KPI. The reason is you can only increase your bid that much (due to budget limitation) and QS improvement has an upper limit (i.e. 10).
If your business focuses on growth and expansion, don't use Impression Share.
Finally, let's turn back again to Google's Help article.
As Google has stated, impression share is an estimate. It was mentioned 4 times in 3 paragraphs explaining how impression share is counted.
Would an estimate be a good KPI?
A KPI has to be measurable, accurate, specific and indicative of the bottom line i.e. the result of attaining the KPI impacts the bottom line positively.
An estimate is neither of the above.
The fact that it's an estimate means it:
Impression share as a KPI is kind of like going into a casino and estimating that you have a 40% win rate at roulette.
Your win rate can't be measured. There are too many factors involved in gambling.
It's inaccurate. The win rate of one day vs the other can swing greatly and have a wide range of values.
It's not specific. The range deviates wildly based on the dealer, the casino you went and maybe your luck (who knows).
The 40% estimate does not guarantee or indicate that you will win 40% of the time - it's an estimate.
The point is, you should never set an estimate as a KPI. It's a gamble.
I hope that you're convinced that Impression Share is a bad KPI and are looking for a new KPI.
What would be a great KPI then?
Following are my recommendations:
As an advocate of growth and expansion, the KPI I always recommend is growth in leads or sales quarter on quarter. Use net response rate and CPA as your efficiency and optimisation measures.
Use unbounced or qualified visits if you don't have a specific conversion point on your website e.g. a pure marketing website. You may also use other Google Analytics metrics like average time spent and pages/session.
Finally, use CTR if you have no other choice. Note though that CTR is easily manipulable using a clickbait ad copy.
Though I have shot down Impression Share quite hard, I believe there are some limited use cases for it.
It can be used to track a fixed set of priority keywords that you bid for strategic purposes or for the purpose of market presence.
The keyword is a fixed set. Keep it fixed to a specific set of exact terms to limit the fuzziness of Google's algorithm and limit the growth of the denominator.
For example, your brand has a brand name and 5 key product terms that you need to be present on as you're well aware of competitive and online retailers' presence.
Keep the 6 keywords in a fixed set where you monitor the impression share. Whenever the number drops, dive in to find out any competitive movements via the auction insights report.
In essence, it's a good monitoring metric.
It's also good as an optimisation metric via its two partnering metrics: Impression Share Lost (Rank) and Impression Share Lost (Budget).
These two metrics tell you your lost opportunity due to insufficient bids or QS and insufficient budget respectively.
Troubleshooting your campaign with these two metrics can be a lifesaver, but again, not a KPI.
I would like to close this long post with this quote above.
While it's important to have a keen focus on data and using it to drive decision making, it's extremely harmful if you look at the wrong data or have the wrong perception of the data.
Data builds your conviction and makes you more convincing. Holding on to the wrong data or having the wrong perception of the data will lead you to make the wrong decision.
Now that you know what Impression Share really is, think twice the next time someone wants to use it as a KPI.
Hopefully, this post and the many others online will help you in your case.