One of the biggest challenges that we face daily while working in paid media is being able to cut down the noise to identify what’s actually working and what isn’t.
We’re constantly swimming amongst a sea of data and metrics. It’s not just easy to get swept away from seeing opportunities, ending up down rabbit holes over optimising campaigns and reading into performance changes – it’s almost an inevitability!
This often leads to the inability to measure success. The consequence is that we lose the ability to make decisions, which can be a huge problem when it comes to business growth. After all, you need to act quickly and double-down on what’s working.
At the time of writing in 2021, we’re in a place where more businesses are being built from bedrooms than boardrooms, agility is order of the day and the white noise from the over-abundance of metrics and data around us is a roadblock to getting things done.
A paper by Forrester Consulting highlighted that:
“Despite decreased budgets due to the pandemic, 63% of brand leaders stand ready to invest in agencies that can produce measurable outcomes”
Marketing directors are under more pressure now than ever to focus on (and implement) activity that achieves a direct return due to economic uncertainty.
This amplifies the challenge for the paid media team. Bringing in opportunities to the table also means inviting risks, which means you need to be armed with a robust case for what you want to roll out, why you want it and how likely it is to deliver a favourable return.
As author of Skin In The Game, Nassim Nicholas Taleb says though:
“You should study Risk Taking, not Risk Management”
There will always be some element of risk in marketing. Taking opportunities is how you’re going to grow and a negative test can sometimes be as useful as a positive one. Having the right insights and forecasting should allow you to mitigate concerns, scale and thrive, even in economic uncertainty, providing you’re measuring what you do on the right metrics.
If you’re not taking the opportunities that are coming your way, you can count on the fact that someone else will be.
This is a complex topic and as such it requires long-form content to get into the detail required, so strap in for a deep-dive! I will outline further detail on this challenge and then offer my solution for the problem.
For context though before we go any further, this blog post is focused solely on performance campaigns to drive sales or leads. For brand building and other similar objectives this content would be different.
Mindset and machine learning
It’s not just information overload that can make measuring success and decision making a challenge. We’re living in the machine learning age of advertising – it’s not the future, it’s the present and it has been for around 2 years.
With automated bidding and dynamic creative (using a vast amount of data signals that we couldn’t process manually, but also don’t have access to), our analysis of what ‘makes sense’ based on legacy metrics is not always the best proxy for success measurement. These strategies are designed to align with business goals and focus on hard outputs not advertising metrics.
I talk about this in a bit more length in the cognitive bias section of my blog post on mindset. Our brains are essentially hardwired to apply bias and rules to condense the amount of information we’re taking in at any one time so that we can make decisions easier. From a digital marketing perspective, this creates a problem in terms of measuring success and reporting on what’s working and what’s not.
Our brains prefer ways of thinking that ‘make sense’ and very often now, those ways of thinking often aren’t aligned to advertising in the machine learning world.
This was the best we had to work with prior to machine learning and to some extent it is correct (in the sense that you need impressions to drive clicks and clicks to drive conversions). However, the problem here is that it then can very sneakily lead us to the conclusion that:
‘Clicks went up = success’
‘Clicks went down = failure’
This isn’t always the case. This can lead to us trying to make our creative overly engaging to obtain the highest possible number of clicks. So, you end up optimising for people who just click on your ads, instead of people who are most likely to click on your ads to buy.
Facebook for Business has done some interesting research to back this up showing that:
“Not only are people who frequently click on ads not necessarily more likely to buy, they’re also more expensive to reach.”
What I see happen very often, as an example, is when you take a campaign from manual bidding to Smart Bidding in Google Ads, you will likely see a drop in impressions and clicks but providing you have the correct set up, you see a significant increase in conversion rate so your business goals output goes up significantly for the same budget.
It’s exactly what is supposed to happen, you’re bidding based on a particular user’s propensity to convert or convert at target cost, so you’re ducking out of auctions where that’s less likely to happen, which can mean less impression and clicks.
Machine learning based bidding strategies aren’t optimising for ‘more in the top equals more in the bottom, they are optimising for the bottom and cutting out the middle-man. This can defy our pre-existing ideas about measuring success if we’ve been working in marketing for 3+ years where reporting tends to focus more heavily on advertising metrics.
Moving to example two – another common situation historically would be that as advertisers, we would work to reduce avg CPC. The idea here was if we reduce CPC, CPA will go down and if CPA goes down, we can get more conversions for the same budget, again makes sense right?
The problem here again is that we aren’t bidding for clicks anymore, we’re bidding for customers and customers at target cost.
Often CPC will increase when you move to a machine learning based strategy. CPA often doesn’t though as again, conversion rate goes up as you’re bidding for users who are likely to convert not for clicks.
tldr: Correlation does not equal causation.
Indicative metrics such as impressions, clicks, CTR and avg CPC are still useful for forecasting, optimisation and troubleshooting, of course.
However, they are now more indicative metrics than they are measurements when we’re talking about reporting on success or failure of performance paid media activity.
These situations will occur often and you should expect to find yourself having these conversations.
You can apply the Five Whys technique before you decide on if a campaign is really working or not (and definitely before planning changes and optimisations).
This technique is designed to get to the root cause of a problem and is incredibly helpful.
I’ve adapted the Wikipedia example linked above for the context we’re talking about.
Why? – We need to mix-up our ad copy. (First why)
There is so much value in this, you might not need all five questions, it is just about getting to the root cause and then making decisions based on the correct information. Even if you go through the process and realise you were right you’ve done your due diligence, now you can make decisions more assertively.
When you see something positive or negative you need to apply critical thinking to the situation. Ask yourself, “is this an indicator of performance or a measurement?” before you start to action change or report off of the back of it – further diagnosis may (and likely) will be required.
In the world of machine learning paid media where business goals are involved, our measurement metrics are those metrics that align to business goals and not anything else.
Our indicative metrics are anything else along the journey. They aren’t nothing, but they aren’t the measurement of success or failure when we’re looking to see a direct return in paid media.
The metrics you use specifically will depend on the setup that your business or client has because you may not be privy to certain information, or may not have the facility to know it via a CRM for lead gen, for example.
I’ve given the outline below as a guide. Try to get as close to the right hand side as you possibly can and focus your inputs and optimisations on those as measurements and not anything else.
As we will explore in the next section, this way of thinking will streamline what you need to focus on and will ensure that when you make changes to improve performance or to scale, you visit the right areas to do so.
Even metrics like conversions and conversion rate can paint the wrong picture when there are multiple actions being counted under that one umbrella term.
If there are 4 different actions falling under ‘conversions’, and 3 out of the 4 are things that are either not business goals or are not what you or your campaign is being measured on, then you have a disconnect.
Using data for data’s sake and importing metrics because they are just there is arbitrary.
Take this as a rule: Don’t be general when you could be specific.
Aligning metrics with business goals
So let’s dive deeper at this point and look at context. It’s not as simple as updating our reports to different metrics and moving on.
As Brian Clifton states in his book Successful Analytics:
“Analysts thrive on data; executives need insights.”
A huge part of the paid media advertisers job now is to take the data that measurement metrics provide and convert them into insights for the business that they are working on to grow. This is important to align strategy, implementation and optimisation with business goals.
This adds an immense amount of value and takes you from being a service provider in the context of creating campaigns, to actually ensuring that the business is making money through the activity that you are running.
I’ll give you a couple of examples of how this works.
This business is an ecommerce store. The product is something that doesn’t benefit for customer lifetime-value so direct return is what the client is looking to see.
Return on ad spend: 4:1
For all intents and purposes this looks pretty good if you were looking at benchmarks. After all, you’re returning more than you’re spending.
Regardless, the problem is that depending on the margin of the product(s), this may or may not actually be profitable for the business. ROAS & revenue are very good indicators but they are still indicators nevertheless.
If we add in average order value, margin and profit to get to ROI, we can start to get a clearer picture of a definitive measurement of success on this activity.
Average order value: £100
Margin Value: £40
In this example we’re seeing a positive return overall which is great, but we still need to use it as the starting point to a conversation around what to do next.
We now have our insight rather than data to be able to do this effectively:
‘This month we generated £10,000 of revenue with a 4:1 return, which based on an average order value of £100 and margin of 40% means we achieved an ROI of £1500’.
Is this profit enough for the activity to be viable for the business? Does it meet expectations?
Following this part of the conversation, from a strategy and optimisation perspective:
Do you want to look to increase ROAS to make the current spend go further? How would you do that without reducing sales volume?
Do you need to scale revenue? What’s the best way to do that whilst maintaining target return?
The outputs of this conversation can give you the steer you need in terms of where to spend your time and which future direction to take – no more guesswork.
ROAS needs to increase without reducing sales volume and changes post-click to increase conversion rate would offer the most effective way of achieving this. If we increase tROAS in the bidding strategy, we will likely duck out of auctions, leading to potentially a lower sales volume. Similarly, if we pause keywords with a lower ROAS that are converting, we will also likely reduce average monthly sales volume. CRO it is. Get the tasks in, make it happen, rinse and repeat.
Lead gen example
This business is a B2B service provider who is looking to grow their customer base through generating qualified leads. They have a CRM system and track lifecycle-stage through from MQL to customer.
Cost per lead: £50
Neither CPL nor lead volume in isolation can be a definitive measurement here. One without the other doesn’t mean a huge amount in the same way that achieving a 10:1 ROAS, but only making one sale per month likely doesn’t either.
Seeing how far these leads progress through the lifecycle-stage is very important to actually measure the success or failure of what you’re doing. Anyone can generate ‘leads’, but not everyone can generate qualified leads that lead to opportunities, sales, and a profitable return.
Cost per lead: £50
Lead to customer rate: 5%
Average deal size: £2,000
I’ve skipped some steps such as lead to MQL rate, MQL to SQL rate etc as the example is just to illustrate the point. We’re now measuring our activity based on business goals such as actual sales opportunities and deals rather than more indicative metrics such as ‘leads’.
Of course, you can and should take this example further also to profit, it’s slightly more challenging generally with lead gen as sales cycles can take longer periods of time to complete, so you may have to wait a while to see the actual deals going into the CRM. This is where this forecasting method can bridge the gap. The key takeaway here is that we’re measuring based on business goals.
The conversation and outputs again should dictate your next steps. The way you optimise for higher qualification will likely be different to optimising for volume. Having this steer away from the business means that your optimisations are more likely to meet expectations and everyone is singing from the same hymn sheet – trying to grow the business in the right way.
We need to increase lead volume, we have plenty of room to play with in CPL having reviewed lead to customer rate, revenue and profit. We increase our CPL target in the bidding strategy to allow machine learning to bid more aggressively and to enter more auctions, leading to a higher lead volume.
From blog posts to boardrooms
It’s likely starting to become clear that our role as paid media specialists no longer ends in the ads account. We’re closer to c-suite than ever before and what we do needs to align with the boardroom. This is not just so that we can measure activity on the right outputs, but also so that we can feed them back in and do our day-to-day jobs in the most effective way.
Advertisers need insights on business goals and businesses need opportunities (risks) to take as well as being told about current performance.
Taking what we learned from the example client situations in the previous section, you now can and should bring those learnings into your strategy moving forward.
The conditions that you expect automation to work under have a huge bearing on your outputs. If you know you need a certain ROAS in order to hit target margin to be profitable, that data needs to be input so that results align with expectations. You can structure your campaign changes around what’s going to help your bidding strategy achieve the goal it needs to and then let it do what it’s very effective at – achieving that business goal.
We’re months, not years, away from being able to bid for profit margin. Bidding for conversion value at target ROAS is so effective it’s more than likely on the horizon. For the time being you can achieve the same thing anyway but doing the manual calculation work we looked at in the last section so there’s no need to wait.
When it comes to lead gen there’s also a huge amount of opportunity to feed in more of the right signals to drive more of the right results.
If you have a compatible CRM system such as HubSpot, you can now push your first-party data back into your ads account (such as MQLs, SQLs, opportunities and deals) as offline conversions, so that you can utilise machine learning to bid for not just leads but leads who are most likely to become customers and this is massive.
The line between marketing and sales here is really starting to thin which is great news for everyone.
This is something that will also start to come into its own with restrictions on cookie-based tracking following data privacy changes. If you lose the ability to track the ‘lead’ on your website you can’t use it as a measurement anyway so best we get ahead and focus on end value.
The business goals focussed way of measuring success ensures that reporting is focused on the highest possible value metric. This leads to more confidence in being able to double-down on what’s working while taking new opportunities through faster and more effective decision making either yourself or collaboratively with your business or client.
To refer back to the report I cited in the introduction, this is incredibly important in the current climate where businesses are being more cautious with investment and there is more of a focus on direct return than ever before.
When it comes to optimisation it takes away a lot of the guesswork and requirement for ‘best-practice’ and templates. It gets you focused on what you need to do right here and now, through giving you ‘the why’ of the bigger picture at play and means that what you do is far more likely to meet expectations.
Get to the root cause of what you’re trying to do. Use the Five Whys, look at your business goal outputs and think:
“What’s the best way to achieve the required increase or decrease?”
Avoid the temptation to favour complexity to make a change hoping that it causes a change, that then leads to the change that you need.
It can be tough to let go of legacy metrics because it requires a mindset shift, but once you’re able to, performance starts to really come into its own.