My aim for this content is to condense some of the key learnings I have accumulated over the past few years into principles based resources to help those looking to grow their understanding of online business growth.
Learning how to think more than what to think will enable you to understand ‘the why’ behind what’s being done, which helps with diagnosing and fixing issues if they occur, alongside maximising performance and taking advantage of opportunities.
It will also enable you to learn how to apply what you know more easily to different products and services across a variety of verticals, and finally keep up with the ever-changing platforms, as they continue to evolve.
Why business goals?
There has never been more alignment between paid media and business goals, which is great news for both PPC consultants and business leadership teams!
As Frederick Vallaeys, author of Digital Marketing in an AI World: Futureproofing Your PPC Agency and CEO of Optmyzr very aptly said:
“PPC is becoming easier and harder at the same time”
Following the introduction of machine learning into paid media platforms, most prominently in the form of Smart Bidding in Google Ads, campaign creation and management has been simplified, this means in many instances day-to-day manual changes are often unnecessary and can be more detrimental to performance than conducive.
This doesn’t mean we can just “set it and forget it” though.
Often campaigns and channels don’t perform as expected, which is one reason why our day-to-day value as advertisers has moved to have more of a focus on other areas, a key one of which being able to create effective strategy to achieve the desired outputs.
Strategy can be a broad term though that often gets chucked around to mean all kinds of different things.
Essentially what I’m talking about here is taking insights in the form of business goals and performance metrics from advertising channels, forming an approach and then creating specific actions to achieve the desired results.
Strategy is as much about identifying what not to do as it is what to do, so, research and planning is a must.
Research, planning, and understanding business goals
The research and planning part of campaign strategy is probably the most important in predicting success and ensuring that your campaign performance will meet expectations – whatever those expectations may be.
Historically, we would bid for impressions or clicks and would have to make correlations, for example, that each click was as valuable as the last on a particular keyword or search term. With the introduction of machine learning, we are able to tap into predictive modelling based on online behaviour and bid based on the likelihood of that particular customer at that particular time to convert for our desired business goal.
This is a huge difference as bidding at the click level also meant applying blanket adjustments which although could be beneficial some of the time, would inevitably lead to overpaying for lower intent prospective customers and underpaying for higher intent prospective customers.
This is great news as we are now able to be even more focussed on the return we bring.
For e-commerce this has always been easier as we can see the direct return we are generating but we shouldn’t stop there, we have to become more commercially minded and start to factor in lifetime value and profit, taking performance measurement past generic in-account metrics such as “Conversions”.
Lead generation should also be no different, we need to be mindful of how many of those leads turn into SQLs and then how many of those turn into customers, what the average deal size is for those customers, the customer lifetime value and much more. Our measurement shouldn’t end when someone becomes a lead if we truly want to do our best work.
To start with we will need to gather information about specifically what we are trying to achieve, if it’s possible, what further opportunity is there and more.
Of course, there are benchmarks online which can be referred to as an indicator but what good looks like is something that will vary for every business so I can’t and neither can anyone else, tell you what good really is based on benchmarks alone. You need to do some work to find out what your baseline performance looks like for each channel – and that’s what we’re going to do.
Just a handful of things to consider:
- Product/service margin
- Advertising budgets
- Levels of competition in the auction
- Post-click experience (website)
- Market share
- Other marketing activity
- Length of the sales cycle
- Lifetime value
- The volume of sales/leads required
- Product lifecycle
- Market trends
- Brand maturity
Already this is a complex set of variables to be thinking about but it’s important that we do. The conditions we expect our campaigns to perform under and the inputs that we feed in are hugely important and are often the biggest predictor of success or failure.
Volume or cost-efficiency?
Let’s simplify this, our starting point needs to be:
What is the more important factor, growth at volume or growth at target cost-efficiency?
Driving out and out volume almost always comes with a compromise on cost-efficiency (ROAS & CPA).
From a bidding strategy perspective, if you have a target, machine learning will look to duck out of auctions where it doesn’t think it is possible to meet your target ROAS or CPA. As mentioned earlier this is because people have different levels of intent to convert, even if they are searching the same query. What you need to pay for one, will be different from another, even if both are deemed likely to convert.
From a wider strategy perspective, we can also see a parallel, when it comes to actual campaign strategy itself. For example, retargeting campaigns usually have the lowest CPA and highest ROAS.
Because the individual has already interacted with the product or service in whatever format you have created the audience from so has a higher propensity to convert.
If we created a display campaign using broad demographic audiences, we would expect to see a much higher CPA and lower ROAS because these users are likely to be earlier on their journey to conversion, if on the journey yet at all.
How the content is served in the campaign is also a factor, someone actively looking for a product or service and then seeing a paid search ad (providing that the ad is relevant) is having a fairly frictionless experience.
Someone who has a display (or social) ad interjected in front of, or alongside the content they are engaging with can be slightly more disruptive, even if the content is relevant and they are in-market for the product or service but not actively looking at that time.
“Let’s just do retargeting and sack off the top of the funnel then!”
With each trade-off on efficiency comes an associated trade-off in volume.
For example, a retargeting audience is finite, there will only be a certain number of add-to-cart hits, website page views etc over a given time frame so comes with a cap in volume. Our broad demographic audience doesn’t have that problem.
We need to do away with connotations of what’s good or bad. It’s not helpful. Linear ways of thinking like that will restrict your strategies and approach, limiting performance.
They are simply different tools for different jobs.
So, let’s get back to the point, are we looking for growth at volume or cost-efficiency?
There isn’t usually a black and white answer to this, usually, there is a bit of compromise required each way and this is something that will change as the business grows – this is normal. Try to get a starting point at least and then look to optimise over time.
But surely if we reduce CPA we can get more conversions?
Yes, if the cause of the higher CPA is something that you have control over to change and that it is what your goal is.
If a particular auction cost is high because of competition or it’s an expensive channel generally, even with the best set up in the world and an efficient post-click experience, you may not be able to reduce your CPA or improve your ROAS.
All of these types of metrics also lose context when focussed on and talked about in isolation.
Having a CPA or ROAS target with no associated volume of leads or sales isn’t very useful, similarly, neither is having a volume target, if you are hitting it but not making any money.
Having an awareness of the need to balance metrics like leads and CPA, sales and ROAS will allow you to ensure they are optimised for or measured with context.
Spreadsheets at the ready
If your business or client already has their tCPA, tROAS, and volume requirements then that’s great news as you’re a step ahead. If not, this process should give you an idea of what’s possible, which can be the starting point to the conversation around what would meet expectations.
We need to do some calculations. If you have historic account data this is a lot easier.
If you’re looking at paid search as I expect many of you will be, get your list of seed keywords or use the website to pull the monthly search volumes from the keyword planner in your Google Ads account. Make sure you remove any rows for keywords you feel aren’t relevant.
For paid social the same applies, forecasting is usually a bit tougher but if you start to build out your audience in a platform such as Linkedin, it will give you a monthly estimate of traffic which you can use as your starting point in the same way.
If you haven’t run any activity yet, you can use a benchmark but be wary this is most likely to not be a particularly accurate forecast in that scenario. Take it as a “best guess” to get an idea and then look to update once you have some live performance data.
Once the core data is in, I always rework the cells into formulas using the percentages, this will allow you to dynamically adjust certain areas. Feel free to use my example sheet here to get you started, it’s not extensive but shows the methodology below in use.
Once the core data is in, I always rework the cells into formulas using the percentages, this will allow you to dynamically adjust certain areas.
If you use your CTR and impressions cells to calculate your clicks and then your clicks and conversion rate cells to calculate your expected sales or leads, for example, you can start to tweak the numbers to see if you impacted performance metric x, what would happen to value output y.
An example could be that to return the amount of revenue you need, you increase the conversion rate percentage and see that it would need to increase by 5% to achieve your goal.
You can then build out a direct action:
“In order to achieve that revenue goal, we would need to increase conversion rate on average by 5%, to do this we would need to invest in a landing page tool such as Unbounce to increase lead form submissions via this campaign.”
A different example could be that you see you’re off on meeting your volume requirement but the potential is there (based on something like monthly search query volume) you just have a lower engagement rate currently:
“In order to achieve our volume target, we need to increase CTR by 2%. In order to do this we will need better creative assets for this Facebook ads campaign. This means we need to brief in the creative team for 4 new images at a cost of x, the benefit of this to performance is likely to be y, if we achieve our KPI.”
The important thing here is to work through all of these metrics until you get to the point of value.
So for lead gen you will likely have something along the lines of:
Impressions > Clicks > Leads > MQLs > SQLs > Opportunities > Customers > Average Deal Size
For e-commerce it will likely look something more like this:
Impressions > Clicks > Sales (AOV) > ROAS > Profit
Once we have this information we can start to look to answer all sorts of other interesting questions.
Is the total conversion volume possible to achieve based on the demand there is?
If you are able to achieve this volume, what would it cost? Is this budget in line with the budget that is able to be spent?
Not enough budget to achieve your goals?
More scope than you initially thought?
Only forecast to get a few conversions?
All of this information should really help to see how close you are to the desired goals. You may even see that it’s possible to surpass them. If that’s the case then great, you can go in with more investment or at least know the scope is there to expand.
Using the same approach we can start to model out what our cost-efficiency might look like. If the business has a tCPA of £3 and the avg CPC in auctions for those keywords is £5 then you have a problem. Perhaps it isn’t possible to see a direct return in this auction and channel with those parameters.
Conversely, if you’re predicting an avg CPA of £10 but the target is £30 you can see there is some room here to perhaps expand out to broader keywords or other campaign formats/channels to max out on volume, whilst maintaining your target.
For ROAS this works in a similar way but you can go a step further, using average order value you can model out, if we achieved a ROAS of 5:1, for example, with an average order value of x and a margin of y, we would return z.
If it looks like activity isn’t likely to be profitable you’re better off knowing and having that conversation than running the campaign and wondering why it didn’t meet expectations.
If this is well within profitability, then we can afford to try other keywords, other audience types, and other channels to allow the maximum volume of sales to come through on a monthly basis.
Using this type of approach means that you can see how likely your activity is to perform in-line with business goals before you start creating campaigns. It’s not perfect, and it will likely change very quickly based on the particular market and auction, so, it’s by no means a guarantee but what it is, is a great way to set expectations and validate ideas before testing.
It’s worth mentioning that performance metrics will vary for different campaign types and channels. The average conversion rate you see for search will likely be much higher than what you see for display or a Facebook awareness campaign. To make this effective, it’s important to have this level of segmentation otherwise certain campaigns or channels may end up under or over forecasting your performance, when aggregated together.
Moving back to bidding strategies, once you have the parameters you need, you can plug them straight in and once the learning phase is complete, hopefully you will be presented with results that meet expectations.
Finally, armed with this data, you can come back and forecast to your marketing director or client, we think for your investment of x, we can return y. If you’re delivering great work, you want to be able to shout about it to show your value and the value of the activity.
Optimisations and scaling
This methodology also plays a part for advertisers in terms of looking at optimisations and what to do next. The way that you optimise for volume will often be very different from cost-efficiency.
For example, if a particular keyword is converting but not as efficiently as another, if we’re within our cost-efficiency goal at the campaign level and we take it out:
Are we ok losing that extra 5 sales or leads per month?
Is that what the client or business would want?
If we’re currently getting a 3:1 ROAS and our target is 6:1, we can perhaps start to gradually change our targets within the bidding strategy, understanding that we will trade off some degree of volume and also have to allow for the learning phase to restart and complete before analysing performance post-change.
Asking yourself these types of questions and also asking the business is an on-going process, taking this mindset and approach will help you to be as aligned as possible in everything you do.
When it comes to scaling the same thing applies, let’s say we’ve got our search campaign performing in-line with business goals as we’ve discussed. What’s next?
Again we can come back to the question of volume or cost-efficiency to take some more calculated next steps.
Model out as best as you can your ideas and then discuss with your business or client to help show them what’s possible and set expectations before heading off to test. Forecasting will give you an indicator of risk and potential return, of course, once activity goes live you can update with the actual performance metrics and re-evaluate.
This methodology is particularly useful for testing and trying to get buy in for new channels. If you’re moving to uncharted territory, such as testing YouTube for the first time, it may be that certain types of targeting such as affinity audiences don’t work out to align with business goals in the forecast but something like custom intent audiences may do. Making this assessment is a calculated way to test and experiment for growth by validating ideas with data. You’ll be able to better test the validity of the channel as you have a targeting method that is in-line with the desired outputs.
Often too many variables are introduced when it comes to testing and experimentation, this makes it very hard to assess what did and didn’t work and can easily lead to something like a new channel being dismissed when there is actual viable opportunity there. In the example I have given, if you are new to YouTube advertising, you may be in new territory with targeting methods as well as advertising on the channel generally. Thinking about what targeting method is in-line with business goals allows you to at least reduce one variable which means you have a better on if there is validity in the channel itself.
If you are using Google Ads, there is a great new tool that does a lot of this hard work for you. It’s called ‘Performance Planner’, you will find it in tools and settings and it will allow you to model out for existing search or shopping campaigns (with enough conversion data) what your forecast looks like for various timescales.
Performance Planner “Simulates relevant ad auctions over the last 7-10 days, including variables like seasonality, competitor activity, and landing page” so it is able to make much more accurate predictions and you can use the slider to see what a trade-off in ROAS would do for conversion volume, alongside, of course, increases and decreases in ad spend.
We can see in the example below, if we reduce tROAS from 13.12 to 10.13 and add an additional £1.6k of spend, we can potentially return an additional approx £10k of revenue.
It’s great when it works, it’s not always available though, usually due to a lack of conversion data to be able to make accurate predictions.
I hope this article has been of use to you, what you need to do now is go away and review your own activity.
I’ve added some example questions below for you to get the cogs whirring, the crucial part to this though is to take the information and insights you find and work them into real-world actions. Start with your insights, move them into ideas and then create actions.
Are you using the right bidding strategy?
Do your current campaigns and advertising channels align with your business goals?
How can you forecast and align strategy with business goals for the next month, quarter, year?
Can you trade off some cost-efficiency for more volume?
Should you restrict some volume to improve cost-efficiency?