Automated Bidding and Covid-19 Lockdown: An Expensive Lesson
Darko•September 22, 2020
An Expensive Lesson in Automated Bidding We’re About to Learn
As marketing campaigns that screeched to a halt during COVID-19 / Coronavirus lockdowns begin to get going again, we will witness many businesses inefficiently spending millions (billions?!) by doing what many of us would think is natural: simply flicking a switch.
Why? Automated bidding.
Don’t get me wrong, automated bidding is a fantastic tool. Allowing machine learning to micromanage campaigns within constraints that we set, is great – it gives marketers time to focus on the strategic level. We have more time to look at the business decisions that give us new hypotheses, new constraints and new avenues to explore. We get to learn and know more about the bigger picture.
So why is flicking the switch bad? In short: because we shouldn’t be spending money teaching algorithms what we already know about the pandemic.
(Approximately) How the micro-decisions work
Freeing up our time to focus on the strategic and learn more here naturally means that you’re spending less time in the nitty-gritty. If you asked a good PPC manager what the effect of a £2 to £2.20 bid was on one keyword in a massive account, or how a very specific Display network placement performed yesterday, they wouldn’t immediately know. Nor should they! They should be focusing on the parameters and boundaries set to allow automation to flourish, not be swimming in the details.
Automated bidding adapts to changes in circumstances by:
Detecting that something isn’t working as it used to, or isn’t working as expected. Depending on budgets and speed of delivery, this detection could take place very quickly or take weeks, depending on your industry and keyword search volume
A hypotheses is then formed based on how the search engine thinks user behaviour has recently changed. This could be, “If I focus on slightly different Display placements at a lower CPM, I will get a higher ROI”, or perhaps “I can more aggressively target fewer conversions by bidding higher to secure position 1 more often” – or many other possibilities based on data
These tests endlessly repeat until something that works to your constraints (ROI / conversion / CPA / target share of voice) is achieved. Even then, automation is constantly experimenting with a portion of your budget to see if there’s something even better out there that hasn’t been found yet
What the COVID-19 / Coronavirus pandemic and subsequent lockdown has done to many industries is short-circuit this process, because consumer behaviour has changed so rapidly (during a period where many marketing campaigns were shut down, remember – no data being gathered) that a majority of hypotheses (step #2 above) that could be formed by automated bidding are doomed to fail from the outset.
Why automated bidding can get very (temporarily) expensive
The Travel industry is an easy example to pick here because of how severely it has been impacted, but there are many more sectors where the vast majority of consumer behaviour changed overnight. Things have changed drastically with Search for millions of people simultaneously, such as:
Geolocation of searches (office vs home)
Consumer intent (I want to look up a holiday refund or to defer my holiday, not to book a new one!)
Consumer sentiment (I still want to buy a car, but am now researching something more economical)
Ad networks and search engines use this data to work out crucially how likely you are to become a conversion. A simple example might be a B2B vs B2C query. You are more likely to become a B2B lead if you are conducting your search from a known office location. If everyone disappears overnight and we are now all working from home, that piece of historical data is (at least temporarily) worth a lot less to the search engine. It becomes a bit harder to work out what B2B or B2C intent might be, and new hypotheses for new behaviours need to be formed.
If you simply flick the switch and turn everything on as before, automation will eventually catch up. While it does, the data is gathered about our ‘new normal’ and many hypotheses of what’s profitable get rejected. Lots of testing happens at different bid points – for every ad, placement or keyword in every campaign. Naturally, your business is paying for this.
A fairly typical mid-sized business with a £75k/mo budget could (and many are) spend a budget up to and over this, so that automation can re-learn and ‘get to the right place for the new normal’ – which works, but can be horrendously wasteful.
Working closer with marketing teams and using business hypotheses can save a lot of money
If you are in a position where campaigns have been on ice for months now, adjusting them to the new world before hitting unpause can be a big benefit. The historical data that your campaigns would rely on to make bidding decisions, is for a world that looks very different. Why spend money re-learning this when we can make intelligent guesses ourselves to help it along?
Talk closely with your marketing and product teams, agencies and consultants wherever you can. Take the time to chat about how your business has changed. How do we believe consumer sentiment has shifted? Do we think that different products or services are more appealing to our audience than before? Spending a day or two discussing how business is now different, will lead to you being able to adapt campaigns before any money gets spent.
Mid-funnel ads for display or search keywords can be put in separate top-of-funnel campaigns if buying habits have shifted
Messaging can change to reflect more flexible policies for new customers
Landing pages can be moved to higher-funnel consideration phase for bigger purchases where intent has changed (such as vehicle hire/purchases)
Target CPAs / ROIs / conversion metrics can be adjusted based on results seen elsewhere in the business (such as offline sales, organic traffic, or other channel), minimising the chance that any degree of automation is trying to find efficiencies or targets that don’t exist
Having these discussions means that when something changes quickly elsewhere in the business, you’re in a better position to be informed and act quickly on this information – for many organisations this will be a very important habit to get into!
Continue to test hypotheses and don’t be afraid of being wrong – but put business hypotheses first when sculpting your campaigns for re-launch. This is precisely where paid media management that works closely as an extension of your in-house teams will benefit you far more than management that is further removed and is reliant only on search data.
Doing the above and more, before any additional money gets spent, will help you to avoid spending a lot of money for automated campaigns to learn what you either already know, or can deduce offline.
Our Related Insights
March 17, 2021
CRM Marketers are the Real Growth Hackers No One Talked About…Until Now