The year is divisible by four. The -mber months are here. It can only mean one thing - a US election is on the horizon.
The quadrennial battle pits two of America's largest digital marketing teams against one another. To the losers, a hefty Facebook invoice to wipe away the tears of defeat with. To the winners, a chance for that guy who appeared in all their ads to run the country for four years.
If the past decade is anything to go by, the digital war that's waged over the election is going to be fiercer than ever in the run-up to November.
But that's pretty much all we can take from elections past. Social media advertising changes at such a fast pace that four years may as well be an eternity.
What worked then won't work now. Campaigns teams will have to come up with new digital tactics to win over America's ballots.
So, with the election imminent, what sort of tactics can we expect to see on the ad front? What shady strategies might come to light after the election? These are a couple of possibilities.
One of the takeaways from the Trump campaigns' 2016 strategy is that there are two things you can try to change with advertising:
When we think of the purposes of political advertising, most of us instinctively think of number 1.
The problem with this though is that it's incredibly difficult to change someone's political opinions with a Facebook Ad. A much easier challenge is to change whether or not someone will vote at all.
Trump's 2016 campaign targeted demographics that were overwhelmingly likely to vote Democrat, and tried to convince them that Hillary wasn't worth voting for. Not that Trump was a better alternative; just that Hillary wasn't worth a trip to the polling booth for.
In November 2020, it could be far easier to convince people that it's not worth a trip to the polling booth. With COVID case counts likely to still be high around election time, I'd be surprised if political campaigns didn't use this to their advantage.
By targeting demographics that are most likely to turn out for the opposition party with ads that instil fear of COVID, campaigns could aim to turn opposition voters away from the polling booths.
I think the chance of this happening is so high that I've already built a script which requests and ad related to COVID from Facebook's political ad library; the tool they operate to allow researchers insight into political ads.
Even in early August, at time of writing, there's been a huge growth in the number of political ads published per day related to COVID:
Spend per day has also skyrocketed:
Having looked through some of the highest-spending ads, I don't believe any that are currently running are aimed at voter suppression.
The threat is most certainly there come November though, and it's hard to believe that no campaigns will resort to this tactic.
For all the furore over Russian interference in the 2016 election, you'd think Russian influence on Facebook Ads would have been routed out by now.
To some extent it has been. You have to verify your permanent address in order to run political ads now, providing a momentary stumbling block to any wannabe Russian meddlers.
There is one aspect of Facebook Ads though in which Russian firms still have an unparalleled advantage; audience targeting.
To see why, let's say you want to target Facebook members with a specific ethnicity. Facebook has just removed the last of its "multicultural affinity audiences" from its platform, meaning you won't have any vanilla targeting options to help you.
But, my election-meddling friend, there is still hope. There are tools out there, like LeadEnforce, that allow advertisers to target members of specific Facebook groups. When you realise that many people join groups which over-index on traits like ethnicity (examples here, here, and here), it's easy to see that there's a problem.
By targeting these groups, advertisers can effectively still target (or choose not to target) certain protected characteristics.
These tools work by scraping public lists of group members, and running them through 3rd party databases like the Russian CatchID, in order to 'enrich' them.
Enrichment is the process of adding additional information, such as addresses and emails, to the data that's already been scraped. LeadEnforce is then able to send the enriched data into ad platforms like Facebook, ready for advertisers to target.
Maybe this isn't a huge deal, you think. After all, not everyone of a certain race is going to join a Facebook group that explicitly calls out their race?
Sure, if a malicious advertiser wants to target a certain ethnicity, they might only get access to 1 or 2% of that ethnicity's population in the United States. But that's still a huge problem.
The reason for this is that, once you have a source audience of people that you know (or at least highly suspect) have a certain characteristic, you can teach ad platforms such as Facebook to find more people who are likely to have that trait.
This is a concept known as a Lookalike Audience. It's more typically used in traditional marketing settings, where an advertiser might create a lookalike audience of their current customers, in order to find more people that are likely to become customers.
But there are no checks and balances on the source audiences that you use to create lookalike audiences. This means it'd be all too easy for a malicious advertiser to effectively teach Facebook how to find people with a certain protected characteristic, in order to target them with ads.
One of the key headlines from the ads war of 2016 was to do with the sheer number of ads that the Trump campaign ran; some 5.9 million variations in the end.
While 5.9 million ad variations is no small feat, and likely required a fair amount of automation to produce, the production would have been largely driven by humans.
One of the broader shifts within digital advertising over the last 4 years is a shift away from painstakingly designed, manually crafted creatives, and towards machine generated content.
There's already a raft of platforms which are working on machine-designed ads and landing pages, and it's hard to believe that these technologies won't play a role in 2020 and beyond.
Bear this in mind when you next see an election ad. The ad's main asset, be it a static image or video, may not have been designed by a machine just yet, but there's an increasingly high chance that it's copy will have been generated by one.