ArticleDay Trading In Paid Media
Whether it’s in-person at conferences or events, one-on-one in calls, or in forums/groups/Twitter, one thing is abundantly clear: the frustrations with Google and Meta are real. They are shared by everyone from business executives to PPC specialists, CMOs to marketing directors, agency and brand side alike. And those frustrations are growing:
Just about everyone loathes – not just dislikes – Google Analytics 4.
Many brands don’t trust Smart Bidding.
Most advertisers (in-house, consultants, agency) are leery of “smart” anything – whether it’s AI-driven creative, or Performance Max, or ASC campaigns on Meta.
There’s mounting frustration from brands and agencies alike on the growing disconnect between inputs and outputs – between levers pulled in-platform and results observed from those changes.
All of this has been exacerbated by recent outages on Meta, sluggish ad performance on Google, and trying to decipher any of it within the riddle wrapped inside an enigma that is GA4.
My (somewhat unpopular, someone hot) take on this situation is: these are all features – not bugs – of the “next-generation” digital ecosystem. Today’s issue is going to focus on why this is happening, how we got here and most importantly, what we can do about it.
Things Used To Be Simple
If we rewind, back to the early days of digital media, the value proposition was clear and compelling: digital is more measurable, more transparent and more dynamic than traditional media. Digital advertising enabled us to attribute outcomes (sales, leads, whatever) to individual clicks, and those individual clicks to specific campaigns and ad groups, each with specific audiences, queries or placements. This, in turn, was what digital marketers took to their bosses, executive teams, and investors as justification for continued (and expanded) investment.
Back in these good old days, things were linear and controlled: a specific campaign targeted a specific keyword that served a specific ad that linked to a specific page on a website, all of which ultimately resulted in a specific action being taken. And, as those conversions rolled in, marketers could readily scale campaigns by increasing budgets – and see immediate impacts in the form of more/higher placement, more clicks, more conversions, more sales, etc.
But over time, ad platforms realized the above situation was untenable: users aren’t always predictable. Brands (usually) don’t know who to target. ~20%(ish) of searches are net-new. Not every person searching a given term wants/needs to see the same ad. Each one of these suboptimal realities created arbitrage opportunities in the form of radically under-priced audiences, keywords, placements, whatever.
And it’s these arbitrage opportunities that led to the rise of what I’ve come to call “day-trading PPC” – where marketers would (for all intents and purposes) day trade in their (or their client’s) accounts. It may sound odd, but the reality is that day-trading in PPC and day-trading in the stock market are/were fundamentally similar: in both cases, the trader is trying to out-smart or out-trade a system in a quest for alpha. The only difference is that, instead of buying stocks or options, PPCers were setting bids, turning on/off keywords, manually hunting around for newly-released audiences, seeking out under-bought geos, testing ads during times-of-day when the competition was absent, despite their target audience still being in the mood to buy.
Here’s the rub: arbitrage is (essentially) value leakage. It is inefficiency. It is uncaptured ad platform revenue. For advertisers + PPCers, it was glorious – what could be better than buying an impression with an expected value of $0.12 for $0.02, especially if there are 20,000 of them for sale on the average day? That’s $2,400 of value for $400 – it’d still be a steal at twice the price. Brands lapped it up, consultants/experts made millions….it was a golden era for marketing.
But from a platform’s perspective, it was an unmitigated disaster. Google + Meta wanted the delta between ad cost paid + net value created to approach $0.00, as this is what maximizes their revenue. So, over the past 5+ years, these platforms did exactly what one would expect, given that singular objective: they systematically removed or dismantled the systems that enabled it to occur in the first place:
- Match types have been (effectively) destroyed
- Audience options have been consolidated + removed
- “Fixed” assets (like ETAs) have been removed in favor of responsive assets
- Audience expansion is now a default // opt-out (vs. opt-in)
- Search Terms have been removed / adjusted
- Placements have been limited + aggregated
- Attribution has shifted from last click to data-driven
- Bidding has moved from manual to automated to smart
- Optimization has shifted from the input level to the outcome/data level
The list goes on and on. The common thread that runs through all of these changes is ad platform’s desire to create a more efficient market – and in so doing, maximize their value capture.
Broader match types + audiences mean that more advertisers are entered into every auction – thereby raising bids. Search Term + Placement removals hinder advertiser’s ability to opt-out of various auctions. Data-driven attribution allows platforms to bid on higher-funnel terms + still receive credit later.
The End of Arbitrage-Only Ad Accounts
While ad platforms have invested billions in creating more efficient markets, this doesn’t mean they are perfect at it (far from it); but it does mean that many of the traditional “day trading” approaches are no longer viable.
The days of spreadsheet + script PPC (where you’d calculate bids for every keyword in excel, then use a macro + script to dynamically update Google) are long – and thankfully – over. While it was (at times) fun, the reality is that most PPCers were dreadful at it – just as most day-traders would be much, much better off just buying VTI or VTAX (this isn’t investment advice).
Advancements in machine learning + automation have given us bidding algorithms that can calculate expected value based on millions of data points in fractions of a second. There is no human on the planet who can out-trade a modern algorithm – and so, the frequency and value of arbitrage opportunities in paid media is declining, even as investment in platforms skyrockets.
Evolution + Revolution Inside Legacy Structures
The other challenge – one which I spoke about several times over the past month – is that all of the above evolution is happening inside of legacy structures + terminology.
Most marketing leaders today (CMOs, VPs, Marketing Directors, Brand Owners, etc.) spent their early careers working in the “Things Used To Be Simple” ecosystem from above – and so, when they hear (for instance) marketers today talking about “keywords,” they expect that because the name is the same, the underlying functionality is similar. The same is true of “targeting” and “optimization” and “placement” and so on and so forth.
Candidly, this creates a massive challenge for marketers today: not only do we have to navigate changes in an increasingly opaque ecosystem, we have to do so while trying to explain (delicately) to clients, leaders and shareholders that the very concepts on which this ecosystem is based have all fundamentally changed. This is why it’s often easier to explain net-new functionalities/campaign types (i.e. Performance Max, Demand Gen, ASC) than it is to explain revisions to legacy campaign types (Search, Shopping, etc.).
And then, there’s the cherry on top of all of this: attribution.
Honore de Balzac once wrote, “Every great fortune is founded on a great crime.” In the case of digital advertising, that great crime was last-click attribution. The lie that was sold a million times over was that the “simple” thing above was ever that simple or that straightforward. Ad Platforms (and marketers, and everyone else) used attribution to convince everyone – from marketers to executives to investors – of the value of digital advertising.
Attribution was the catalyst that opened the financial floodgates while setting the expectation in the minds of a generation of leaders that marketing investment can be reduced to “$1 in gets $2 out.” – as if things were ever that simple.
The reality is that things were never that simple. Attribution was always a fugaze. Something as complex and multi-layered as human behavior can’t be reduced to a single interaction. Ad platforms knew this all along…which is why they’ve shifted to data-driven attribution, or 7-day click + 1 day view, or some variant thereof. Doing so maintains continuinity of the terminology (it’s still ROAS!), while (once again) fundamentally shifting the meaning of the underlying concept.
While I’ve never been a fan of attribution (I think it’s a fundamentally unsolvable problem, but one that’s nevertheless worth trying to solve anyway), I think – for once – platforms got it right with this change. It forces digital marketers to grapple with the fact that, under all those fancy terms and ad platforms, they’re still doing marketing. The same rules still apply.
And that’s where I think the next generation of Ad Accounts are going. They are leaning into marketing, not into arbitrage.
So, Now What?
Despite everything I’ve written above – I am more bullish on the future of digital media today than I ever have been (and, between us, that’s saying something).
Arbitrage – the kind I’ve discussed until now – is one way to create value as an advertiser / brand, but it isn’t the only way. The changes to ad platforms are forcing us to confront the hard truth that exploiting arbitrage opportunities is not equivalent to durable value creation.
The most successful ad accounts + brands today do both:
- Data-Informed Arbitrage. Successful paid media accounts uncover relevant arbitrage opportunities by integrating their data + understanding with ad platforms’ bidding abilities. With traditional targeting levers no longer functioning as they once did, optimization has shifted to the data level – the brand with the best data wins. The more (and better) 0P/1P data you can share with platforms, the higher the probability that the machine will be able to find value that other advertisers have missed.
- Set Actual Targets. One of the shortcomings of “old school” bidding was that it relied on a human’s ability to derive value based on a faulty attribution mechanism. Contrast that with smart bidding today: instead of having to calculate a bid based on value, costs + uncertainty, I can now spend all of my efforts setting actual, real targets based on business value (tCPA, tROAS) – then allowing the machine do the actual bidding on an user level – something no PPCer could ever hope to do with a modicum of accuracy. The best accounts pass back actual user / margin data to ad platforms, so we’re getting closer and closer to people & profit – something that allows me (as a marketer) to spend more time on strategy, offers, angles, creative, audience understanding and value creation, and less time setting bids.
- Build Durable Advantages. The best brands out there build durable advantage by obsessing about the entire journey, not just a click or keyword or a creative – from detailed audience research to remarkable creative, conversion-worthy websites/landers and remarkable experiences before and after the conversion.
- Do More Marketing. One of the remarkable things that emerges when you examine purchase behavior, drivers & triggers is that the vast majority of it happens outside the ad account – approximately 94% of the reason for a user clicking on one of your ads has nothing to do with the targeting, ad content or landing page (source: The Secret Life of Search). Most of it is stuff outside the ad account – the brand, the offer, the experience, the existing relationship with the brand, etc. The great unlock (at least for me) with smart bidding is the ability to spend more time + more mental energy moving the needle on that 94%, and less time on the 6% (which the machine can do better than I can, anyway).
- Evaluate Holistically. I’m absolutely delighted that more brands are thinking bigger + broader about their media evaluation – moving from conversations focused on ROAS, attributed revenue and other vanity metrics to conversations based on MER (or aMER / rMER), Contribution Dollars + (ultimately) incrementality. The future of marketing measurement looks a lot like the marketing measurement of old: Marketing Mix Models – only the ones we have now are far more robust, they’re automated, and they work hand-in-hand with digital platforms.
Let’s all be honest: we were never good day traders. But we can all be great marketers if we stop trying to be passable day traders.
Until next time,
Sam
PS. About GA4:
Before wrapping, I want to share one other point I’ve been thinking about for some time: if Google wanted to make GA4 look + function comparably to UA, they could have done so. They certainly have the expertise and the tech. But they didn’t. They made GA4 a tool geared toward data analysts, not brand owners or marketers.
My sneaking suspicion is that they made that choice with the intention of getting those people out of the data evaluation business, so they could do more of the things I’ve mentioned above. I have no way of knowing whether or not that’s true, but it certainly does fit.
