Article

Your Daily Budgets Are Killing Your Campaigns

Economic Analysis google ads Meta

November 13, 2023

Over the last few weeks, I’ve reviewed a handful of ad accounts – and noticed a particularly problematic trend: daily budgets killing growth. As a result, this issue is dedicated to addressing that issue, along with how you can avoid falling victim to it. Let’s get into it.

One of my core premises as a media buyer has always been to approach digital ad platforms the same way an investor approaches markets. That starts with understanding three core principles: (1) as an advertiser, you are fundamentally a market taker; (2) that expected value is the primary constraint on ad markets & (3) market supply is volatile.

Market Makers vs. Market Takers

In economics, healthy markets are populated by two core roles: market makers and market takers. Makers create orders and wait for them to be filled (in digital ad parlance: makers are the ad platforms placing inventory for sale at the best available price and collecting a spread on each transaction). Market takers are advertisers who place orders for various ads (based on your targeting, performance targets, inventory restrictions, budget, etc.) and wait for them to be filled by the market makers (the ad platforms).

In order for an order to be filled, it must meet the following conditions: (1) the impression must be within the advertiser’s targeting, (2) the expected value of the impression must exceed the minimum-acceptable return for the advertiser AND (3) the advertiser must have sufficient funds available to complete the transaction.

There’s a lot to unpack there, but let’s start with the expected value.

Expected Value

From a high-level perspective, expected value is a way to determine the risk-adjusted value of an asset; in the digital advertising ecosystem, those assets are impressions (the fundamental units of inventory that are bought and sold in the markets described above – even if you buy on CPC, you’re still buying on CPM – you’re just trusting someone else’s math).

Now, as anyone in advertising knows well, not all impressions are equal. Some are objectively better than others. The relative value of an impression depends on hundreds (if not thousands) of variables, with the core ones being:

Who is served the impression?
Where is the person being served the impression located?
Where on the page is the impression served?
How many other advertisements are being served on the page?
What time is the impression being served?
What is the content of the ad being served?
What is the offer contained within the ad?
How well does the advertiser’s site perform?
How often do users similar to the one being served the ad accept the advertiser’s offer?
How well has the ad performed historically?

Ad platforms have reduced these (and many other) variables into four core considerations, which are used (in combination with the bid + budget of the advertiser) to determine which ads to serve each time an impression is available:

Bid & Budget
Ad Relevance
Expected CTR
Landing Page Experience & Expected CVR
Optional: Expected AOV / Revenue (used for tROAS bidding)

Using these variables, each ad platform (Meta, Google, LinkedIn, Microsoft, etc.) computes the “Total Expected Value” of a given impression, along with the “Minimum Acceptable Expected Value” (“MAEV”) for each advertiser, based on the constraints you’ve put in place (tROAS, tCPA, Cost Cap, Bid Cap, etc.). If the Total Expected Value of an impression is greater than the MAEV you’ve set, your ad will show; if it isn’t, it won’t. This is, fundamentally, how target-based bidding works.

It’s important to note that this is inherently a probabilistic calculation, made under significant uncertainty: platforms don’t know *exactly* how likely a given person is to convert, nor do they know *exactly* how much a given user is willing to spend for a given product. Sometimes platforms get it wrong. Sometimes they make mistakes. And that’s why you should take your free insurance in the form of bid limits (which defend against those blowup CPCs/CPMs). But, by and large, digital ad platforms are pretty good – especially over the long run – at getting it right. If you want proof, check out this graph from Mike Ryan over at Smarter Ecommerce:

This shows that, for a sampling of his client accounts, that as the number of optimization events (conversions) increases, the actual cost approaches or exceeds the target set. Or, in other words, a digital ad platform’s machine learning is pretty damn good at hitting or exceeding your performance target, when given sufficient data.

If you’ve been reading some of my previous issues on Enhanced/Offline Conversions + Data Passback – this is the point where these concepts connect: the incremental data provided to platforms via the processes is used to hone the values for each of these variables – which, in turn, allows you to bid more accurately than your competitors who are not using these tactics. It may seem like a minor thing, but the impact of having 10% or 20% better information than your competition is *staggering* when considered on the scale of an ad account. Data advantages, like financial advantages, compound over time.

Supply Volatility

To be blunt, the supply of people who want to buy what you’re selling (whether that’s a product, service, whatever) is not consistent every day. The number of people active on a given platform at a given time is certainly not consistent. The number of other advertisers who want to reach the same people that you do is not consistent. Each of these things impacts the price of digital ad impressions (and, correspondingly, the expected value of those impressions). As a market taker, you’re at the mercy of digital ad platforms – some days, there will be a LOT of

impressions available that satisfy both your targeting constraints and your expected value constraints; other days, there will be few (if any) impressions available that do the same.

The beauty of automated, target-based bidding is that it performs the calculations described above (computing the Expected Value of an eligible impression & comparing it to your MAEV) in a fraction of a second, using more data than any of us have access to.

This is cool & all, but how does this relate to daily budgets?

To put this simply: daily budgets act as an additional constraint, directly impacting an advertiser’s ability to participate in qualified auctions. In a vacuum, this isn’t a bad thing – after all, most advertisers have budgets allocated on a daily/weekly/monthly basis, and don’t want to exceed them. The problem is that fixed budgets (whether they are daily, weekly or monthly) assume that supply is consistent, which we know is not true.

To illustrate this, consider the below graph, which shows available impression volume (in red) over time, relative to the daily budget (in gold):

The area below the red line and above the gold line (shaded in green) represents impressions with an expected value equal to or greater than this advertiser’s MAEV that the advertiser could not obtain because of the daily budget. Conversely, the area below the gold line and above the red represents days where there was not a sufficient supply of qualified impressions to spend the advertiser’s budget.

To be quite blunt: the entire area above the gold & below the red is money left on the table because of an unnecessary constraint. For an advertiser similar to this one (I used anonymized data here), this could represent hundreds of thousands of dollars in unrealized contribution dollars (not revenue!) – and this is just one month. Over the course of a year, that could be $2M, $3M or even more in bottom-line dollars that went un-captured, all for the sake of a daily budget.

There’s a reason Google & Meta are actively weakening daily budget controls (while keeping monthly limits in place) – it’s because this scenario is becoming more and more common.

If you’ve set your targets properly, and if you’re giving ad platforms excellent data (enhanced + offline conversions, CAPI, etc.), then your daily budget should only be a “emergency stop” constraint (i.e., set so high that you’ll only hit in the event of a platform-wide meltdown).

What this means, in practicality, is that there will be days where your account spends $1,000, days where it spends $10,000 and the occasional day where it spends $50,000. And, with the right targets, all of those days are good days. They’re good days because you’re taking all of the available market on each of those days – not leaving qualified impressions for your competitors due to a daily budget.

I fully understand that this is a major shift in advertising and budgeting. I get that, from an advertiser/brand perspective, it is probably downright terrifying to give a digital ad platform carte blanche, with only a tCPA/tROAS/Cost Cap/Bid Cap. But, as any smart investor will tell you (and Warren Buffett has): great investors are conservative right up until the moment when they’re aggressive. They hunt for great opportunities, and when they find them, they push as much capital into those opportunities as possible, until the expected return of the next incremental dollar falls below their return threshold. Everything I’ve described above is doing exactly that – just with the aid of automated bidding.

I Don’t Have Performance Targets / Cost Caps – Does This Apply?

One of the dirty secrets of digital ad platforms is that bidding strategies without a performance target (so, Max Clicks, Max Conversions, Highest Volume) simply aim to spend your entire budget each day, at the best return possible for that day. But, going back to the above – not all days (and not all impressions, and not all clicks) are created equal.

On days with supply gluts (i.e. everyone is shopping + relatively few advertisers competing for impressions), a target-less strategy might perform on-par with a tCPA/tROAS strategy – and even might exceed the strategy with a target *if* you get lucky on some impressions with an Expected Value lower than your MAEV going your way. But, on days with high demand + limited supply (lots of advertisers, relatively few buyers), the platform will still spend the targetless-advertiser’s *entire* budget – whether or not it yields anything. For advertisers with a performance target, the platform will simply not spend on days like this. And that’s a good thing.

If you remember anything from reading this newsletter, it’s that spending poorly is worse than not spending. Why? Because when you spend poorly, you must generate a higher-than-expected return on all subsequent dollars spent just to get back to your original performance target.

There Are Other Good Reasons To Use Daily Budgets – Like Business Factors!

Of course, impression volume is only one constraint on a business – there’s inventory, fulfillment, capacity, cash flow, sales cycles, team availability, etc. that all must be taken into account. This is, perhaps, the most common objection I hear to the above – daily budget is just an (admittedly, imperfect) method of accounting for these other factors.

My response to this is simple: if that’s the case, then you’re throwing money away in a different way. Why? Remember that, all things being equal, each impression/click/conversion is more expensive than the last. It follows that, if you’re budget-capped, you could *lower* your targets (i.e. make them more restrictive), get the same conversion volume, and spend *less* to do it. So, in this case, you’re still throwing money away (probably less, but still – every dollar counts).

Do I have a chart for that? Of course:

In this case, simply making the target more restrictive, while removing the daily budget constraint, is projected to yield the same volume as a more aggressive target cost/conversion, with the advantage of substantial per-conversion savings.

Embrace using performance targets as your primary spend control levers. Not only will your ad accounts perform better over the long run, but so too will your (or your client’s) businesses.

Thanks for reading!

Cheers,

Sam

Related Insights