Asking the RIGHT questions in your next ad agency interview

How boring would Mad Men be if it was 2024 and D. Draper is running Facebook Ads?

The advertising ecosystem has undergone significant changes since we exited our first venture in 2021, and it has become even more dynamic since we started Igloo. One of the most common phrases we hear from founders whose earnings were heavily dependent on Facebook is: “My ads have never been the same since iOS 14.” In essence, the new software update gave users the ability to opt out of tracking, which was fundamental to Facebook’s audience-targeting algorithm. As a result, the entire ecosystem shifted, forcing businesses to find new ways to acquire customers.

The debate over whether Apple played its best cards is ongoing, but if I were Zuck, I’d probably have Tim Cook on speed dial—just in case he decides to tank my share price by another 60%.

Nonetheless, in the last 18 months, we’ve seen the rise of AI-driven algorithms such as Advantage+ (Meta) and Performance Max / Demand Gen (Google) and as a result, increased skepticism about the reliability of tracking and attribution. Why? Because over time, advertisers have been forced to give more power to the algorithms, knowing that manual targeting results have weakened (typically). Ideally this allows operators to spend more time focusing on creative development with the mindset that “the algorithm will do the work for me, if I feed it the right assets.” How an agency has strategically adapted to the changes in recent years, is also a great opening topic.

The devil’s in the details

As a founder, vetting an agency candidate could be one of the most important decisions you make. So we’ve curated a blueprint list of Facebook Ads questions to ask in your audits, and test their level of expertise. Let’s dive in.

7 Facebook Ads questions:

1. Should we focus on cost per purchase (CPA) or ROAS to assess our profitability and scalability?

Answer: Depends on the product category. If you know you have a lifetime value (LTV) of each customer that far exceeds 1 purchase (for example: fashion brands, subscription businesses, CPG products etc.), then you want to be able to manage an account knowing the profitability of acquiring users across a lifetime. If you’re managing budget based on first purchase profitability it may limit your growth. Of course, it’s important to acknowledge cash flow dependancies and whether or not you can afford the runway between payment cycles.

1. When should we use a maximize volume bid versus a cost per result bid? Which one is more suitable for our goals?

Answer: Well, it depends. Maximize volume is ideal when we're trying to rapidly grow our audience whereas cost per result bid strategy is ideal for campaigns where CPA/ ROI is crucial. Best would be to start with maximize volume & gradually move towards cost per result bid to have more control on our costs & CPAs.

1. None of our creatives seem to be effective. We’ve tried static images, videos, UGC, and more. What approach would you suggest instead?

Answer: Leverage a creative strategy platform (optional but helpful), to see where your audience is engaging best and worst within your hard and soft KPI’s (hook rates, engagement rates, play time, CTR etc), and how you can pipeline a creative strategy workflow that leverages your learnings. Having a consistent pipeline of Concept Strategy > Creative Development > Testing > Analysis > Repeat is vital. Log your learnings, insights and test and iterate like your life depends on it. Creative strategists can be the most important hire in this process.

1. What is the recommended budget for running user acquisition campaigns?

Answer: This is variable, and methods can change. Ideally, you should start with a test budget that allows us to gather sufficient conversions for insights. If your budget’s too low, it’ll take too long to reach your target CPA. If your budget’s too high, you may spend more than you need to identify creative winners. Ideally a daily budget that’s 3-4x your target CPA, if you’re a new account, is sufficient to start. Meta’s algorithm needs “results” for it to identify higher converting audience segments, whether ATC or purchase actions, you want to have enough budget to at least draw conclusions.

1. Advantage Plus seems like a black box. How can I guarantee Advantage Plus isn’t just targeting previous buyers?

Answer: Advantage plus has the ability for you to put existing customer caps in the settings at the ad set level. This is something that’s often overlooked, and a big red flag if agencies don’t do this by practice. This feature allows you to put % caps, so you can limit the amount of “existing customers” you target. Implementing 3 audience segments at the ad account level: Engaged Customers, New Customers, Existing Customers, is also very important. Controlling that distribution is vital and should be mentioned in the campaign set up process.

1. What if Facebook allocated almost all the budget to one ad in the ad set and the others just don’t get any spend?

Answer: If the goal is to test new creative winners, do so in a separate campaign, then move successful creatives to the core campaign, keeping in mind that Facebook's algorithm allocates spend based on early performance signals. Mitigating spend allocation to the top performing assets is a challenge - the agency should acknowledge this. There’s no perfect answer, but looping in a test and scale method, where you’re isolating tested creatives from scaling creatives (we call these “snowballers”) in separate campaigns is the best way to force spend to an asset and see the results.

1. How often do you like to meet with clients to review the account?

Answer: The correct answer is that every account is tailored to the client. Some founders are very knowledgeable about ads and prefer to stay involved by emailing and closely tracking the data as well as the management. Others just want operators to implement their strategy. So, unfortunately there’s no one-size-fits-all solution. However, agencies should always encourage weekly, if not bi-weekly catch-ups by default, where clients have the chance to ask questions, learn more about the methods and execution, and discuss the plan of action moving forward.

Agency Pricing - The Hybrid Model

An agency's pricing structure should be fair (relative to the business size and team) and, if there’s a percentage-of-ad-spend model, it should be incentivized and regulated. Many agencies that work with large companies embed this into contracts as a way to make extra money without necessarily earning it. Ideally, you want to have your agency on a structure where they earn a little extra when you help scale the business. Agency rates are normally negotiable, though they often seem rigid during introductory calls.

The best approach is to use a hybrid model. For example:

Brand A spends $50,000 per month on average and is struggling to acquire new users at higher spend thresholds within their target ROAS/CPA. The agency's baseline retainer could be $2,000 per month per channel. But if the agency operators scale the spend to a particular milestone that shows growth within your KPI goals, then—and only then—should they receive additional remuneration. In my opinion, it’s good to set this bar 20-25% above the average monthly spend. Also, anything more than 5% additional remuneration is unnecessary. If your agency's cut is 10-20% of ad spend, you need to negotiate.

Here’s an example of a sound process:

Brand A is spending $50,000 per month. The agency retainer is $2,000 per platform per month plus 5% of ad spend above $60,000 per month (this is the agreed growth milestone). So if the business spends $75,000 one month and the agency has successfully scaled the account to $25,000 more than their prior ceiling, the agency would receive a total of:

$2,000 + (($75,000 - $60,000) * 0.05) = $2,750. The agency has made an additional $750 in additional fees, for scaling the business by 50%.

If the business spends less than the $60,000 milestone, the fee remains the flat retainer. If you’re a Q4-heavy brand or heavily seasonal, you’ll need to figure out a hybrid model that’s perhaps a bit more dynamic.