Paid Advertising Interview Questions

Performance Max is a goal-based campaign type that serves ads across all Google inventory (Search, Display, YouTube, Gmail, Discover, and Maps) from a single campaign using Google's AI to optimize toward a conversion goal.

Unlike traditional campaigns where you control placements, match types, and bidding per channel, PMax hands the distribution decisions to the algorithm based on asset groups and audience signals you provide.

The tradeoff is less granular control and reduced transparency into which placements are actually driving conversions.

In 2026, PMax is table stakes for most ecommerce advertisers, but smart practitioners pair it with brand exclusion lists, negative keyword feeds via the campaign-level negative keyword support, and supplemental Search campaigns to protect branded terms.

Meta Advantage+ Shopping consolidates what would traditionally be multiple ad sets (different audiences, placements, and creative variations) into a single automated campaign where Meta's AI controls audience targeting, budget allocation, and placement across Facebook and Instagram. You provide a product catalog, a daily budget, and creative assets, and the system optimizes delivery toward purchase events using its own signals rather than your defined audience segments. The main benefit is that it leverages Meta's first-party data and behavioral graph, which is especially valuable post-iOS 14 when advertiser-side signals degraded.

The limitation is that you sacrifice audience-level reporting and cannot isolate what is working by demographic or interest segment without running separate tests alongside it.

Incrementality testing measures the true lift a paid channel drives, meaning conversions that would not have happened without the ad exposure, as opposed to conversions the platform claims credit for.

The most common methods are geo-based holdout tests, where you suppress ads in matched regional markets and compare conversion rates, and synthetic control models that reconstruct what performance would have looked like without the spend. It matters more now because iOS 14+ and cookie deprecation have broken the multi-touch attribution chains that last-click and even data-driven attribution models depended on. Platform-reported ROAS is increasingly inflated because platforms can only see a fraction of the touchpoints, so they over-attribute.

Incrementality testing, combined with media mix modeling (MMM), gives you a ground-truth read on which channels are actually moving the needle versus just getting credit.

ATT, rolled out in iOS 14.5 in 2021 and now mature across the iOS install base, requires apps to ask users for permission before tracking them across other apps and websites using the IDFA. Opt-in rates settled around 25-30% globally, meaning Meta, Snap, and other mobile-focused platforms lost visibility into the majority of post-click conversions happening inside iOS apps and on mobile Safari.

The direct impact was that reported ROAS dropped, audience matching degraded, and lookalike models became less precise. Advertisers have adapted through three main levers: implementing the Meta Conversions API (CAPI) to send server-side events that bypass browser and app-level blocking, shifting toward broader audience strategies that rely less on behavioral retargeting, and investing in first-party data infrastructure (email lists, CRM data, and customer match uploads) to seed lookalike audiences from owned data rather than pixel-tracked behavior.

ROAS-based bidding tells the platform to optimize for a revenue-to-spend ratio, treating every dollar of revenue equally. Profit-based bidding weights conversions by their actual profit contribution, so a high-margin product gets a higher target bid than a low-margin one even if the revenue value looks similar. ROAS bidding is simpler to implement and works well when your product margins are relatively uniform, but it actively optimizes against your own interests when you sell products with vastly different margins because it chases revenue, not profit. Profit-based bidding requires passing product-level margin data to the platform via custom conversion values or a value rules layer, which adds setup complexity.

In 2026, most sophisticated ecommerce teams have moved toward contribution margin bidding, passing net margin values into Google's cart data or Meta's CAPI event payloads to align algorithmic optimization with actual business outcomes.

Problem: A sudden unexplained ROAS drop with no intentional changes is almost always caused by one of four things: a tracking break, an external event, a policy enforcement action, or a competitive shift.

Approach: Start with the measurement layer first, not the campaigns. Check that the Conversions column in Google Ads still shows conversion events firing at the same rate as the prior period; if conversions dropped but clicks held steady, the problem is tracking, not performance. Verify the Google Tag fires correctly using Tag Assistant and check that the Conversions API endpoint is returning 200s. If tracking looks intact, pull the Auction Insights report to see if new competitors entered the auction overnight, which can spike CPCs and suppress ROAS without any action on your part. Check the Change History log for any automated rules or Smart Bidding strategy shifts that may have fired.

Result: Most overnight ROAS drops resolve to either a broken conversion tag, an automated bid strategy over-correcting, or a paid media policy flag, all of which have clear remediation steps once identified. If the cause is competitive, you adjust bids, tighten audience signals, and accept temporarily lower efficiency while the algorithm relearns.

Problem: A DTC brand that over-indexed on pixel-based Facebook retargeting has lost its core audience matching and attribution infrastructure, so both targeting precision and measurement accuracy have degraded.

Approach: The first priority is rebuilding the signal infrastructure by implementing Meta's Conversions API server-side, which restores event matching to roughly 90-95% even for iOS users who opted out of tracking. Simultaneously, you audit the CRM to identify which customer segments exist as owned first-party data and upload them as Custom Audiences to seed Advantage+ Shopping campaigns and lookalike generation. On the measurement side, you set up a geo holdout test to establish true incrementality baselines before scaling spend, because the platform-reported ROAS is likely inflated and you need a reliable read before making budget decisions.

Result: Within 60-90 days of full CAPI implementation and first-party audience activation, most brands recover 60-75% of their pre-ATT retargeting performance while also building a more durable data foundation that is not dependent on third-party tracking.

Problem: A large discrepancy between platform-reported attribution and an MMM output creates a real decision risk: cutting spend based on the wrong model either wastes profitable budget or keeps inefficient spend running.

Approach: Before recommending any budget change, you run a geo-based holdout test to get a direct incrementality read. Split matched geographic markets, suppress Facebook entirely in the holdout group for three to four weeks, and measure the conversion rate difference. This gives you an empirical incrementality figure that neither the attribution tool nor the MMM can dispute, since it is a direct causal measurement rather than a modeled estimate.

Result: In most cases the incrementality test lands between the two estimates, and you use that as the basis for the budget decision. If Facebook's true lift is closer to 20-25%, the right action is a modest trim rather than a 60% cut, and you document the methodology so future budget calls are grounded in holdout data rather than model disagreement.