CTO Series 2: Retail Media Pacing: Why “Even Spend” Is the Wrong Goal

By Francisco Larrain, Co-Founder & CTO, Topsort

‍

Start from the beginning: CTO Series 1: Why Static Floors Break →

Pacing is often misunderstood.

Many systems default to a simple idea: spend the budget evenly over time to ensure smooth delivery and avoid spikes. That works in stable environments.

Retail media is not stable.

Traffic shifts throughout the day. Conversion rates vary by placement. Demand intensity changes with promos, seasonality, and competition. When the environment is volatile, even pacing is not control. It is a drift away from opportunity.

The Real Goal of Pacing

The goal of pacing is not even spend. It is predictable delivery under uncertain opportunity.

In retail media, opportunity is not distributed evenly over time. Some moments carry more value than others, and high-intent windows concentrate disproportionate performance.

If pacing is blind to opportunity, two things happen:

• Budgets deplete early during high-demand windows
• Underdeliver during strong performance periods

Both damage advertiser trust.

Pacing Is a Feedback System

In modern retail media infrastructure, pacing is not a schedule. It is a feedback loop.

It continuously evaluates:

• Are we ahead or behind expected delivery?
• Has available opportunity changed?
• Has auction competition shifted?
• Are we maintaining performance while meeting spend targets?

Based on these signals, the system adapts in real time within defined constraints. The objective is responsiveness without instability.

If the system reacts too slowly, it misses opportunity. If it reacts too aggressively, it introduces volatility into the auction itself.

Why This Matters in Volatile Markets

Volatile markets amplify pacing mistakes.

If the system reacts too slowly, it underdelivers and misses valuable inventory. If it reacts too aggressively, it overspends and distorts auction dynamics. Both outcomes degrade marketplace health.

Retail media at scale requires pacing that:

• Accounts for demand variability
• Protects advertiser performance
• Avoids creating auction distortions
• Works without constant human intervention

This is infrastructure behavior.

The CTO View

At scale, pacing becomes a control problem.

You are balancing delivery targets, performance constraints, and market conditions simultaneously in real time. If those forces are managed through static schedules, the system will break under volatility.

Retail Media 3.0 requires pacing logic embedded into the auction layer itself. Not as an external constraint, but as part of the decision system.

This is what we built at Topsort. Our pacing layer doesn't sit outside the auction. It's embedded in it, continuously adapting to traffic shifts, ROAS targets, and demand signals in real time.

Because in retail media, timing is value.