Case Study · Agentic AI

AI-Powered Campaign Pacing & Performance Monitoring Case Study

An AI pacing and performance monitor replaces a media agency's manual pacing workbook, flags drifting campaigns and suggests the fix.

The result: every live campaign across DV360, Amazon DSP and The Trade Desk is watched on a set schedule, drifting campaigns are flagged and routed to the concerned user, and AI recommendations suggest the fix while every correction stays a human decision recorded against a name.

The Customer

A media agency trading desk

The agency's trading team runs campaigns across DV360, Amazon DSP and The Trade Desk, and monitors pacing and performance for every advertiser it manages.

At a glance

Customer: Media agency trading team
Platforms: DV360, Amazon DSP and The Trade Desk
Data: Harmonized in Adverity, pushed to an application database on a set schedule
Users: Users, Executives and Admins
AI pattern: Goal Setting and Monitoring, with a human in the loop

The Problem

Pacing was monitored in a manual workbook

The agency's trading team monitors campaign pacing in a manual workbook. There is no reliable way to know which campaigns are drifting off plan until someone opens the file and works through it, and no record of who noticed a problem or what was done about it.

Current process

A trader exports delivery data from each DSP and pastes it into a shared workbook. Pacing is worked out by hand against flight dates and budgets that are also entered manually. The team scans the sheet to find campaigns that are behind or overspending, fixes them in the DSP, and moves on. Nothing outside the sheet records that the fix happened.

Pain points

Coverage gaps when the person who owns the sheet is unavailable
No alerting, so problems surface only when someone looks
No record of who acted on a campaign or what they changed
Manual entry across several DSPs, each reporting in a different structure
Leadership has no view of overall advertiser health without asking the desk
Problems are often found after the budget is spent rather than while there is time to act

The AI Solution

How the AI pacing monitor works

The AI pacing monitor automates pacing and performance monitoring. Delivery data from DV360, Amazon DSP and The Trade Desk is harmonized in Adverity and pushed to an application database on a set schedule. Monitoring runs on that data, raises a flag when a campaign moves outside agreed thresholds, and routes it to the concerned user. The user fixes the campaign on the DSP and logs the action, which is recorded against their name. Flags clear themselves once the metrics recover, so the status on screen always reflects the data rather than someone's judgment.

DV360 · Amazon DSP · The Trade Desk
Adverity
Application database
Monitoring
Flag routed to the user
Fix on the DSP, action logged

Agentic AI design pattern: goal setting and monitoring

The platform follows the Goal Setting and Monitoring pattern, with a human in the loop for every correction.

Goal

each flight has a measurable target, its budget delivered across its dates

Monitor

every scheduled run measures progress through the pacing index and benchmarks

Goal met?

pacing bands and benchmark thresholds decide whether a campaign is on plan or drifting

Act

a drifting campaign raises a flag, and the user fixes it on the DSP and logs the action

AI recommendations suggest the fix but never apply it, so every correction remains a human decision, recorded against the user's name.

What the AI platform does

The platform replaces the manual pacing workbook the trading desk maintains today. It watches every live campaign on a set schedule, flags delivery and performance that move outside agreed thresholds, and forecasts when budget will deplete and where the campaign will land. AI recommendations suggest the fix at campaign and advertiser level, informed by what the team has done before, and an in-app assistant answers questions on the data directly. Users act on the DSP and log what they did, so every action is recorded against their name.