System design · Platform overview
How advertising platforms work
An advertising platform brings a campaign into something a person is already doing, such as searching, watching a video or asking for advice. How the platform picks and shows an ad affects what you can learn from the campaign report.
On this page
Use this guide to
- Follow an ad from campaign setup to the performance report.
- See which actions show up in the data you can collect.
- Choose what to explore next, from collecting reports to testing campaign changes.
Access
No account or API key is needed for this guide.
Start with the user experience
Begin with what the person is doing: searching for an answer, watching a video, browsing a feed, comparing a product or asking for advice. An ad occupies a particular place in that experience. Where the ad appears shapes what it can look like, what people can do with it and what you can measure.
Ask where an ad can appear, which ads qualify and which settings you can change. Then check what the platform counts. An impression records an ad being shown. A conversion is an action you measure, such as a purchase or signup. Each platform has its own rules for counting them.
From setting up an ad to reading the report
- 01Set up the campaign
- 02Choose and show ads
- 03Measure what people do
- 04Review results
You set the campaign goal, choose the ad content (the creative), set limits on who can see it and decide how much to spend. The platform chooses which ad to show from the ones that qualify. It records actions and groups them into results you can use.
Results feed back into later decisions about bids, budgets and ads. How useful they are depends on your goal, how well the numbers capture it, how quickly they arrive and which settings you can change.
What to ask about each platform
Scroll across to compare all columns →
| Platform | Where people see ads | Questions to ask |
|---|---|---|
| Google Ads | Search, YouTube and other places Google shows ads. | How do campaign type and placement affect where ads show, how reports group results and which metrics you can compare? |
| Meta | Advertising within social experiences, including Facebook and Instagram. | How do the ad, campaign goal and measured actions relate to the results in the report? |
| Amazon Ads | Advertising around shopping and media experiences. | What can you learn from campaign reports, and when do you need to combine data or build audiences? |
| TikTok | Ads as people discover and watch videos. | Which views, interactions and later actions can your software collect? |
| ChatGPT Ads | Advertising in the context of exploration and decisions in a conversation. | How are ads chosen, what can advertisers change and what can they measure? |
| Programmatic: DV360 & The Trade Desk | Buying and measuring ads across many publishers’ sites and apps. | Which tools let you set up campaigns, create reports and download every batch of results? |
Use these questions to compare platforms. Check each platform’s rules for choosing ads, organising campaigns and deciding which ads get credit for a conversion.
Amazon Marketing Cloud helps advertisers analyse data and build audiences alongside their campaign reports. It’s a clean room, a controlled space for working with data under privacy rules. Amazon Marketing Cloud overview
Connect your own software to a platform
Setting up ads, sending records of actions and collecting performance results are separate jobs. They may use different APIs, which are the ways your software asks the platform for data or actions. For example, Google directs DV360 performance reporting to the Bid Manager API. Campaign management uses the Display & Video 360 API. DV360 reporting boundary
- Choose the account and dates. Decide how to group results, such as by campaign and day, and check each metric’s units and meaning.
- Keep track of which batches you’ve collected, which reports are still running and why any data is missing.
- Save the original results, avoid counting a record twice and make it possible to pick up after an interruption.
- Flag late or revised data. A successful request can return figures that will still change.
- When you use a report to make a decision, explain what it covers and what is still uncertain.
Pagination fetches the next batch of the same report. Splitting a request asks for several smaller reports. Streaming lets you read a large file as it arrives. Whichever approach you use, check that you have every batch you asked for.
Start with ChatGPT Ads
The ChatGPT Ads guide follows an ad from the conversation where it might appear to the report an advertiser can read. The reporting example shows how your software can ask for daily campaign results and flag gaps.
Start with the platform guide to see how the parts fit together. If you’re already working on a reporting tool, start with the API design principles.
Platform sources
Official sources used in this guide. Check your own ad account for the features you can access.