Choose an online advertising channel by the job it must do: answer an active search, introduce an unfamiliar product, demonstrate a process, or help a shopper compare a specific item. The choice also depends on the creative you can produce, the destination you can support, and the evidence you can collect.
A platform with a large audience is not automatically a good first test. A small campaign should answer a specific question about the offer and its audience. Start with the business outcome, then decide which paid environment gives that question a fair test.
Compare the work required, not just the reach
On a small screen, scroll the table sideways to read all columns.
| Channel | Useful job | Preparation and limitation |
|---|---|---|
| Paid search | Respond to a query related to the offer. | A clear destination and query review. Search interest may be informational rather than ready to buy. |
| Display and contextual placements | Introduce or reinforce a message in relevant content. | Recognizable visual creative, placement controls, and delivery checks. Impressions do not establish attention. |
| Paid social | Show a product or idea within a social feed. | Creative that makes sense quickly in that setting. A precise audience label does not guarantee a receptive audience. |
| Online video | Demonstrate a process, experience, or product distinction. | Video, captions, sound-off clarity, and an appropriate next step. View definitions differ and a completed view is not a sale. |
| Retail media and sponsored listings | Help shoppers discover or compare a purchasable item. | An eligible, accurate listing, stock, price, and fulfillment. Reported sales may include demand that already existed. |
These are channel roles, not mutually exclusive platform categories. Google's campaign-type overview describes both individual formats and campaigns that reach several kinds of inventory. Check the actual networks, placements, bidding objective, and reporting controls behind a product name.
Retail media can extend beyond a retailer's own pages. For example, Amazon describes Sponsored Products as CPC ads for individual listings that can appear on Amazon and other destinations, with clicks leading to the advertised product detail page. Eligibility and availability depend on the product and marketplace; do not assume every retail-media campaign has the same destination or measurement scope.
Check the destination before buying the visit
A person should be able to understand the offer, assess important limitations, and complete the next step. Test the whole path on a phone: ad preview, landing page, form or checkout, confirmation, and the team's response. Make pricing and eligibility easy to find when they determine whether the offer is suitable.
Confirm that the product can actually be supplied. Inventory, service area, calendar capacity, and support response time are campaign constraints. A qualified lead that nobody answers is still a missed opportunity, even if the ad platform reports a conversion.
List the claims the creative makes, including implications created by images or comparisons. For U.S. advertising, the FTC's substantiation policy explains that advertisers need a reasonable basis for objective claims before disseminating them. Keep the supporting evidence with the approved creative instead of asking a writer to invent a stronger promise.
Also check rights, tracking choices, and landing-page accessibility. When video carries essential information, provide captions and a text explanation that lets people understand it without sound. The media purchase should not be the first time anyone attempts to complete the intended action.
Work through three different starting points
A familiar service with an urgent search
A fictional repair business has a defined service area and appointment capacity. Paid search is a plausible first test when queries describe a service it can provide. Separate relevant repair requests from jobs, training, and unsupported equipment. The objective is a qualified booking, not any form submission.
An unfamiliar product that needs explanation
A fictional reusable packing insert is difficult to understand from a static headline. A short demonstration may explain how it fits and what it protects. Video or social placements could test that explanation, while a detailed page answers size and compatibility questions. Expect some people to need more research before purchasing; instant checkout is not the only diagnostic signal.
A stocked item already sold through a retailer
A fictional kitchen-tool seller has an accurate listing and reliable stock. A sponsored listing could test whether relevant shoppers discover the item. Review product-level contribution, returns, retailer fees, and existing sales alongside attributed orders. More advertising cannot fix a misleading size chart or an unavailable product.
These scenarios are decision examples, not recommendations to buy from a particular platform. If the offer has no clear audience or the destination cannot answer basic questions, improve those foundations first. The online marketing guide supplies a broader planning process.
Define a first test you can interpret
- Write the question. State the audience, offer, and uncertainty you want to resolve.
- Choose one primary outcome. Define exactly what counts and what is excluded, including duplicates or canceled orders.
- Set the scope. Record channel, geography, creative version, destination, spending limit, duration, and review owner.
- Check the measurement. Test events and record attribution windows, date basis, and any gaps.
- Set decision rules. Name operational reasons to pause and what evidence would support another test. Avoid treating the first favorable day as proof.
Download the first-test brief (CSV). It includes blank planning fields and a fictional repair-service example. It is intentionally a brief, not a platform configuration file.
Reserve money and staff time for preparation and review as well as placement. Keep creative, agency, and platform costs visible. For the arithmetic linking click costs to customer contribution, use Cost per Click.
Separate evidence from the platform's score
A campaign optimized for visits can produce more visits without producing more customers. A view, click, lead, and purchase are different outcomes. Record which one the buying system is trying to increase and whether the business outcome is measured separately.
Allow the chosen customer group time to finish its buying cycle. Google's conversion-window documentation explains why the period after an interaction affects counting. Compare like periods and avoid adding each platform's credited sales as though they were all different customers.
When an interface offers automated audience expansion, bidding, or creative variations, inspect what can change and which controls remain available. AI assistance can help draft variants from approved facts, but check the actual combinations shown. Keep an accountable person responsible for claims, exclusions, budget, and the decision to continue.
At review, connect the result to the original question. You may learn that the channel delivers relevant visitors but the offer needs clarification, or that the test is too small to distinguish outcomes. That is more useful than declaring a channel universally good or bad from one campaign.
Related resources
- Select and review contextual placements
- Plan a streaming-audio campaign
- Plan a credible creator endorsement
- Manage campaign work and approvals
- Explore all marketing resources
Researched and updated September 6, 2026. Use the linked primary sources to check the definitions and policies that apply to your tools and audience.