Where to find customers: search or social

Does your buyer start with a question or notice you in a feed?

Picture an actual buyer. What happened before they started looking? How did they describe the problem? How urgently do they need to act? “Our audience is aged 25 to 55” provides little guidance. “An operations manager needs a replacement supplier before opening a branch” gives you a hypothesis to investigate.

When buyers actively search for a solution they can name, search advertising is worth examining. When the offer is unfamiliar or easier to demonstrate than describe in a query, social advertising offers a different starting point. These are research hypotheses, not rules that assign entire industries to one channel.

Choosing among every option, including maps, marketplaces and direct sales, is covered in our separate guide to the first acquisition channel. This article deals with a narrower and very common choice: Google search ads or Meta social ads.

How the two tests differ before you spend anything

Google Search can show ads alongside results for related queries; it is one campaign type within Google Ads. Meta describes inquiries through forms, messaging and calls. Specify the actual campaign and inquiry route, and check whether the required format is available in your market.

Meta Ads Manager offers six simplified campaign objectives. The Leads objective collects inquiries through messages, phone calls or sign-ups. Beyond the objective, you choose the conversion location, such as a website or messaging app, and the performance goal: the outcome the system optimises delivery for. That choice determines who the algorithm goes looking for.

Google versus Meta is not an A/B test

A conclusion that social works worse than search does not follow from a test like this

Within one platform you can run a genuinely controlled test. In a Meta A/B test, each version is shown to a separate audience segment so that nobody sees both, and Meta recommends the same budget for each version. Google Ads experiments split traffic between a control campaign and a treatment campaign.

No such split exists between Google and Meta. The people, auctions and moments of purchase all differ. The accurate description is: “we compared two acquisition hypotheses using one outcome rule.” A conclusion that social works worse than search does not follow from a test like this.

Turn the idea into one testable hypothesis

Return to the illustrative company selling warehouse software implementation. Its search test reaches people already choosing a system. Its social test shows warehouse managers how to spot stock discrepancies and offers a discussion. The two groups may be at different stages of buying, so comparing form-submission costs alone can mislead the owner.

Write down the offer and the next action you need. For example: “Managers facing this inventory problem will discuss implementation after seeing this evidence.” Specify the information that will establish whether an inquiry fits. If those facts require a conversation, arrange the conversation and decide how its outcome will be recorded.

A one-page brief reduces the chance that marketing and sales judge the same test by different standards. You can draft it in an hour with the steps below.

1. Define the audience and the situations deliberately excluded from the test.
2. Write down the product, conditions and next step the ad promises.
3. Name who handles each inquiry and what they record in the CRM.
4. Set one definition of a qualified inquiry for both channels.
5. Fix the spending limit and the date of the interim review.
6. State how long buyers normally need and when sales can be assessed.

The illustrative test: a cheap inquiry, an expensive customer

The cheap lead was cheap because it was easy to submit, not because the customer was cheaper

Here are the numbers from the opening. They are an illustrative example, not company results or a market benchmark. Each channel received $6,000 over four weeks, every inquiry was observed for the same 60 days after first contact, and one qualification rule applied to both channels.

Search produced 40 inquiries, 24 completed conversations, 12 qualified prospects and 3 purchases. Social produced 80 inquiries, 32 conversations, 10 qualified prospects and 2 purchases. Cost per inquiry: $150 versus $75. Cost per qualified inquiry: $500 versus $600. Ad spend per purchase: $2,000 versus $3,000.

Most social inquiries never reached a conversation: only 40% did, against 60% for search. Only 12.5% of social inquiries became qualified prospects, compared with 30% from search. The cheap lead was cheap because it was easy to submit, not because the customer was cheaper.

The cost at each stage changes the winner

Judged on the first column alone, social wins. At the qualified-prospect stage, search is already $100 cheaper, and at the purchase stage $1,000 cheaper. But the last stage rests on three contracts and two. One extra social purchase would make ad spend per purchase identical: $6,000 / 3 = $2,000.

For an interim decision, cost per qualified inquiry is more dependable: it rests on 12 and 10 observations, not 3 and 2. Treat purchases as a check on direction rather than a final verdict on the channel. Calculate the full customer cost, including sales effort and tools, separately, as in our guide to CPL, CAC and ROAS.

Is there enough data for the algorithm, and for you, to learn?

Meta says an ad set exits the learning phase once delivery stabilises, which usually takes about 50 results in the week after the last significant edit. During learning, performance is less stable and cost per result is usually higher. Editing the ad set during this period resets learning.

In the illustrative test, social delivered 20 inquiries a week and only 2.5 qualified prospects. Optimising delivery for qualified prospects would leave the ad set far short of 50 results a week. Getting 50 inquiries a week at $75 would take about $3,750 a week instead of $1,500. That is arithmetic on teaching numbers, not a promise that the cost per inquiry holds as spend grows.

1. Estimate how many results a week your budget can buy at the expected cost.
2. If that is well below 50, optimise for a more frequent event and count qualified prospects separately in the CRM.
3. Avoid changing ads, audiences and budget without a reason while learning is under way.
4. Compare channels only after an equivalent observation period.

Finish the test with a decision, not a report

If inquiries are scarce, check whether the offer was understandable and received enough exposure to examine the hypothesis. If inquiries do not fit, review the promise and screening criteria. If suitable prospects fail to progress, examine actual conversations and purchase conditions. Each pattern calls for a different investigation.

In the illustrative example, a sensible decision is to keep search running, rewrite the social ad so its promise screens out unsuitable buyers, and test it for another four weeks under the same rules. Moving the entire budget to the channel with the cheaper lead would have been a mistake.

Ask the team for three conclusions alongside the platform report: what was tested, what evidence was collected and what will change next. Preserve the test conditions. The next budget can then examine a new assumption instead of repeating the same attempt under a new campaign name.

This journal is published with the involvement of WEBNOSTRA, a company that provides advertising services. You can run the test plan in this article in-house or with any agency.

Key decisions

1. Describe the buying situation before choosing a platform
2. Compare equivalent outcomes over equivalent observation periods
3. Decide in advance how each test result will change your next action

What to discuss with your team

• A specific buying situation is documented
• Campaign type, performance goal and inquiry route are selected
• Marketing and sales agree on qualification
• Spending and observation limits are recorded
• Every percentage in the report shows its underlying count
• The result leads to a specific next action

Sources

1. Google Ads: About the Google Search Networksupport.google.com · checked 29/10/2026
2. Meta: forms, messaging and calls, November 2023about.fb.com · checked 31/10/2026
3. Meta Business Help Center: How to choose the right ad objective in Meta Ads Managerfacebook.com · checked 31/10/2026
4. Meta Business Help Center: About A/B testingfacebook.com · checked 31/10/2026
5. Meta Business Help Center: About the learning phasefacebook.com · checked 31/10/2026
6. Google Ads API: Experiments overviewdevelopers.google.com · checked 31/10/2026