Black Friday Creatives: Judge by Sales, Not Clicks

Key takeaways

• CTR and CPL measure interest; attributed CPA and conversion rate per ad measure sales. Decide on the second group.
• Compare creatives in the same period, on the same audience and with the same offer; changing all three at once invalidates the test.
• Read by cohort: leads that came in through ad X and bought on any date, not sales that landed on the day.
• Leads with no origin (no ad id or ctwa_clid) go under "unattributed" and are never distributed among creatives.
• Do not pick a winner with fewer sales than the minimum agreed in the brief; a single sale is a hint, not proof.

Why don't CTR and CPL tell you which Black Friday creative sells?

Because Meta's messaging campaign optimizes for conversations started, and the creative with the most clicks is often the one that leaves the conditions least clear. CTR measures attention; CPL (cost per conversation) measures how much it cost to bring someone into WhatsApp. Neither knows whether the person bought.

Three common situations where a high CTR misleads:

• The ad promises "up to 70% off" and the lead finds out on WhatsApp that the discount applies to one item. Lots of clicks, lots of short conversations, few sales.
• The ad omits the price and attracts the curious. Procon-SP (São Paulo's consumer protection agency, 2025) recorded that 7.26% of Black Friday complaints were about offers not honored or misleading advertising, and article 37 of the CDC (Brazil's Consumer Protection Code) bans misleading advertising, including by omission.
• The ad uses a scarcity trigger without genuinely limited stock. High conversations, low trust, low sales.

Decision rule: if one creative's CTR is far above the others and its sale conversion rate is below, the creative is selling the wrong promise. Fix the ad before blaming the sales team. The same pattern shows up with unqualified leads from click-to-WhatsApp ads.

Which per-creative metrics decide on Black Friday?

The metrics that decide are attributed CPA, conversation-to-sale conversion rate and attributed revenue per creative, read together with attribution coverage. All of them require the sale to carry the originating ad id.

Two rules for reading the table:

1. If coverage is below 50%, no comparison between creatives is reliable, because half the sales could belong to any creative. Fix tracking first; the article on tracking from the ad to WhatsApp lists what to check.
2. Different currencies are never added up in attributed revenue. A client selling in reais and in dollars has two creative reports.

How do you compare two creatives by the path to the sale?

Compare by the cohort's attributed CPA, with spend, conversations, qualified leads, sales and revenue side by side. A hypothetical example, step by step:

1. Video A: spend R$ 1,200, 300 conversations, 150 qualified, 12 attributed sales, average order R$ 240.
2. Image B: spend R$ 1,200, 150 conversations, 105 qualified, 12 attributed sales, average order R$ 240.
3. CPL: A = 1,200 ÷ 300 = R$ 4.00; B = 1,200 ÷ 150 = R$ 8.00. On CPL, A wins by a mile.
4. Qualification rate: A = 150 ÷ 300 = 50%; B = 105 ÷ 150 = 70%. B attracts a better-fitting lead.
5. Sale conversion: A = 12 ÷ 300 = 4%; B = 12 ÷ 150 = 8%.
6. Attributed CPA: A = 1,200 ÷ 12 = R$ 100; B = 1,200 ÷ 12 = R$ 100. A tie on cost per sale.
7. Hidden cost: A brought 150 extra conversations to the sales team without generating extra sales. If each conversation costs 5 minutes of an agent's time, A consumed 12.5 extra hours of team time for the same result.

Decision: with equal CPA, the creative that uses less of the sales team wins, because on Black Friday the queue is the scarce resource. The Harvard Business Review (2011) showed that companies responding within 1 hour are about 7 times more likely to qualify the lead than those that take longer; 150 extra conversations in the queue worsen the response for every lead, including those from creative B.

Important caveat: 12 sales per creative is a small sample. If the brief set a minimum of 20 sales per variation (hypothetical), neither one has a winner yet; the decision is to keep both until the minimum is reached.

How do you set up a creative test that does not mislead?

A valid test changes one variable, keeps audience and offer fixed, runs in the same period and has a minimum number of sales defined up front. Without those four points, the difference you find may come from the auction, not from the creative.

Procedure:

1. Write the creative's hypothesis: "show the product in use", "explain the payment terms" or "answer the delivery-time objection". One hypothesis per variation.
2. Fix audience, placement and offer. If the offer changes during the test, end the test and start another.
3. Make sure each variation has its own ad with its own ad id; the lead that arrives from the click brings the ctwa_clid, and that identifier is what links the conversation to the creative. Trackable links with utm parameters serve paths outside the ad (bio, website); they do not replace the ad id.
4. Define the minimum sales per variation and the observation window (compatible with the product's decision cycle) before switching it on.
5. Read by cohort at the end of the window, and only then pause the loser.

Common mistake: swapping the creative in the middle of Black Week because it "isn't performing" after 6 hours. Why: the 6-hour cohort has not had time to buy, and Meta restarts delivery of the new creative in the most competitive auction of the year. In November, compare cost with the account's own previous November, not with October.

What does the conversation reveal about the creative?

The conversation shows which promise the lead understood, and that points to what to fix in the creative. Read a sample of conversations per creative, not just the numbers.

Questions for the reading:

• What is the lead's first question? If 30% (hypothetical) ask "which product does the discount apply to?", the creative is vague.
• Does the lead mention something the creative did not promise? A sign of an ambiguous creative or one reused from another campaign.
• How many disappear after the first reply? If the drop-off rate after the price is high for one creative and low for another, the first is attracting people who will not accept the price.

A CRM with attribution, such as RASTRO, keeps the conversation history alongside the ad origin and shows acquisition analysis per creative, which lets you cross the number with the sample. Any tool will do, as long as it links the conversation to the ad id and marks as "unattributed" whatever arrived without it.

Which creative analysis mistakes cost the most?

The most expensive mistakes are the ones that pick the wrong winner with confidence. The list, with the why:

1. Switching off the ad with the lowest CTR. A lower CTR can come with a more qualified lead; check attributed CPA first.
2. Picking a winner on a single sale. With a small sample, the difference is noise; audiences, budget and campaign stage explain more than the creative.
3. Splitting unattributed sales among creatives. Splitting is invention. "Unattributed" is a category, not a shame.
4. Comparing creatives from different periods. The November auction changes the cost; a creative that ran in October does not compete with one from November.
5. Changing the offer and the creative together. Impossible to know what caused the change.

How these numbers go into the client report is in a report that helps the client decide, and the budget decision per creative in Meta Ads budget for Black Friday.

Action plan: what to do this week

Use this sequence to test creatives by sale, not by click.

Frequently asked questions

Sources

• Procon-SP, Black Friday 2025 partial report (2025)
• Harvard Business Review, The Short Life of Online Sales Leads (2011)
• Código de Defesa do Consumidor (Brazilian Consumer Protection Code), Lei 8.078/1990