Costs
Part of Paid social advertising, tested against experience
Paid social advertising case studies that hold up to scrutiny
Paid social advertising case studies need review for source, date, objective, audience, creative, measurement, costs, selection, transfer limits, and missing facts.
What to take away
- A paid social case study is a selected account, not an experiment. It holds up only when the page gives a baseline, a period, a spend figure, a denominator and a counterfactual.
- The cases worth reading by name are LinkedIn Scaling Champions, Dreamdata, Duolingo, Airbnb and Gymshark. None of them isolates paid social.
- Audit one claim at a time and rebuild it as a record a colleague can reproduce from your notes.
- Vendor numbers are leads for a local test. Compare the structure with two paid media strategy cases before copying a tactic.
Five named cases worth reading
Five write-ups appear again and again in paid social conversations. Treat them as claims with authors, since the author usually sells what the case describes.
| Case | Publisher | What it supports | What is missing |
|---|---|---|---|
| LinkedIn Scaling Champions | LinkedIn Marketing Solutions | An always-on mix of organic and paid activity for a B2B audience | Spend, sample, and pipeline that would have arrived anyway |
| Dreamdata | LinkedIn customer story | Employee advocacy, video and Thought Leader Ads tied to conversion tracking | Reported influence, not a tested effect |
| Duolingo | 2021 IPO prospectus | Word of mouth carried most learner growth; revenue reported near $251 million for 2021 | Paid social is not separated from organic TikTok reach |
| Airbnb | 2020 prospectus and later results | Marketing cut hard in 2020; revenue reported near $3.4 billion, then $6 billion | Paid social share not disclosed; no counterfactual |
| Gymshark | Company history and 2020 funding reports | Instagram and athlete partnerships built a brand valued near £1 billion | Brand scale, not a measured social-only effect |
Sport England's 2015 This Girl Can campaign ran on Facebook, Instagram and Twitter beside broadcast advertising. Its evaluation used a commissioned tracking survey rather than platform-attributed conversions, which separates reach from measured behavior.
Keep vendor percentages beside published paid social benchmarks with caveats, which give you a range to argue against. Add a read date beside each row in your copy, because vendor pages get edited without changelogs.
Example: the Airbnb claim rebuilt
The claim as it usually gets repeated: Airbnb cut performance marketing hard and demand came back anyway.
That sentence is testable, and the public record rebuilds it into a table.
Airbnb claim rebuilt
- Claim
- Bookings and revenue recovered while performance marketing stayed down
- Baseline
- 2019 revenue reported near $4.8 billion
- Period
- 2020 through 2021
- Spend
- Deep cuts to performance marketing in 2020; the paid social share is not broken out
- Denominator
- Nights and experiences booked, a company-wide measure
- Counterfactual
- None published
The record weakens the claim. Company-wide revenue cannot be assigned to one channel, and Airbnb changed pricing and refund rules in the same window. It still beats most vendor pages, because its figures can be checked.
A six-step record for one claim
Turn a published claim into a record someone else can reproduce.
Audit published evidence
- Apply FTC substantiation policy to performance claims
- Check reviews guidance for hidden relationships
- Separate provider, customer, measured, independent claims
- Record selection criteria and missing evidence
- Reproduce one material claim with local data
- Label unverifiable methods as leads for testing
- Save the page and write the read date on your copy.
- Copy the claim sentence exactly, using the publisher's own metric name.
- Fill in baseline, period, spend and denominator, and mark every gap as not disclosed.
- Note who measured the result and who paid for the publication.
- List what else changed in the same window, including seasonality and other channels.
- Write the transfer conditions and one local test with a stop rule.
A claim with four blank fields is a lead, not evidence. Worked paid social examples that show their reasoning beat headline lifts, because you can see which variable the team changed.
US and Canadian rules on advertising claims
The FTC requires a reasonable basis for objective advertising claims before they publish, under its advertising substantiation policy. Ask the advertiser for the evidence file behind the percentage, not a summary deck.
The FTC's guide to reviews and endorsements covers paid incentives and hidden relationships. Ask whether the customer was paid, given credit, or picked from the success list.
In Canada, PIPEDA covers the audience data a social pixel collects, as the Office of the Privacy Commissioner explains. If a case study credits retargeting, ask what consent covered those people.
Common questions
Can a vendor case study be good evidence?
Yes, for the account it describes, when the period, spend and denominator are stated. It stops being evidence the moment you move it to another audience.
Should I use a case study result as my benchmark?
No. Treat it as a hypothesis, then test it on your own traffic and judge it on your own cost per acquisition.
Which number is missing most often?
The counterfactual. Without a holdout or a clean baseline, a percentage describes a change, not a cause.
What should I do with a claim that survives?
Run a smaller version yourself and attach a stop rule. Fixing paid social advertising starts with isolating one change rather than rewriting the whole account.
