How to Structure Paid Social Testing Properly in iGaming
How to structure paid social testing properly
To structure paid social testing properly, iGaming teams should start with one commercial question, isolate the most important variable where practical, define success before launch and evaluate results using player-quality metrics rather than front-end conversions alone.
Paid social budgets are often wasted because campaigns attempt to answer too many questions at once.
If the audience, creative, offer, landing page, optimisation event and budget strategy all change together, a strong result may still tell the team very little about what actually caused the improvement.
For regulated operators, testing also needs to account for market eligibility, approved messaging and the difference between cheap registrations and valuable depositing players.
In short: a strong paid social testing programme uses a clear hypothesis, controlled variables, sufficient budget, pre-agreed decision rules and downstream quality checks. The goal is not simply to find a winning advert. It is to build a repeatable learning system that makes each future test more intelligent.
Start with a commercial question
Every paid social test should answer one defined question connected to a commercial outcome.
Questions such as:
“Which advert performs best?”
are too broad.
A better question might be:
“Does football-focused product creative produce a lower cost per qualified first-time depositor than a generic sportsbook offer among eligible prospective bettors?”
Or:
“Does product-led casino creative generate stronger D7 value than welcome-offer creative within the same prospecting audience?”
These questions make the test actionable.
They specify:
What is changing.
Who is being tested.
Which outcome matters.
What the team will learn.
What decision may follow.
Diagnose the funnel before choosing the test
The right experiment depends on where performance is constrained.
Weak reach or delivery
Possible issues include:
Audience too narrow.
Market size.
Bid strategy.
Budget.
Platform eligibility.
Excessive campaign fragmentation.
The priority may be audience or account-structure testing rather than creative.
Strong impressions but weak CTR
Potential causes include:
Weak opening hook.
Irrelevant proposition.
Poor creative fit.
Low sporting relevance.
Message fatigue.
Unsuitable format.
Creative testing may be appropriate.
Strong clicks but weak registrations
Potential causes include:
Advert-to-page mismatch.
Low-intent traffic.
Poor landing page.
Confusing offer.
Registration friction.
Changing the advert alone may not solve the issue.
Strong registrations but weak FTD conversion
Potential causes include:
Low player intent.
Verification friction.
Payment issues.
Misunderstood promotion.
Overly broad targeting.
Creative that attracts curiosity rather than deposit intent.
The test may need to focus on proposition or audience quality.
Healthy FTDs but poor retained value
This may suggest:
Bonus-led traffic.
Weak product fit.
Unsuitable audience.
Low repeat behaviour.
Poor acquisition economics.
The correct test should focus on player quality rather than reducing front-end CPA.
Write the hypothesis before building the campaign
A useful hypothesis should contain:
The variable.
The audience.
The expected behaviour.
The commercial success measure.
For example:
Hypothesis:
A short-form vertical video showing the sportsbook product within the first three seconds will generate a lower cost per qualified FTD than a static offer-led asset among prospecting users in the target market.
This gives the campaign team a clear reason for the test.
It also creates an audit trail when stakeholders later ask:
Why was this asset produced?
Why was this audience selected?
Why was budget allocated here?
Why was the variation scaled?
The answer should come from the hypothesis rather than hindsight.
Separate the variables in paid social testing
Paid social campaigns contain several variables.
The most useful groups are:
Creative
This includes:
Concept.
Opening hook.
Visual treatment.
Talent.
Product interface.
Static versus video.
Video duration.
Copy.
Call to action.
Sporting context.
Audience
This may include:
Broad delivery.
Available interests.
Lookalikes where appropriate.
Existing-customer exclusions.
Market segmentation.
Product-based audiences.
Remarketing segments.
First-party seed groups.
Offer and proposition
This may include:
Welcome offer.
Free bet.
Odds boost.
Free spins.
Product feature.
Game range.
Payment proposition.
Trust message.
Brand-led message.
Conversion path
This includes:
Landing page.
Registration route.
Promotional-code journey.
Web versus app.
Optimisation event.
Verification flow.
Delivery mechanics
This may include:
Campaign objective.
Bid strategy.
Budget structure.
Placement.
Attribution configuration.
The cleaner the experiment, the easier the result is to interpret.
Test one primary variable where possible
If the objective is to test creative, hold other major factors as consistently as possible.
Keep stable:
Audience.
Offer.
Landing page.
Optimisation event.
Budget.
Market.
Placements where appropriate.
Then vary the creative.
If the objective is to compare audiences, use the same proven:
Creative.
Offer.
Landing page.
Optimisation event.
This isolates the audience difference more effectively.
Perfect isolation is not always possible, but unnecessary variables should be controlled.
Know when exploratory testing is appropriate
Not every test needs to isolate one variable immediately.
When entering a new:
Market.
Product.
Audience.
Platform.
the business may not yet know which combination is viable.
In this situation, an exploratory test may compare broader strategic packages.
For example:
Route A
Sportsbook-led creative.
Broad eligible sports audience.
Product proposition.
Sports-focused landing page.
Route B
Casino-led creative.
Different eligible audience.
Welcome proposition.
Casino landing page.
This can identify which direction deserves further investment.
But it should be labelled correctly.
An exploratory test tells the team:
“This combination appears more promising.”
It does not necessarily prove which individual component caused the difference.
Follow successful exploratory routes with cleaner validation tests before reallocating substantial budget.
Use a paid social testing hierarchy
Not every question deserves the same priority.
A practical testing roadmap has three levels.
Level 1: Prove the fundamentals
Before testing small creative differences, establish whether the campaign has a commercially viable foundation.
Test questions such as:
Which market can support the campaign?
Can the account generate sufficient conversion volume?
Which audience breadth is viable?
What optimisation event should be used?
Which broad proposition generates qualified interest?
Is the landing-page journey suitable?
Can the campaign produce acceptable FTD economics?
The purpose is to find a route with enough potential to justify refinement.
Level 2: Improve the proposition and creative
Once a viable campaign structure exists, test:
Product-led versus offer-led messaging.
Creative concepts.
Opening hooks.
Sporting context.
Product features.
Different value propositions.
Trust-led messaging.
Video versus static.
Different creative styles.
This is where much of the meaningful creative learning happens.
Level 3: Optimise marginal gains
After the main proposition and structure are validated, refine:
Formats.
Video duration.
Placements.
Call-to-action wording.
Creative refreshes.
Frequency management.
Landing-page variants.
Minor copy differences.
Testing small details before proving the fundamentals can waste budget.
A button or headline variation has limited value if the campaign is optimising towards registrations that rarely deposit.
Choose the deepest reliable conversion event
One of the most important testing decisions is the event used to evaluate performance.
Cost per registration is often useful diagnostically.
It is rarely sufficient as the only commercial success measure.
Possible conversion events include:
Completed registration.
Verified registration.
KYC approval.
First-time deposit.
Qualified first-time deposit.
Second deposit.
Early retained player.
Net gaming revenue after a defined period.
Early player-value band.
The ideal event is:
Commercially meaningful.
Accurately measured.
Available quickly enough.
Frequent enough to support learning.
Balance speed against certainty
There is a natural trade-off.
Fast signals
These may include:
CTR.
Landing-page views.
Registration.
Cost per registration.
They arrive quickly but provide limited certainty about value.
Mid-funnel signals
These may include:
Verification.
First deposit.
Payment success.
First bet.
Cost per qualified depositor.
These provide stronger commercial evidence.
Mature signals
These may include:
Second deposit.
D7 retention.
D30 retention.
Bonus-adjusted revenue.
Net gaming revenue.
Cost per retained player.
Predicted value.
These provide better information but arrive later.
A practical testing system uses both early and mature signals.
For example:
During the test:
Pause clearly inefficient creative using registration cost, deposit conversion and FTD data.
After initial completion:
Compare the remaining variants using qualified FTD rate and D7 value.
When cohorts mature:
Revisit the decision using D30 value or retained-player economics.
This prevents teams from waiting too long to make obvious decisions while still checking whether early winners produced valuable players.
Define success before the test starts
Do not decide what counts as success after seeing the results.
Agree the rules beforehand.
These may include:
Minimum spend.
Minimum impressions.
Minimum clicks.
Minimum registrations.
Minimum qualified conversions.
Acceptable CPA.
Minimum FTD conversion.
Quality threshold.
Maximum loss before pause.
Minimum duration.
Required player-value check.
This reduces bias.
Without pre-agreed rules, teams can:
Declare a favourite asset the winner too early.
Keep weak variations running because someone likes them.
Change the success metric after seeing the data.
Overreact to one strong day.
Build quality guardrails into every test
A paid social result should be protected by downstream quality metrics.
Useful guardrails include:
Verification rate.
First-deposit rate.
Payment success.
First-to-second deposit.
Bonus cost.
D7 retention.
Net revenue.
Cost per retained player.
Rejected or suspicious conversions.
Relevant player-protection outcomes.
For example:
A variation may produce:
20% lower cost per registration.
But also:
30% lower FTD conversion.
Higher bonus dependency.
Lower D7 value.
That creative should not automatically be treated as the winner.
Build a paid social test matrix
Before launching, record:
Test name.
Commercial question.
Funnel issue.
Hypothesis.
Market.
Product.
Platform.
Audience.
Control.
Variation.
Variable being tested.
Variables held stable.
Offer.
Landing page.
Optimisation event.
Budget.
Primary KPI.
Quality guardrails.
Minimum spend.
Minimum conversion volume.
Test period.
Stopping rule.
Approval status.
Decision owner.
This helps prevent overlapping experiments and makes results easier to interpret later.
Example: paid social creative test
Commercial question:
Can product-led sportsbook creative improve qualified acquisition compared with the current welcome-offer campaign?
Hypothesis:
Product-led video will generate a lower cost per qualified FTD than offer-led creative within the existing prospecting audience.
Control:
Current welcome-offer video.
Variation:
Product demonstration focused on a sportsbook feature.
Stable variables:
Audience.
Market.
Budget.
Landing page.
Optimisation event.
Placements.
Primary metric:
Cost per qualified FTD.
Quality guardrails:
Verification.
First-deposit rate.
D7 retention.
Bonus cost.
Decision:
Scale the product concept if the CPA remains within threshold and downstream quality is stronger.
Example: audience test
Commercial question:
Does a broader eligible audience outperform an interest-based audience without reducing player quality?
Hypothesis:
Broad delivery will reduce cost per qualified FTD while maintaining the agreed D7 value threshold.
Stable variables:
Creative.
Offer.
Landing page.
Optimisation event.
Market.
Variable:
Audience construction.
Primary KPI:
Cost per qualified FTD.
Quality guardrails:
Registration-to-deposit rate.
D7 activity.
Early player value.
This makes the result much easier to interpret than testing a new audience with different creative at the same time.
Structure campaigns to protect the test
The exact structure depends on the:
Platform.
Market.
Budget.
Conversion volume.
The core principle is consistent:
Give each test enough delivery to learn without creating unnecessary fragmentation.
Creative testing structure
Where practical, test creative variations inside one stable campaign environment.
This helps keep:
Audience.
Budget conditions.
Optimisation event.
Market.
more consistent.
Avoid creating separate campaigns for every creative unless there is a clear reason.
The more structure is fragmented, the harder it becomes to separate creative performance from delivery differences.
Audience testing structure
Audience tests may require separation at ad-set level.
This is only useful when each audience receives enough:
Budget.
Reach.
Conversions.
If the budget is spread too thinly across many audiences, none may receive enough data to produce useful evidence.
The result becomes noise rather than learning.
Avoid unnecessary audience overlap
Running multiple near-identical campaigns against similar audiences can create:
Auction overlap.
Unstable delivery.
Higher frequency.
Internal competition.
Harder cost interpretation.
It can also encourage constant manual budget changes.
A test becomes difficult to interpret when the campaign structure changes every day.
Avoid excessive manual intervention
During a controlled test, avoid repeatedly:
Moving budgets.
Changing bids.
Adding creative.
Removing placements.
Changing targeting.
Replacing landing pages.
unless a clear stopping rule or operational issue requires it.
Every major intervention changes the conditions of the experiment.
If the test is constantly altered, the final result may no longer answer the original hypothesis.
Match the budget to the optimisation event
A test needs enough spend to generate the event being evaluated.
If first-time deposits occur infrequently, a small daily budget may not produce enough conversions to distinguish performance from random variation.
Possible responses include:
Increase the budget.
Extend the test.
Broaden the audience.
Consolidate campaigns.
Use a higher-volume proxy temporarily.
The correct solution depends on the commercial context.
Use proxy events carefully
A higher-volume event such as registration can help when FTD volume is too low.
But a proxy is useful only if it predicts the deeper outcome.
Before using registration as the optimisation event, analyse whether:
Registration correlates with verification.
Verification correlates with first deposit.
The relationship remains consistent by channel.
It remains consistent by campaign.
It remains consistent by market.
A proxy that generates volume but poorly predicts player quality can push the algorithm in the wrong direction.
Do not fragment low-budget accounts
One of the most common testing mistakes is creating:
Too many campaigns.
Too many ad sets.
Too many audiences.
Too many creatives.
for the available budget.
A £200 daily budget split across twenty competing variations may look sophisticated but provide weaker learning than a simpler structure with four meaningful variations.
Concentration usually improves the quality of the evidence.
Treat compliance as a testing constraint
In iGaming, compliance defines what is available to test.
It should influence:
Creative concept.
Offer.
Wording.
Talent.
Market.
Audience.
Landing page.
Age settings.
Promotional conditions.
Calls to action.
Do not build a test around a concept that cannot be approved consistently.
Create approved creative territories
Teams can define pre-approved strategic territories for each market.
Examples may include:
Product-led sportsbook.
Event-led sports content.
Brand trust.
Casino game selection.
App experience.
Approved promotional routes.
Each territory can contain reusable:
Copy modules.
Terms treatments.
CTA options.
Age messaging.
Safer-gambling elements.
Market-specific restrictions.
This allows the creative team to test new hooks and executions without reopening the entire compliance discussion each time.
Record market restrictions with test results
A creative may produce strong performance but be approved only:
In one market.
On one platform.
For one product.
During one promotional period.
That result is not automatically scalable.
The testing log should record:
Approval status.
Market restrictions.
Platform restrictions.
Offer expiry.
Product restrictions.
This prevents teams from treating a locally successful asset as a universal winner.
Use comparable testing windows
Paid social performance can fluctuate because of:
Day of week.
Fixtures.
Major tournaments.
Paydays.
Promotional calendars.
Competitor activity.
Platform changes.
Audience saturation.
Market news.
A sportsbook asset tested during a major final may not represent ordinary league-week performance.
A casino campaign may perform differently around a large promotion.
Testing context should be recorded and considered.
Do not use a fixed testing duration blindly
There is no universal rule that every paid social test should run:
Three days.
Seven days.
Two weeks.
Duration should depend on:
Spend.
Conversion volume.
Delivery stability.
Market.
Decision importance.
The test should continue long enough to generate enough evidence across meaningful delivery periods.
Weak variants should not continue indefinitely simply to reach an arbitrary day count.
Avoid declaring winners during early volatility
The first day of a test can be misleading.
Early performance may be influenced by:
Platform exploration.
Small samples.
Uneven delivery.
A single large conversion.
A particular fixture.
Time-of-day effects.
Treat early performance as directional.
Wait for enough delivery and conversion volume to support the decision.
Maintain a paid social testing log
Every test should create reusable organisational knowledge.
Record:
Hypothesis.
Dates.
Platform.
Market.
Audience.
Creative.
Offer.
Landing page.
Budget.
Optimisation event.
Primary metric.
Quality metrics.
Result.
Decision.
Context.
Limitations.
Context may include:
Major fixtures.
Paydays.
CRM activity.
Landing-page changes.
Tracking changes.
Offer changes.
Platform issues.
Without this context, a future team may misinterpret why the result occurred.
Record losing tests as well as winners
Negative results are valuable.
They can show:
A proposition does not improve quality.
A creative format attracts weak traffic.
A narrow audience cannot scale.
A promotion increases bonus dependency.
A landing-page variation reduces verification.
Recording negative outcomes prevents teams from repeatedly paying to discover the same thing.
Validate a winner before major scaling
A first test should identify a candidate for scaling.
It should not automatically trigger a large budget increase.
Validate the result in:
A fresh time period.
A broader audience.
A comparable sporting window.
A larger budget.
Another placement.
Another market where appropriate.
This helps determine whether the original result was durable or contextual.
Turn the winner into the new control
Once a variation has been validated, it becomes the benchmark.
The next test should compare a new challenger against the best current approach.
For example:
Test 1:
Offer-led creative vs product-led creative.
Product-led wins.
Test 2:
Product-led control vs alternative product-led opening hook.
New hook wins.
Test 3:
Winning hook vs shorter video execution.
This creates cumulative learning.
The account gradually improves rather than repeatedly restarting from scratch.
Scale in stages
A sensible scaling process may look like:
Stage 1: Validate
Confirm the result with additional evidence.
Stage 2: Increase budget gradually
Test whether performance survives higher spend.
Stage 3: Expand reach
Move into broader eligible audiences.
Stage 4: Test transferability
Assess whether the idea works across:
Markets.
Formats.
Placements.
Products.
Stage 5: Monitor player quality
Confirm that increasing scale does not weaken:
FTD quality.
Retention.
Bonus efficiency.
Net value.
A campaign that works at £500 per day may not behave the same way at £5,000 per day.
Expect quality to change as scale increases
As budgets expand:
Reach increases.
Frequency changes.
Lower-propensity users may enter.
CPM may change.
Creative fatigue can increase.
Conversion quality may weaken.
Scaling should therefore be monitored through marginal economics.
Ask:
Is the next £1,000 producing comparable player quality?
Is CPA rising?
Is FTD conversion falling?
Is D7 value weakening?
Is the audience becoming saturated?
The goal is not maximum spend.
It is the highest sustainable level of profitable acquisition.
Separate scaling from creative fatigue
A winning asset will not perform forever.
Monitor:
Frequency.
Reach.
Days live.
Spend.
CTR trend.
CPA trend.
FTD conversion.
Player quality.
If performance declines, determine whether the cause is:
Creative fatigue.
Audience saturation.
Market change.
Competitor activity.
Landing-page issue.
Offer weakness.
Do not discard a successful proposition merely because one execution has become stale.
Maintain challengers while scaling
A mature paid social account should not allocate 100% of spend to current winners.
Keep a controlled proportion available for:
New creative.
New hooks.
New propositions.
Emerging audiences.
New landing-page ideas.
This prevents the programme from becoming dependent on one creative or campaign structure.
The exact allocation depends on:
Budget.
Market maturity.
Creative production capacity.
Performance stability.
Connect paid social tests with player value
Paid social testing should not exist separately from internal reporting.
Each test should ultimately connect with:
Registration.
Verification.
First deposit.
Repeat deposit.
Bonus cost.
Net revenue.
Retention.
Player value.
This allows the team to answer:
Which creative produces the best players?
Which audiences retain?
Which offers create bonus dependency?
Which landing pages improve qualified conversion?
Which campaign structures create sustainable value?
Without downstream data, paid social optimisation remains incomplete.
Connect paid social testing with landing pages
A creative test can be invalidated by a poor destination.
Check:
Message match.
Offer consistency.
Market.
Product.
Mobile usability.
Registration.
Verification.
Payment.
If one creative sends users to a different page, the test may actually compare complete journeys rather than creative alone.
That can be useful if intentional.
It should be documented as a journey test rather than interpreted as a pure creative test.
Connect paid social testing with CRM
Creative insight can also help CRM teams understand player motivations.
For example:
If product-led football creative consistently produces stronger players, that may inform:
Welcome messaging.
First-bet journeys.
Event-led communication.
Product education.
If bonus-led casino creative produces weaker repeat behaviour, CRM and acquisition teams can review whether the proposition attracts the wrong cohort.
Testing should create insight beyond the platform.
Use automation to improve paid social testing
Automation can reduce the manual work involved in managing experiments.
Useful applications include:
Test naming.
Asset tagging.
Reporting.
Data reconciliation.
Pacing alerts.
Conversion anomaly detection.
Creative-fatigue monitoring.
Quality guardrails.
Testing logs.
Result summaries.
AI can support:
Drafting hypotheses.
Summarising test outcomes.
Identifying unusual performance.
Grouping creative themes.
Preparing test backlogs.
Comparing player-quality metrics.
Automation should not independently decide:
Which promotional claim is compliant.
Whether a player segment should be targeted.
Whether a campaign should be scaled significantly.
Whether a test result is commercially transferable.
Whether player-protection controls can be overridden.
These decisions require specialist human judgement.
Common paid social testing mistakes
Common mistakes include:
Testing without a hypothesis.
Asking several questions in one campaign.
Changing audience, creative and offer together.
Declaring winners on CTR.
Optimising only towards registrations.
Ignoring player quality.
Splitting limited budgets across too many ad sets.
Running overlapping campaigns.
Moving budgets constantly.
Using proxy conversions without validating them.
Ending tests after one strong day.
Using fixed test durations regardless of volume.
Failing to maintain a control.
Scaling before validation.
Treating one-market performance as universal.
Ignoring market and platform restrictions.
Forgetting sporting context.
Recording winners but not failures.
Failing to monitor player quality after scaling.
The stronger approach is a smaller number of controlled questions with clear decision rules.
Practical paid social testing process
1. Find the funnel constraint
Identify the commercial problem before deciding what to test.
2. Write the hypothesis
Specify the variable, audience and expected result.
3. Choose the primary variable
Decide whether the test is about:
Creative.
Audience.
Proposition.
Conversion path.
Delivery.
4. Hold other variables stable
Control unnecessary differences.
5. Choose the deepest reliable metric
Use the most commercially meaningful conversion event available at sufficient volume.
6. Add player-quality guardrails
Protect the result with FTD, retention, bonus and value metrics.
7. Define the decision rules
Set minimum spend, volume, thresholds and stopping rules before launch.
8. Protect the experiment
Avoid unnecessary campaign changes during the test.
9. Record context and results
Preserve the learning whether the test wins or loses.
10. Validate before scaling
Confirm the result before committing substantial budget.
Where Cognaix fits
This is where Cognaix’s role sits: helping iGaming acquisition teams turn paid social testing into a more structured and measurable operating process.
The value is not simply launching more advert variations.
It is helping teams:
Define commercial hypotheses.
Structure tests around controlled variables.
Connect paid social with player-quality data.
Standardise reporting.
Build testing logs.
Monitor creative fatigue.
Identify performance anomalies.
Reduce manual analysis.
Preserve learning across markets.
Turn winning experiments into controlled scaling plans.
For operators and affiliates, the objective should be a paid social programme that becomes more intelligent with each experiment rather than a collection of isolated campaign results.
Final thoughts
The strongest paid social programmes are not built around one breakthrough advert.
They are built around a queue of commercially useful questions.
Each test should tell the business something specific about:
Audience.
Creative.
Proposition.
Conversion.
Player quality.
The process is straightforward:
Question → hypothesis → controlled test → quality check → decision → validation → scale → next question
That discipline matters because paid social platforms can always produce more data.
The competitive advantage comes from knowing which data actually changes the next decision.
When testing is structured around clear commercial questions and downstream player value, paid social becomes a repeatable acquisition system rather than a sequence of expensive guesses.
FAQ
How should paid social testing be structured?
Start with one commercial question, identify the primary variable, hold other major factors stable and define the success metric and stopping rules before launch.
Should paid social tests change one variable at a time?
Where practical, yes. Isolating one primary variable makes the result easier to interpret. Broader exploratory tests can be useful in new markets but should be followed by controlled validation.
What should iGaming paid social campaigns optimise towards?
Use the deepest commercially meaningful event that can be measured accurately at sufficient volume, such as verified registration, first deposit or qualified FTD.
Is cost per registration a good testing metric?
It can be useful as an early diagnostic metric, but it should not determine winners without checking verification, deposits, retention and player value.
How much budget does a paid social test need?
The budget should be sufficient to generate enough examples of the conversion event being evaluated. Low-frequency events such as FTDs generally require more spend or a longer test than registrations.
How long should a paid social test run?
Duration should depend on spend, conversion volume and delivery stability rather than an arbitrary number of days. The test should cover enough meaningful delivery periods to support the decision.
When should a paid social variation be paused?
Pause rules should be defined before launch using minimum spend, efficiency thresholds, conversion volume and quality metrics.
What is an exploratory paid social test?
An exploratory test compares broader combinations of creative, audience or proposition to identify promising directions. It is useful early in a market but does not isolate the cause of performance as cleanly as a controlled test.
Should winning ads be scaled immediately?
No. Strong variations should normally be validated in a fresh period, larger audience or higher-spend environment before substantial budget is committed.
How should creative fatigue be handled?
Track frequency, reach, spend and downstream performance. Refresh the execution where necessary without automatically abandoning a proposition that has already demonstrated value.
Can AI automate paid social testing?
AI can support test planning, tagging, anomaly detection, reporting and summaries. Human specialists should retain control over compliance, commercial interpretation and major scaling decisions.
What is the biggest paid social testing mistake?
The biggest mistake is changing too many variables at once and then treating the resulting performance difference as clear evidence about one of them.