MishaBook a demo

Aug 14, 2026

Stop Guessing on Creative

Creative evaluation framework: a systematic approach to measuring ad asset performance against quantified thresholds (CTR, CPC, ROAS, view-through rate) rather than aesthetic preference or gut feel.

Why Guessing Fails

Most DTC operators evaluate creative through a filter of personal taste or 'what feels right.' This introduces bias and delays scaling decisions. A founder who loves minimalist design may kill a high-performing product demo video. A marketer attached to brand voice may run a lower-converting testimonial ad for weeks.

The cost is compounding: every day a creative underperforms is a day the budget doesn't flow to the winner. Every creative that gets greenlit without threshold checks is a sunk test cost. Over a quarter, this adds up to 15 - 30% of wasted ad spend.

Guessing also breaks team alignment. Without shared metrics, creative briefs become arguments. The designer doesn't know what 'better' means. The paid lead doesn't know when to escalate.

Define Your Baseline Thresholds

Before running any creative test, establish minimum performance gates. These are not targets - they are the floor below which a creative is considered failed and should be paused or killed.

  • CTR (Click-Through Rate): Minimum 0.8 - 1.2% for feed ads, 1.5 - 2.5% for search. Below this, the creative is not resonating at impression level.
  • CPC (Cost Per Click): Should not exceed 30% above your historical average. If your baseline CPC is $0.80 and a creative drives $1.10+, pause it.
  • ROAS (Return on Ad Spend): Minimum 2.0x for profitability on most DTC. Below 1.8x, the creative is not covering CAC + margin.
  • View-Through Rate (for video): Minimum 25% for 3-second views on YouTube/TikTok. Below 20%, the creative loses attention too fast.
  • Frequency cap: Pause creative after 3 - 4 impressions to the same user if ROAS drops >20% week-over-week. Fatigue is measurable.

The Test Window and Sample Size Rule

Operators often kill creative too early or let losers run too long. Use this decision rule: run creative for 3 - 5 days minimum or until you hit 100 - 200 conversions (whichever comes first). This eliminates noise from daily variance.

If a creative hits your minimum thresholds within this window, scale budget by 20 - 30% and re-evaluate after another 100 conversions. If it misses thresholds, kill it and move to the next test.

Exception: If a creative is bleeding money (ROAS < 1.2x after 50 conversions), kill it immediately. Don't wait for the full sample.

Failure Modes to Watch

Creative can fail in predictable ways. Knowing these patterns accelerates diagnosis.

  • High CTR, low conversion rate: The creative attracts clicks but the landing page or offer doesn't convert. Problem is not the creative. Audit the funnel.
  • Low CTR, high CPC: The platform is showing the creative to the wrong audience or at the wrong time. Tighten targeting or adjust bid strategy.
  • High ROAS day 1 - 2, cliff on day 3: Audience saturation or fatigue. Expand audience size or rotate creative.
  • Consistent underperformance across all platforms: The core message or product angle is weak. Go back to the brief.
  • High ROAS but declining incrementally: This is normal. Set a floor (e.g., 2.2x) and pause when it hits 2.0x to preserve margin.

Creative Rotation and Scaling Checklist

Once a creative passes thresholds, follow this sequence to avoid plateau:

  • Passes threshold test (100 - 200 conversions): Increase budget 20 - 30%, monitor for 3 days.
  • Holds ROAS after scale: Increase again by 25 - 50%, run for 5 - 7 days.
  • ROAS remains >2.0x: Move to 'always on' status. Reduce frequency cap to 2 - 3 impressions to extend lifespan.
  • ROAS drops below 2.0x: Pause and rotate in a new creative variant. Keep the winner in a holdout audience (10 - 15% of budget) to measure decay.
  • Holdout ROAS declines >15%: The creative is fatigued. Archive it and test a new angle.

Build a Creative Scorecard

Document every creative test in a single source of truth. This prevents re-testing losers and surfaces patterns.

Minimum fields: Creative ID, platform, audience, start date, end date, impressions, clicks, conversions, CTR, CPC, ROAS, status (pass/fail/paused), notes. Update weekly.

Review the scorecard monthly. Look for: Which message angles (product demo vs. lifestyle vs. testimonial) consistently outperform? Which platforms have the highest ROAS? Which audience segments convert best? Use these insights to brief the next round of creative.

When to Break the Rules

Thresholds are guardrails, not absolutes. Break them in two cases:

Brand awareness campaigns: ROAS targets are lower (1.2 - 1.5x acceptable). Measure success by reach, frequency, and brand lift survey instead.

New product launch: First 50 conversions may miss thresholds due to low intent. Run for 7 - 10 days minimum before killing. Collect feedback and iterate the creative, not the product.

Questions

FAQ

How do I know if my thresholds are too high or too low?

Compare your thresholds to your historical best performers. If your top 20% of creatives average 3.5x ROAS, a 2.0x floor is reasonable. If your average is 1.8x ROAS, raise the floor to 1.6x to avoid killing marginal winners. Adjust quarterly based on market conditions and CAC trends.

Should I test creative on one platform or multiple platforms at once?

Start on one platform (usually Meta for most DTC). Once a creative passes thresholds there, test it on secondary platforms (TikTok, Pinterest, Google). Platform audiences differ; a winner on Meta may underperform on TikTok. Don't assume parity.

What's the difference between pausing and killing a creative?

Pause: ROAS is still >2.0x but declining. Hold it in a small holdout audience (5 - 10% of budget) to measure decay and extend lifespan. Kill: ROAS is <1.8x or has declined below your floor for 3+ days. Archive it and move budget to the next test. Never resurrect a killed creative without a significant brief change.

How many creatives should I test at once?

Run 3 - 5 variants in parallel if budget allows ($500 - $1,000 per variant minimum). This accelerates learning. Once one passes thresholds, scale it and test 2 - 3 new variants against it. Avoid testing >5 at once; you'll dilute budget and extend the test window.

Want this on your account?

Thirty minutes. Bring the number that keeps you up.

Book a demo

More from the blog