Banner Blindness at $18K: When a CEX's Display Ads Became Invisible to Crypto Traders

The Situation

A mid-tier centralized exchange (CEX) targeting Southeast Asian traders allocated $18,000 over 8 weeks to Google Display Network campaigns. The strategy was straightforward: banner ads on crypto news sites and finance blogs, promoting a "0% trading fee" offer for new users.

The creative team produced four static banners (300x250, 728x90, 160x600, 320x50) with identical messaging and a dark purple brand color scheme. These were the only creatives used for the entire campaign duration.

What Went Wrong

The Google Display Network serves billions of impressions daily, but the crypto-interested audience in Southeast Asia is finite. With only four static banners and no frequency capping, the decay was predictable:

Week 1-2: CTR 0.8%, CPA $28 — Strong start. Crypto news site placements performed well. Week 3-4: CTR 0.4%, CPA $45 — Early signs of fatigue. The same users were seeing the same purple banners repeatedly. Week 5-6: CTR 0.12%, CPA $71 — The algorithm expanded to non-crypto placements (general finance, news aggregators) to find fresh users. Week 7-8: CTR 0.04%, CPA $94 — Complete banner blindness. Average frequency among converting audiences hit 12.3x.

The final two weeks were particularly wasteful: Google's algorithm, struggling to find users who had not already been exposed multiple times, began delivering ads on completely irrelevant placements — cooking blogs, weather sites, mobile game apps. These placements had negligible CPMs but generated zero conversions.

A placement report for the final week showed:

• 34% of impressions on "weather.com" variants
• 18% on mobile game interstitials
• 12% on recipe blogs
• Only 8% on crypto/finance-related sites

Diagnosis

The diagnosis was straightforward:

1. Zero creative variation — Four sizes of the same design is not creative variation. Users saw the identical message in every context, creating what psychologists call "perceptual habituation" — the brain literally stops processing familiar stimuli.
2. No responsive display ads — Google's Responsive Display Ads (RDA) automatically combine headlines, descriptions, and images into hundreds of variations. The team used only static uploads, eliminating Google's ability to test and optimize creative combinations.
3. No frequency management — Without a frequency cap, high-intent users (who visit crypto sites daily) were hammered with 12+ impressions before their next action.
4. No placement curation — The campaign used open targeting with no managed placement list or topic exclusions.

The Fix

We rebuilt the display creative strategy from the ground up:

1. Responsive Display Ads: Created RDA campaigns with:

   • 15 headline variants (mixing benefit-led, curiosity-led, and urgency-led angles)
   • 5 long headlines
   • 5 description variants
   • 8 image assets (4 landscape, 4 square) with different visual styles
   • Total permutations: 3,000+ unique ad combinations

2. Weekly creative refresh: Replaced the bottom 3 performing headlines and 2 images every Monday based on the previous week's CTR data.

3. Frequency cap: Set at 5 impressions per user per 7-day window. This forced Google to find new users rather than over-serving existing ones.

4. Managed placement list: Built a curated list of 120 crypto and fintech websites. Allocated 70% of budget to managed placements and 30% to open targeting for discovery.

5. Creative A/B structure: Ran two ad groups simultaneously — one with "0% fee" messaging and one with "Earn up to 12% APY" messaging — to prevent single-message fatigue.

Results

The recovery was fast — within 10 days:

• CTR recovered from 0.04% to 0.92% (23x improvement)
• CPA dropped from $94 to $39 (58% reduction)
• ROAS improved from 0.6 to 2.8
• Frequency stabilized at 3.8x (down from 12.3x)
• Crypto-relevant placement share increased from 8% to 73%

The $18,000 loss was modest compared to some cases, but the lesson applies at any budget level: on Google Display, creative is not a set-it-and-forget-it asset. It is a depreciating resource that requires constant reinvestment.

Early Warning Signals: Display Fatigue Decays on a Schedule

Display fatigue is more predictable than almost any other failure type — the decay curve in this case (0.8% → 0.4% → 0.12% → 0.04% CTR across eight weeks) is the typical pattern we see whenever static banners run against a finite niche audience without rotation. The signals to watch:

1. Week-over-week CTR halving. One soft week is noise. CTR that halves twice in a row on stable spend is habituation — the audience's brains have literally stopped registering your banners.
2. Frequency above 6x on Display. Niche audiences (crypto readers, finance site visitors) are small; without a cap, frequency compounds fast. Past 6–8x weekly, incremental impressions are close to worthless.
3. Placement drift into irrelevant inventory. When the algorithm exhausts fresh users on relevant sites, it maintains delivery volume by expanding to whatever inventory is cheap — weather sites, game apps, recipe blogs. A falling share of category-relevant placements is fatigue expressing itself as placement decay.
4. Static-only asset library. If every live asset is a fixed-size upload of the same design, you have zero variation for the system to rotate. Fatigue is not a risk in this setup; it is a countdown.

The Recovery Playbook

1. Read the placement report before blaming the creative. Quantify what share of impressions still lands on category-relevant sites. This tells you how much of the decay is habituation versus inventory drift.
2. Switch to Responsive Display Ads. Multiple headlines, descriptions, and images give Google thousands of permutations to rotate — structural resistance to habituation that static banners cannot provide.
3. Impose a frequency cap immediately. Five impressions per user per week is a sane default for niche B2C audiences; it forces the system to find new users instead of hammering existing ones.
4. Curate placements. A managed list of category-relevant sites for the majority of budget, with a capped open-targeting slice for discovery.
5. Establish a refresh cadence. Replace the weakest headlines and images on a fixed weekly schedule, using the asset report — not when someone remembers.
6. Run two message angles in parallel so no single value proposition saturates the audience alone.

Prevention Checklist

• [ ] CTR trend reviewed weekly against your own trailing baseline
• [ ] Frequency cap set on every Display campaign at launch — never left unlimited
• [ ] RDAs used as the default format; static banners only as supplements
• [ ] Placement report checked weekly for relevance drift
• [ ] Fixed weekly asset refresh cycle with documented kill rules
• [ ] At least two concurrent message angles at all times

Where Your Numbers Should Be

Median crypto performance on Google runs around 2.5% CTR (search-weighted), a $42 sign-up CPA, and 3.9x ROAS — Display alone will land below those blends, but a Display CTR collapsing to hundredths of a percent means the channel has stopped functioning, whatever the vertical. Benchmark your account on the Crypto Google Ads benchmarks page, or test your rotation discipline with the creative fatigue detector.

Related reading: Meta Ads Creative Testing Framework (the testing logic transfers directly to Display) · Crypto Exchange User Acquisition Strategies · Same failure type elsewhere: Creative Fatigue: Meta E-commerce