How Broad Match Keywords Drained $30K From a Crypto Google Ads Campaign — And the Exact Match Strategy That Fixed It

Opening Hook

The search term report told the entire story. A crypto exchange spending $30,000 on Google Search ads, targeting what they believed were high-intent trading keywords. When we pulled the actual search queries triggering their ads, the top terms included "is bitcoin a scam," "free cryptocurrency," "what is blockchain technology," and "crypto news today." Not a single one of these queries indicated purchase or trading intent. The advertiser was paying $8.50 per click to educate curious browsers who had zero intention of opening a trading account, let alone depositing funds. Their CPA had ballooned to $420 per verified user — against a target of $80.

The Setup

The client operated a regulated cryptocurrency exchange targeting retail traders in Southeast Asia and Europe. Their platform supported spot trading, futures, and staking across 150+ token pairs. The unit economics were sound: average first deposit was $340, 90-day LTV was $1,200, and organic conversion rate from their landing page sat at a respectable 4.2%.

Their in-house marketing team had launched Google Search campaigns to accelerate growth beyond organic and referral channels. The keyword strategy was built around what seemed like common sense: target terms that contain "crypto," "bitcoin," "trading," and "exchange." They set up three campaigns — one for Bitcoin-specific terms, one for general crypto terms, and one for competitor brand terms. Total daily budget was $700 across all three.

The match type selection was where the foundation cracked. All keywords were set to broad match, following Google's own recommendation to "let the algorithm find the best queries." The team trusted Smart Bidding (Target CPA at $80) to handle the optimization. No negative keyword lists were built. No audience signals were layered. No search term review process was established.

What Went Wrong

For the first two weeks, the numbers looked deceptively acceptable. Google's reporting showed a $95 CPA against the $80 target — close enough that the team felt optimistic. But this number was a mirage. Google was counting "sign-up" as the conversion event, not "KYC-verified depositing user." The actual CPA for the metric that mattered was already $280 and climbing.

By week four, the campaign had spent $18,000. The search term report — which no one had reviewed — revealed the catastrophe. Here is a representative sample of what the broad match keywords were actually matching to:

• "bitcoin" (broad) matched: "bitcoin price prediction 2027," "bitcoin documentary netflix," "bitcoin inventor," "is bitcoin dead"
• "crypto exchange" (broad) matched: "crypto exchange comparison reddit," "best free crypto exchange," "crypto exchange hacked news"
• "buy cryptocurrency" (broad) matched: "should I buy cryptocurrency now," "buy cryptocurrency with no fees," "buy cryptocurrency under $1"

Over 73% of click volume came from informational queries with no transactional intent. The algorithm was finding cheap clicks — which happened to be from people researching, not trading. Smart Bidding dutifully optimized for the cheapest sign-ups, which were overwhelmingly users who registered out of curiosity, never completed KYC, and never deposited a single dollar.

By day 45, $30,000 had been spent. Total verified depositing users: 71. Actual CPA: $420. ROAS: 0.5x.

Root Cause Analysis

The failure stemmed from four compounding errors:

Broad Match in a High-Noise Vertical. Cryptocurrency is one of the noisiest search verticals in existence. The ratio of informational-to-transactional queries is approximately 15:1. Terms like "bitcoin" and "crypto" are overwhelmingly searched by people seeking news, education, or entertainment — not by people ready to open a trading account. Broad match in this vertical is essentially paying to fund a crypto education service.

Wrong Conversion Event. The campaign optimized for sign-ups (email registration), not for KYC-verified deposits. This is the Google Ads equivalent of the Meta case — asking the algorithm to find tire-kickers instead of buyers. Google's Smart Bidding model learned that the cheapest sign-ups came from curiosity-driven informational queries, so it aggressively bid on those.

Zero Negative Keyword Architecture. The account had literally zero negative keywords. No exclusions for "free," "scam," "news," "price prediction," "documentary," "reddit," "what is," or "how does." In a vertical where 73% of query volume is non-commercial, this is the equivalent of leaving the vault door open.

No Search Term Review Cadence. The team operated on a "set and forget" model. No one reviewed the search term report for 45 days. In a broad match campaign, the search term report is the single most important diagnostic tool. Without reviewing it, there was no mechanism to catch the query drift that was hemorrhaging budget.

The Fix

The restructuring was executed over a focused 21-day sprint:

1. Search Term Audit and Negative Keyword Build (Days 1-3). Downloaded the complete 45-day search term report — 14,000+ unique queries. Classified each query by intent: transactional, navigational, informational, or junk. Built a master negative keyword list of 2,200+ terms organized into themed lists: educational terms, news terms, scam/fear terms, free/cheap terms, competitor comparison terms, and irrelevant crypto terms (NFT, mining, etc.). Applied as shared negative keyword lists across all campaigns.

2. Keyword Restructuring (Days 3-7). Replaced all broad match keywords with exact match and phrase match equivalents. The new keyword set focused exclusively on high-intent transactional terms: [buy bitcoin], [open crypto account], [crypto exchange sign up], [trade ethereum], [bitcoin trading platform]. Organized into tightly themed ad groups with no more than 15 keywords per group. Each ad group had tailored ad copy matching the specific intent cluster.

3. Conversion Event Fix (Days 3-5). Replaced sign-up as the primary conversion with "KYC-verified first deposit." Implemented enhanced conversions with first-party data (hashed email) for improved attribution. Set a 30-day conversion window to account for the typical 7-14 day lag between sign-up and first deposit in crypto.

4. Audience Signal Layering (Days 7-10). Added audience signals (observation mode initially, then targeting mode after data accumulated): in-market audiences for "Financial Services" and "Investment Services," custom intent audiences built from competitor URLs and high-intent search terms, and a Customer Match list of 5,000 existing depositors for similar audience expansion.

5. Landing Page Alignment (Days 7-14). Created intent-matched landing pages for each ad group cluster. Transactional queries ("buy bitcoin") landed on a streamlined deposit-and-trade page. Platform queries ("best crypto exchange") landed on a feature comparison page. The generic landing page that previously served all traffic was retired.

6. Bidding Strategy Reset (Days 14-21). Switched from Target CPA (which had been optimizing for junk conversions) to Maximize Conversion Value with a target ROAS of 300%. This required the 14-day data accumulation from the new conversion event before the algorithm had sufficient signal to optimize effectively.

Results

The recovery was swift once the structural issues were corrected:

| Metric | Before | After | Change | |--------|--------|-------|--------| | ROAS | 0.5x | 3.8x | +660% | | CTR | 1.2% | 4.5% | +275% | | CPC | $8.50 | $2.10 | -75% | | CPA (Verified Depositor) | $420 | $65 | -85% | | Avg First Deposit | $120 | $380 | +217% | | Search Term Relevance | 27% | 94% | +248% | | Impression Share (exact) | 12% | 68% | +467% |

The CPC decrease was counterintuitive — exact match keywords in crypto are typically more expensive per click than broad match. But the quality score improvements from tighter ad-to-keyword-to-landing-page relevance dropped the actual CPCs significantly. Google rewards relevance with lower auction prices.

The depositor quality improvement was equally dramatic. Users arriving via high-intent exact match queries deposited 3.2x more on average, because they arrived with pre-formed trading intent rather than idle curiosity.

Key Takeaways

• Broad match in high-noise verticals is budget destruction. In verticals where informational queries outnumber transactional queries by 10:1 or more (crypto, insurance, legal, medical), broad match without extensive negative keyword lists will hemorrhage budget on irrelevant traffic.

• The search term report is not optional — it is the campaign's nervous system. For any campaign using anything other than pure exact match, search term reviews should happen every 48-72 hours for the first 30 days, then weekly thereafter. Automate alerts for new query patterns.

• Optimize for the event that generates revenue, not the event that is easiest to track. Sign-ups, page views, and add-to-carts are vanity conversion events. If your revenue comes from deposits, optimize for deposits. Accept the longer learning phase — it is worth it.

• Negative keywords are as important as positive keywords. In many verticals, your negative keyword list should be 5-10x larger than your positive keyword list. This is especially true in crypto, where the search landscape is polluted with news, speculation, and educational queries.

• Google's recommendations are not your strategy. Google will always recommend broad match and automated everything because it increases their revenue. Your job is to constrain the algorithm to the opportunity space where your economics work, not to give it unlimited freedom to explore.

Prevention Checklist

Before launching any crypto-vertical Google Search campaign:

• [ ] Keyword match types are exact or phrase match only — no broad match without proven negative keyword coverage
• [ ] Negative keyword list built with minimum 500 terms covering: educational, news, scam, free, comparison, irrelevant crypto sub-verticals
• [ ] Primary conversion event is the actual revenue event (deposit, purchase), not a funnel milestone (sign-up, registration)
• [ ] Enhanced conversions configured with first-party data (hashed email/phone)
• [ ] Search term review process scheduled: every 48 hours for first 30 days, then weekly
• [ ] Ad groups contain maximum 15 tightly themed keywords each
• [ ] Landing pages are intent-matched to ad group themes (not one generic page)
• [ ] Audience signals layered: in-market, custom intent, Customer Match
• [ ] Automated rules set to pause keywords with spend > $200 and zero conversions
• [ ] Competitor brand campaigns separated with dedicated budgets and landing pages