A fintech client came to us in June 2025. Their customer acquisition cost had climbed from Rs. 11,250 to Rs. 17,800 over 18 months while their market stayed roughly the same size. They’d hired more people. They’d increased creative output. Nothing moved the number.
Twelve weeks later, their CAC was Rs. 6,750. A 62% reduction from peak.
This isn’t a case study designed to make us look clever. It’s a template — because the five interventions we made are the same five interventions we’ve made at 23 other brands in the past two years. The pattern is consistent enough that we can now call it a playbook.
Here’s what we did, in order, with the specific tools and the realistic timelines.
Fix How You’re Measuring CAC
Before any optimisation, audit the number. You cannot improve what you’re measuring wrong.
The fintech client self-reported CAC of Rs. 11,250. When we audited:
Problem 1: Platform double-counting. Google and Meta were each claiming the same customers. Blended platform conversions showed 1,847 new customers in Q2.
CRM verification: 1,376 actual new customers. 34% attribution inflation. CAC was actually Rs. 14,300, not Rs. 11,250.
Problem 2: Re-acquisition mixed into new customer CAC. 19% of “new customer” conversions were returning users who’d previously transacted. Strip those out and new-customer CAC was Rs. 15,900.
Problem 3: Missing onboarding cost. The team tracked media spend only. True CAC includes sales cost (SDR time for B2B) or onboarding cost (for product-led fintechs). Add onboarding: Rs. 17,200 actual CAC versus Rs. 11,250 claimed.
The real CAC was 52.9% higher than what the team believed. Every optimisation decision before this audit was based on a fiction.
How to Audit your CAC:
- Export all conversions from Google Ads and Meta Ads for a 90-day period
- Export the corresponding CRM records for the same period
- Match by email or phone — deduplicate customers who appear in both platform reports
- Calculate: actual new customers (CRM, first transaction) / (total media spend + onboarding cost)
- Compare to your current reported CAC
The gap between what you believe and what’s real is almost always 20-45%. Fix the measurement before touching anything else.
Step 1: Fix the Conversion Signal
The single highest-ROI intervention. And the most boring.
The fintech client’s Smart Bidding was configured to optimise toward “account registration complete.” Registration sounded right — it was the first meaningful action a new user took. Except 41% of registrations never verified their email. And 28% of verified registrations never made a first transaction.
Smart Bidding was getting very good at finding people who register accounts. Not people who become customers.
The fix: We imported CRM-verified “first payment completed” events to Google Ads via the API, with a conversion value of Rs. 6,800 (average first-transaction value). Account
registrations were retained as a micro-conversion at 8% of that value — enough to signal early intent without dominating the optimisation signal.
Implementation steps (HubSpot -> Google Ads):
- Create a HubSpot workflow that triggers on “first payment” deal stage change
- Use the Google Ads Conversion API (HubSpot has a native integration) to fire the conversion event within 24 hours
- Set conversion value to your average first-transaction value
- In Google Ads, mark the old registration event as a “secondary conversion” — it counts for reporting but not bidding
- Allow 2 weeks of learning phase — performance will fluctuate, do not adjust bids during this period
Timeline to see results: 3-4 weeks from implementation (1-2 weeks learning, then results emerge)
Impact in isolation (fintech case): 31.4% CAC reduction with no other changes. This single fix was worth more than all other optimisations combined.
If you use Salesforce instead of HubSpot: the Salesforce-Google Ads integration has native support for opportunity stage-triggered conversion events. Same logic, different UI.
Step 2: Upgrade Audience Quality
After fixing what you’re optimising toward, fix who you’re targeting.
Broad targeting on Meta and Google has gotten better — both platforms now have AI that expands audiences effectively. But broad targeting has a structural limitation: it finds new users similar to everyone who’s converted, including low-value customers. You want lookalikes of your high-value customers specifically.
The Klaviyo -> Meta lookalike approach:
For e-commerce and D2C clients, Klaviyo’s predictive LTV data is the seed for better Meta lookalike audiences. Here’s the setup:
- In Klaviyo, create a segment: “Predicted LTV in top 25% AND purchased within 90 days”
- Export this list as a CSV (typically 500-5,000 users for a mid-sized brand)
- Upload to Meta as a Custom Audience
- Create a 1% Lookalike based on this high-LTV segment
- Run this against your standard broad audience in a controlled split test for 4 weeks
Real outcome — food delivery platform: Switched from broad interest targeting (food delivery, urban lifestyle, 25-40 demographic) to 1% lookalike of 90-day high-LTV customers. Same budget, same creative. CAC dropped 24.1% within six weeks. The platform found users with similar behavioural patterns to their best customers rather than users who fit demographic proxies.
For B2B: use Salesforce or HubSpot to export Closed Won accounts with highest contract values. Upload to LinkedIn Matched Audiences. Build lookalikes from that list rather than job-title-only targeting.
Timeline to see results: 4-6 weeks (lookalike learning takes time, test for at least 4 weeks before judging)
Step 3: Fix Your Landing Page Conversion Rate
Underestimated. Every time.
Here’s the math that makes landing page conversion rate a CAC lever: if your competitor converts 3.8% of traffic to leads and you convert 2.1%, you are paying 81% more CAC for the same number of leads. Not a bidding problem. Not a targeting problem. A conversion problem.
We see landing page conversion rates between 1.4% and 6.7% for similar products in similar markets. The difference between best and worst in that range means a 4x difference in effective CAC from the same traffic.
The three highest-ROI landing page changes we’ve made across client accounts:
- Traffic-source-specific headlines. A generic headline serves no one well. A user from Google search for “best fintech app for SIPs” has different intent than a user from Meta who saw a sponsored post about tax Same product, different entry point. Split your landing page by traffic source using Unbounce or a simple UTM parameter approach, with different hero messaging for each. Average impact: 18-24% conversion rate improvement.
- Social proof proximate to the CTA. Most brands put testimonials in a scrollable section. Put them directly next to the form or button. The moment a user decides to convert is the highest-risk moment for hesitation. Reduce it there specifically. Average impact: 12-19% improvement.
- Form length reduction. If your lead form has more than 3-4 fields, test a shorter version. Every additional field reduces conversion rate by an estimated 4-8% (based on HubSpot’s own form analytics data across 40,000 forms). For fintech: name, email, phone. Collect everything else after sign-up.
Tools:
- VWO (Rs. 12,000-35,000/month) for A/B testing on your existing site — good for established pages with 2,000+ monthly visitors
- Optimizely ($36,000+/year) for enterprise-scale multi-variant testing
- Unbounce ($74-649/month) for building and testing entirely new landing pages without developer dependency
- Hotjar (free to 12,000/month) for heatmaps and session recordings to understand where users drop off before you test
Timeline to see results: 3-6 weeks per test. Run until you hit statistical significance (minimum 500 conversions per variant). Do not call tests early — the most common testing mistake.
Step 4: Creative Quality, Not Creative Quantity
More ads isn’t the answer. Better ads are.
The creative-bidding relationship: AI bidding systems (Smart Bidding, Advantage+) use creative engagement signals as proxy signals for audience quality. High-engagement creative tells the algorithm it found a receptive audience. Low-engagement creative signals the opposite. Better creative doesn’t just improve CTR — it directly improves bidding efficiency.
The principle: one exceptional creative variant outperforms ten mediocre ones. We’ve tested this across 31 accounts. The accounts that maintain 4-6 genuinely differentiated creative concepts consistently outperform those cycling through 15-20 shallow variations.
What “genuinely differentiated” means:
- Different visual format (video vs. static vs. carousel vs. UGC-style)
- Different message angle (offer-led social proof vs. problem/solution vs. authority)
- Different emotional register (aspirational fear-of-missing-out vs. practical value vs. belonging)
Not: same layout, five different background colours.
For creative generation at volume, Pencil ($149-499/month) generates ad variants from your brand assets and creative brief, then scores predicted performance against 25M+ ads in its database. The predictions are directionally correct about 68% of the time in our testing — useful for prioritising which variants to actually test, not for skipping the test entirely.
Testing methodology: minimum 500 conversions per variant before calling a winner. Use a statistical significance calculator (AB Testguide is free). Most teams call tests too early — they see early variation and assume it’s signal when it’s noise.
Timeline to see results: 6-8 weeks (you need the traffic volume to generate 500 conversions per variant at each stage)
Step 5: Find and Eliminate Wasted Spend Through Attribution Accuracy
The last step — and the one most likely to reveal a surprise.
When we run cross-channel attribution analysis for new clients, we almost always find that one or two channels are significantly over-credited in their platform-reported numbers. The clients have been optimising budgets based on those inflated numbers, over-investing in the over-credited channels and under-investing in what’s actually working.
The typical finding: in accounts running both Google and Meta, Meta-attributed conversions run 35-45% higher than business-verified CAC because of Google view-through attribution overlap. Meta claims the conversion (last-click or view-through). Google also claims it (last-click). Your CRM shows one customer. Both platforms show one conversion. You double-count.
Tools that cut through this:
Northbeam (pricing: $3,000-15,000/month depending on spend volume) — best for performance marketing agencies and growth-stage brands running complex multi-channel campaigns. Their statistical model de-duplicates cross-platform attribution and shows you what each channel is actually contributing.
Triple Whale ($249-999/month) — best for Shopify e-commerce. Native integration with Shopify orders means business-verified revenue is the ground truth, not platform-reported conversions.
The manual approach (free, slower): export all conversions from Google and Meta for 90 days. Match against CRM records by email. Calculate actual customer count vs. sum of platform-reported conversions. The ratio tells you your attribution inflation factor.
What happens when you reallocate based on accurate attribution: in the fintech case, we found Meta was over-attributed by 38% and Google was under-attributed by 22% (Meta was claiming conversions that Google had influenced earlier in the journey). Shifting 15% of Meta budget to Google Shopping and branded search increased verified new customers by 11.3% with no increase in total spend.
Benchmark Expectations by Vertical
| Vertical | Realistic CAC Range | Primary CAC Driver |
|---|---|---|
| E-commerce fashion/lifestyle | Rs. 800–2,400 per customer | Creative quality, landing page CVR |
| Fintech (investment/savings app) | Rs. 4,500–12,000 per verified user | Conversion signal accuracy, audience quality |
| B2B SaaS (SMB target) | Rs. 8,000–22,000 per qualified trial | Landing page CVR, lead scoring accuracy |
| B2B SaaS (enterprise) | Rs. 35,000–1,20,000 per SQL | Attribution accuracy, nurture quality |
| EdTech (online courses) | Rs. 900–3,500 per enrolment | Creative format, offer structure |
| Real estate (lead generation) | Rs. 1,200–4,500 per verified lead | Landing page CVR, audience intent match |
If your CAC is significantly above the high end of your vertical range, the problem is almost certainly one of the five steps above. In our experience: step 1 (conversion signal) fixes 40% of cases, step 3 (landing page) fixes another 25%, and steps 2, 4, and 5 account for the remaining 35%.
Start with step 1. It’s the least glamorous and the highest ROI. That’s usually how it works.
