It's Monday morning. You open your dashboard and see that last week's revenue came in 18% below forecast. Your stomach drops. You start clicking between tabs - Google Ads, GA4, Shopify, Stripe - looking for the culprit. Everything looks... mostly fine? A little off here, a little different there, but nothing that screams "here's your problem."

This is the most frustrating experience in growth: a meaningful revenue drop that no single tool explains.

What follows is a step-by-step walkthrough of how to trace a revenue problem through every stage of your funnel. We'll follow a realistic scenario through five tools, show what each tool reveals on its own versus what it hides, and demonstrate why the connections between tools are where the real answers live.

The Setup

Let's say you run a DTC subscription brand. Your typical funnel looks like this:

Google Ads & Facebook Ads → drive traffic to your site. Google Analytics (GA4) & PostHog → track what visitors do on-site. Mida → runs A/B tests on landing pages and checkout. Shopify → handles your storefront and cart. Sticky.io → manages subscription billing and rebills. NMI → processes payments. HubSpot → manages customer lifecycle.

Last week's revenue was $127K. This week it came in at $104K. That's a $23K gap. Where did it go?

Step 1: Check the Top of the Funnel (Google Ads + Facebook Ads)

Your first instinct is right - start at the top. Open Google Ads and Facebook Ads Manager.

What Google Ads shows: Spend is flat. Impressions are flat. Click-through rate actually improved slightly, from 3.2% to 3.4%. Cost per click dropped from $1.82 to $1.71. Google Ads says last week was a good week.

What Facebook Ads shows: Spend is flat. Your top-performing campaign - a lookalike audience based on past purchasers - is performing normally. CPA is $14.50, right in line with your target.

Initial conclusion: Traffic sources look healthy. The problem must be downstream.

What you're missing: Google Ads and Facebook Ads can't tell you what happened to those clicks after they arrived on your site. They report "conversions" based on a pixel that fires on a thank-you page, but they don't know which of those conversions actually resulted in a successful payment, and they definitely don't know what happened at the subscription rebill level.

Keep this in mind. We'll come back to it.

Step 2: Check the Middle of the Funnel (GA4 + PostHog + Mida)

Next, look at what visitors did on your site.

What GA4 shows: Total sessions are up 4%. Your key conversion event - "begin_checkout" - is up 6%. Bounce rate improved. GA4 says last week was also a good week.

What PostHog shows: Funnel completion from landing page to checkout submission is consistent at 8.3%. Session recordings look normal. No obvious UX issues.

What Mida shows: You launched a new A/B test on your primary landing page last Tuesday - "Hero Variant B" - which replaced your headline and hero image. Variant B is winning. Checkout starts are up 12% for Variant B compared to the control.

Initial conclusion: On-site behavior is healthy. The A/B test is performing well. The problem must be even further downstream.

What you're missing: Mida is measuring the A/B test winner by checkout starts, not checkout completions or successful payments. GA4 confirms more people are beginning checkout, but neither tool knows how many of those checkout attempts actually resulted in money in your bank account. More on this shortly.

Step 3: Check the Payment Layer (NMI + Shopify)

Now we're getting into territory that most growth teams don't look at closely enough.

What Shopify shows: Order count is down 11%. Average order value is stable. Shopify flags the order drop but can't explain why - from its perspective, fewer people completed checkout.

What NMI shows: Here's where things get interesting. Your overall approval rate dropped from 91.3% to 84.7% last week. That's a 6.6 percentage point drop. For your transaction volume, that translates to roughly 340 additional declined transactions.

Now you're getting somewhere. But NMI's dashboard shows the decline at an aggregate level. The decline codes are a mix: some "do_not_honor," some "insufficient_funds," some "card_declined." Nothing jumps out as a single cause.

What you're missing: NMI doesn't know where these transactions came from. It doesn't know which ad audience drove the customer, which landing page they saw, or which A/B test variant they experienced. It just sees a card number and a transaction amount.

Step 4: Connect the Dots (This Is Where Revenue Intelligence Lives)

Here's where things get powerful - and where most teams get stuck, because no single tool can do what comes next.

Let's connect the signals across all five tools.

Connection 1: A/B test variant → Payment approval rate.

When you segment NMI's approval data by the Mida A/B test variant the customer saw (which requires connecting PostHog session data to NMI transaction data via Shopify order IDs), a pattern emerges:

• Control group (Hero Variant A): 92.1% approval rate

• Test group (Hero Variant B): 81.4% approval rate

Variant B has a 10.7 percentage point lower approval rate. But Mida declared Variant B the "winner" because it drives more checkout starts. This is the core of the problem.

Connection 2: Why does Variant B have lower approval rates?

Digging into PostHog session recordings for Variant B checkout sessions, you notice something: Variant B's new hero copy emphasizes "Try risk-free - cancel anytime" more prominently. This messaging is attracting a different customer profile - more cautious, more price-sensitive buyers.

Cross-referencing with NMI's BIN data, Variant B's customers have a significantly higher rate of prepaid debit cards (23% vs. 9% for Variant A). Prepaid cards have notoriously lower approval rates because they often have insufficient balances and issuers apply stricter fraud screening.

Connection 3: The Facebook audience compound effect.

Remember that Facebook lookalike audience that was "performing normally"? When you segment it by A/B test variant, the story changes. Facebook's lookalike audience sends visitors who are especially responsive to the "risk-free" messaging in Variant B. Of the Variant B sessions, 61% came from the Facebook lookalike. And within that subset, the prepaid card rate is 31%.

So: Facebook is sending a specific audience → that audience resonates with Variant B's messaging → those visitors start checkout at a higher rate → but their card mix causes significantly more payment declines.

The Facebook campaign looks great in isolation. The A/B test looks great in isolation. The payment decline rate looks "elevated but mixed" in isolation. Only when you connect all three do you see what's really happening.

Connection 4: What about the Sticky.io rebill impact?

There's a second, quieter revenue leak. The customers from Variant B who do get approved on their first purchase have a significantly worse rebill success rate in Sticky.io. First rebill approval rate for Variant B customers is 71% compared to 86% for Variant A. Many of these customers' prepaid cards don't have funds loaded when the rebill hits.

This means the A/B test isn't just costing you revenue now - it's reducing your future subscription revenue too.

The Full Picture

Here's the revenue waterfall, traced through every tool:

Google Ads + Facebook Ads - Traffic volume is fine. But audience composition shifted toward price-sensitive segments that engage with "risk-free" messaging. Ad platforms don't flag this because their metrics look healthy.

Mida A/B Test - Variant B "wins" on checkout starts (+12%). But it attracts a customer profile with fundamentally different payment behavior. Mida doesn't flag this because it measures conversion events, not payment outcomes.

GA4 + PostHog - On-site behavior improved. More people are entering checkout. Analytics tools don't flag this because engagement metrics are up.

NMI - Approval rate dropped 6.6 points. But the aggregate number doesn't reveal that the drop is concentrated in a specific customer segment driven by a specific ad audience through a specific A/B test variant. The payment processor doesn't flag the root cause because it can't see upstream.

Sticky.io - Rebill rates are quietly declining for the new customer cohort. The subscription platform doesn't connect this to the acquisition source or checkout experience.

The actual problem: A well-intentioned A/B test, combined with a Facebook audience's natural response to new messaging, created a customer acquisition pipeline that looks great on every surface metric but produces 10% fewer successful payments and 15% worse subscription retention.

Estimated annual cost if left unchecked: $180K+ in declined first payments, failed rebills, and wasted ad spend on customers who will never be profitable.

The Fix

Once you can see the full picture, the fixes are surprisingly straightforward:

Immediate (this week): Pause Mida Variant B or modify it. The "risk-free, cancel anytime" messaging is effective at driving checkout starts, but it's attracting the wrong payment profile. Test a variant that emphasizes value and product quality instead.

Short-term (this month): Adjust the Facebook lookalike audience. Either exclude segments with high prepaid card usage (you can approximate this through demographic and interest targeting) or create separate landing pages for this audience with messaging that better qualifies buyers before they reach checkout.

Medium-term (this quarter): Implement smart retry logic in Sticky.io for soft declines on first rebills. For the prepaid card cohort, consider a shorter trial period or different billing approach that's more likely to succeed.

Measurement change (permanent): Stop measuring A/B test winners by checkout starts. Start measuring by successful payment rate and projected 90-day LTV. This single change prevents this entire class of problems from recurring.

Why This Requires AI (Not Another Dashboard)

Could you theoretically build a custom dashboard that shows all these connections? Yes, with a team of data engineers and several months of pipeline work. But that dashboard would only answer the specific questions you thought to ask when you built it.

The next revenue drop won't follow this exact pattern. Maybe it'll be a processor outage that only affects one card brand, combined with a seasonal traffic shift, combined with a HubSpot email campaign that reactivated a segment with expired cards. The number of possible cross-tool failure modes is essentially infinite.

What you need isn't a static dashboard. It's an intelligence layer that can dynamically trace signals across your entire stack, in response to whatever question you have, and proactively alert you when it detects patterns you haven't asked about yet.

That's what an AI agent does. It doesn't replace any of your tools. It reads across all of them, understands the relationships between them, and gives you the complete picture that no individual tool - and no reasonably-sized data team - can assemble in real time.

Your Revenue Has a Story

Every dollar of revenue - and every dollar of lost revenue - has a story that spans your entire funnel. The ad that brought the customer. The page they landed on. The test variant they saw. The checkout flow they experienced. The payment attempt that succeeded or failed. The subscription that renewed or didn't.

Right now, that story is fragmented across a dozen tools. The pieces are all there. Nobody's reading the whole thing.

It's time to start.

PayRadar.ai connects your ads, analytics, A/B tests, checkout, payments, and subscriptions into one AI agent. Ask any question. Trace any revenue problem. Get answers in seconds.