Why VWO, Hotjar, and Crazy Egg Won't Fix Your Conversion Problem If Your GA4 Is Broken
VWO, Hotjar, and Crazy Egg are all useful tools. This post is not arguing otherwise.
What it is arguing is that all three of them operate at the behavioural and experimentation layer they show you what users do on specific pages, where they click, where they leave, and which variant converts better. None of them go a layer deeper to ask whether the analytics directing their use is accurate.
That gap matters more than most D2C brands realise. And it becomes expensive when GA4 is broken.
What These Tools Actually Do
Before getting into where they fail, a quick, accurate summary of what each tool is for:
VWO is a full-stack optimisation platform covering A/B testing, multivariate testing, heatmaps, session recordings, funnel analysis, and personalisation. Following its 2026 merger with AB Tasty under Everstone Capital, it's now one of the most comprehensive mid-market CRO platforms available, starting around $198 per month. For teams running structured experimentation programmes, VWO handles both the hypothesis testing and the behavioural diagnostic in one place.
Hotjar (now part of Contentsquare, which completed its acquisition in 2025) is a behavioural analytics platform centred on heatmaps, session recordings, on-site surveys, and feedback widgets. Its strength is qualitative depth, understanding the why behind what users do. It doesn't include native A/B testing; for experimentation, teams typically connect Hotjar with a dedicated testing platform.
Crazy Egg is a page-level optimisation tool combining heatmaps, scroll maps, confetti click reports, and native A/B testing in one relatively affordable package. It's the most accessible entry point for teams that want to observe behaviour and run simple tests without a steep learning curve or a large budget.
Together, these three tools cover the behavioural layer of a CRO programme. What they have in common is the thing they can't do: none of them validates the GA4 data informing where you point them.
The Problem: These Tools Inherit Your GA4 Misdirection
Here is how a typical D2C brand uses these tools. They open GA4, look at their funnel, see a significant drop-off at a specific step, and then direct their CRO tools at that step. They open Hotjar recordings filtered to sessions that dropped off at that page. They set up a VWO test to improve that element. They use Crazy Egg's heatmaps to see where users are clicking before leaving.
All of this is reasonable provided the GA4 funnel is showing you a real drop-off rather than a tracking gap. If GA4 is showing you a fake drop-off caused by a missing event, you have just directed your entire CRO tool stack at the wrong problem.
Scenario 1: You focus on the product page when the real problem is at checkout
Your GA4 Funnel Exploration shows a 70% drop-off between product page and add-to-cart. You load Hotjar and spend three weeks watching session recordings on your product page; looking for hesitation, rage clicks, exits near the pricing area. You run a Crazy Egg scroll map to understand how far users get before leaving.
What you didn't check: whether add_to_cart is actually firing consistently. If the event is missing for a portion of users, a common failure after Shopify theme updates, users who add to cart are not being counted at that step. The apparent product page drop-off is a tracking gap. Your real add-to-cart rate might be 40% better than GA4 shows.
No amount of session recording analysis fixes a missing add_to_cart event. Hotjar will show you exactly what users do before they apparently leave but if most of them didn't actually leave, you're watching the wrong sessions and drawing the wrong conclusions.
Scenario 2: VWO declares a test winner that doesn't exist
You set up a VWO A/B test on your product page CTA. VWO is measuring success by purchase conversion pulling that signal from GA4. After three weeks, Variant B shows a 22% lift over control at 95% confidence. You ship Variant B site-wide.
Revenue doesn't move.
The reason: your GA4 purchase event has a duplicate firing issue, an order confirmation page widget triggers it twice for approximately 30% of orders. VWO's test is measuring conversion rate against an inflated denominator. Both control and variant look better than they are, but Variant B happened to trigger the duplicate more reliably during the test window it kept confirmation pages open slightly longer. The "winner" is a tracking artefact, not a real user preference.
We covered the real-world version of this in our post on what happens when A/B testing runs on broken GA4 data. The mechanics are the same regardless of which testing platform you're using: VWO, Convert, Optimizely, or Crazy Egg's native A/B tool. If the conversion event the test is measuring against is broken, the test result is broken. The testing platform has no way to know this.
Scenario 3: Hotjar directs you to a page that isn't your actual biggest leak
GA4 shows a 55% drop-off between the cart and checkout. You filter Hotjar session recordings to users who reached the cart page and didn't proceed. You watch 50 sessions. You identify several users who appear to hesitate on the shipping cost display before leaving. Hypothesis: unexpected shipping costs are causing checkout abandonment. You run a test revealing shipping costs earlier.
The test produces no significant result.
The actual issue: begin_checkout isn't firing on mobile which makes up 65% of your traffic. The vast majority of mobile users who add to cart and proceed to checkout are invisible in your funnel data. The 55% drop-off you saw in GA4 is the gap between the users who added to cart and the (undercounted) users GA4 could see initiating checkout. The sessions Hotjar showed you weren't representative of your real checkout audience you were watching the small minority that happened to trigger the event correctly.
No session recording tool can show you the session that GA4 didn't record. The gap is invisible to Hotjar unless you already know the event is missing at which point you don't need Hotjar to tell you where the problem is.
What None of These Tools Can Tell You
All three tools share the same structural limitation: they show you what happens on the page, not whether the analytics that directed you to the page is accurate.
Specifically, none of them:
Detects whether a GA4 ecommerce event is firing consistently or only some of the time
Identifies duplicate events that inflate your reported conversion rate
Reconciles GA4 revenue against Shopify order data to confirm accuracy
Tells you whether the funnel step you're focused on reflects real user drop-off or missing instrumentation
Validates that test metrics pulled from GA4 correspond to actual order activity
These are GA4 implementation questions and they sit one layer below everything VWO, Hotjar, and Crazy Egg can see. The tools are working exactly as designed. The problem is that they're being deployed against a data layer that hasn't been validated.
The Diagnostic Layer These Tools Skip
The reason this matters specifically for D2C brands on Shopify is that GA4 implementation quality varies enormously and the most common failure modes are the least visible ones.
A missing begin_checkout event doesn't produce an error. A duplicate purchase event doesn't show as a warning in GA4's interface. Broken UTM attribution doesn't flag itself in your acquisition report. All of these produce data that looks plausible numbers in the right ranges, trends that roughly make sense until you cross-reference them against Shopify or run DebugView validation.
Our guide to knowing whether a GA4 funnel drop-off is real or a tracking gap covers the diagnostic process in detail. The short version: before directing any CRO tool at a funnel step, cross-reference that step against an independent data source. If the numbers reconcile, the drop-off is worth investigating. If they don't, fix the tracking first.
The 7 signs your GA4 is hiding a funnel leak gives you the specific checks to run before you set up your first Hotjar filter or launch your first VWO test.
The Right Role for These Tools
To be direct about what this post is and isn't arguing: VWO, Hotjar, and Crazy Egg are valuable when used correctly. Hotjar session recordings pointed at a confirmed real drop-off step produce genuinely useful qualitative insight. VWO A/B tests run against a validated conversion event produce trustworthy results. Crazy Egg's heatmaps on a confirmed high-friction page reveal specific interaction patterns worth testing.
The issue is not the tools. It's the sequence in which they're deployed.
The sequence that works:
Validate GA4 ecommerce tracking, confirm events fire correctly, parameters are complete, revenue reconciles with Shopify
Identify real funnel drop-offs, confirmed by cross-referencing GA4 funnel data against independent Shopify data
Direct Hotjar, Crazy Egg, or VWO at those confirmed drop-off points to understand why users are leaving
Form hypotheses and run tests with conversion events you've validated are reliable
The sequence that wastes budget:
Install Hotjar
Look at GA4, see a drop-off
Watch sessions on the page GA4 pointed to
Form a hypothesis based on what you see
Run a VWO test
Wonder why the winner didn't hold in production
The difference between these two sequences is whether the analytics layer underneath the tools has been validated. That validation is not something VWO, Hotjar, or Crazy Egg can provide. It's what a GA4 implementation audit is built to answer before the CRO tool spend begins.
As we covered in detail in what a CRO audit should actually include for a D2C brand, the CRO programme built on top of a validated analytics layer is the one that produces results that hold. The CRO programme built on top of unvalidated analytics is the one that produces interesting session recordings, inconclusive test results, and a lingering suspicion that CRO doesn't work.
It works. But only on reliable data.
Using Hotjar, VWO, or Crazy Egg on your Shopify store and not confident your GA4 data is reliable enough to direct them correctly? A GA4 implementation audit from FunnelFreaks tells you what your analytics is actually showing — and whether you can trust it.