The optimization industry's structural problem
Discover Why Traditional CRO Agencies Fail and How to Fix It
The industry model sells retainers before diagnosis, A/B tests on pages without enough traffic to reach significance, and 90-day timelines for problems fixable in a week. This is not a critique of individual agencies. It is a structural analysis of the model.
The alternative is not a different agency. It is a different sequence: diagnose first, with evidence you can verify, before paying anyone anything.
The current completed-audit benchmark averages 6.3/10 and 4.6 recorded findings per page. Every one of those failures is observable, specific, and fixable - without a retainer, without a discovery call, and without running an A/B test that won't reach significance for six months. The audit takes under two minutes. The fix takes a day. The retainer is optional.
Six ways the model fails founders.
These are structural failures, not individual agency failures. The model produces them reliably.
Business model
Retainer before diagnosis
The standard agency model starts with a discovery call and a monthly engagement. The diagnosis comes after you've signed. This means you pay for the agency to learn what's wrong - information that should be the starting point, not the output of month one.
The alternative: The audit comes first. It costs less than an hour of agency time, takes under two minutes, and tells you whether there is a structural problem before any money changes hands.
Evidence standard
Generic recommendations over specific findings
"Improve your CTA" appears in every CRO report, written for every page, without reference to your actual CTA. The finding that matters is: "Your primary CTA is below the fold on a 375px viewport. 64% of your paid traffic is mobile. The visitor never sees it." The first is advice. The second is evidence.
The alternative: Evidence-grade findings cite the source: the actual H1 text, the pixel dimensions, the character count, the specific element. You can verify every finding yourself by looking at the page source.
Statistical methodology
A/B testing on pages with insufficient traffic
Running a split test on a landing page with 300 monthly visitors will take 18 months to reach statistical significance at 95% confidence. Most agencies run these tests anyway. The results are noise presented as insight. The founder pauses ad spend and changes nothing, or changes the wrong thing.
The alternative: Fix structural defects before testing. Structural problems - wrong headline, buried CTA, no proof - are not optimization opportunities. They are defects. Fix them with evidence, not experiments.
Scope
90-day timelines for problems fixable in a week
A meta description that is 197 characters can be fixed in 90 seconds. A CTA that doesn't contrast against its background can be fixed in one CSS line. A headline that doesn't match the ad that sent the traffic can be rewritten in an afternoon. Agencies bundle these fixes into multi-month roadmaps because the engagement model requires it.
The alternative: The highest-priority page fixes are almost always structural and fast. Run the audit, find the highest-priority failing signal, fix it, re-audit. The whole loop runs in days, not quarters.
Attribution
Claiming credit for conversion lift they didn't cause
If conversion rate increases during a CRO engagement, the agency takes credit. If it decreases, the issue is attributed to traffic quality or seasonality. Without a defined experiment protocol and a measurement window agreed before the fix is implemented, before/after comparisons are stories, not evidence.
The alternative: The Nebula approach: run the audit, document the failing signal and its evidence, implement the specific fix, run the audit again after a defined window, compare the signal scores. The before/after is the product, not the retainer.
Cost structure
Paying for strategy when the problem is structural
A $3,000/month CRO retainer is the right investment when: you have sufficient traffic to test, your page passes structural checks, and you need creative and copy iteration. It is the wrong investment when your H1 doesn't match your ad, your CTA is invisible on mobile, and your page has no trust signals above the fold. That's a $97 problem, not a $36,000/year one.
The alternative: Sequence the investment. Structural audit first ($0). Fix the structural defects ($97 per leak). Then consider ongoing optimization once the structural layer is clean.
The alternative
What the evidence standard looks like.
Every finding from a Nebula audit traces to a specific, observable condition: the actual H1 text, the computed contrast ratio, the character count, the pixel dimensions, the presence or absence of a specific HTML element. Not a best practice. Not an opinion. A finding you can verify by looking at the source.
When the finding is specific, the fix is specific. When the fix is specific, the before/after is measurable. When the before/after is measurable, you know whether the fix worked. That loop - audit, fix, re-audit - is the product.
The retainer comes after the loop is proven, not before.
Example finding
Signal: Message Match
FAIL
Evidence:
H1: “The AI SEO agency that makes you the answerin AI O”
(H1 truncated mid-word at 53 chars)
Ad keyword targeted:
“AI SEO agency”
Fix:
Rewrite H1 to under 90 chars. Confirm it contains the exact ad keyword. Verify on mobile viewport.
The right sequence.
01
Audit first
Run an evidence-grade diagnostic on the page. Free, under two minutes. Find the specific failing signals before spending anything.
02
Fix structural defects
Message match, CTA visibility, social proof placement. These are not optimisation opportunities - they are defects. Fix them with the evidence report.
03
Re-audit
Run the audit again after the fix is live. Compare signal scores. Document the before/after. This is the evidence that the fix worked.
04
Then consider testing
Once the structural layer is clean and you have sufficient traffic, A/B testing is meaningful. Not before.
Run the diagnostic before you hire anyone.
If the page fails structural checks, fix those first. The free audit takes under two minutes; the score and initial findings appear before email.
Find the Leak →Frequently asked questions
Do CRO agencies actually improve conversion rates?
Some do, over long engagements with sufficient traffic. But the industry model - retainer before diagnosis, generic recommendations, 90-day timelines - means most founders pay for months before any measurable change is implemented. For pages receiving under 5,000 visits per month, the statistical power for A/B testing doesn't exist anyway. The right first step is a diagnostic audit, not an engagement.
What's the difference between a CRO audit and a CRO retainer?
An audit diagnoses what is observable and measurable on your page right now - specific, verifiable defects with evidence attached. A retainer is an ongoing engagement that usually bundles diagnosis, strategy, creative, testing, and reporting into a monthly fee. The audit should come first, cost less, and determine whether the retainer is even necessary.
What should I do before hiring a CRO agency?
Run an evidence-grade diagnostic audit of your landing page. If the page fails observable checks - message mismatch, CTA not visible on mobile, no social proof above the fold - fix those first. They are structural defects, not optimization opportunities. Fix structural issues yourself, then hire for testing and creative if conversion still underperforms.
Why doesn't A/B testing work for most landing pages?
A/B testing requires statistical significance - typically 1,000+ conversions per variant. Most landing pages for founders running paid ads don't have that traffic. Running a split test on a page with 200 monthly visitors will take 18 months to reach significance, and the result will still be uncertain. Fix the structural defects first. Then test when volume exists.
What is an evidence-grade landing page audit?
An audit where every finding is traceable to a specific, observable condition on the page - an H1 that doesn't match the ad, a CTA that is below the fold on 375px, a meta description that is 197 characters. Not an opinion. Not a best practice recommendation. A finding you can verify yourself by looking at the source.