Choosing a CRO Agency for B2B Websites: Methods, Questions, Results

Key takeaways A good CRO agency for B2B does not work with button colors. It works with research, hypotheses and clean measurement all the way to the qualified inquiry. Check four things above all: whether you have enough traffic for tests, whether the agency combines heatmaps and tests into one approach, whether it counts results in your CRM instead of just in the form, and whether it avoids promising fixed uplifts. Without enough visitors, SEO or a clearer offer is often the better first step.

• Judge providers by method, not by how good the portfolio looks
• In B2B, a conversion is a qualified inquiry, not every form click
• Tests need volume. A worked example follows further down
• Tracking and heatmaps need proper consent under the TDDDG (German Telecommunications Digital Services Data Protection Act)
• Questions 1 to 9 in the checklist separate professionals from slide sellers

Table of contents

• What CRO does differently in B2B
• Do you need a CRO agency at all?
• Which types of providers exist
• What a clean CRO method looks like
• How much traffic tests need
• Measurement: What counts as a result?
• Tracking, heatmaps and consent
• Technology: Who implements the changes?
• The 9 vetting questions as a checklist
• Contracts, models and market ranges
• The first 90 days
• Red flags
• When you do not need a CRO agency
• How we do it
• FAQ

What CRO does differently in B2B

CRO stands for conversion rate optimization. It means getting more visitors to your website to take the action that matters to you: a demo appointment, an intro call, an inquiry. The conversion rate is simply the ratio of actions to visitors. You decide what counts as an action, and that is exactly where B2B differs sharply from online shops.

In a shop, things are clear: purchase or no purchase, hundreds of transactions per week, and a test delivers a result in days. In B2B, it usually looks like this:

• Low volume. A software company or a law firm often has a few thousand visitors a month and perhaps 20 to 80 inquiries. For many classic tests, that is not enough.
• Long decision paths. According to the 6sense 2025 report, the average B2B buying cycle was around 10 months, and 94% of buying groups had already ranked a preferred provider before first contact (6sense, press release 2025). Your website has to build trust before anyone even sees a form.
• Several people decide. The visitor on the pricing page is rarely the one who signs. Copy and proof have to work for finance, IT and management as well.
• Quality beats rate. A page that doubles its rate with a prize-draw form but only attracts students has won nothing.
• Less contact with sales. According to Gartner, 67% of B2B buyers prefer a purchase without a sales rep (Gartner, March 2026, n=646). The more buyers research on their own beforehand, the more important website content that answers their questions becomes.

It follows that a CRO agency for B2B has to do more than test. It has to understand how buying works in your market and align the website with that path. Anyone who immediately promises you "20% more conversions" without asking these questions has not looked at your situation.

An honest note on numbers: published conversion rate benchmarks vary widely depending on source, industry and measurement method. The landing page tool provider Unbounce gives a median of 3.8% for SaaS landing pages (Unbounce, Conversion Benchmark Report SaaS). The page states neither a clear survey year nor the sample size for SaaS, and it covers individual landing pages, not whole websites. Values like this help you get a feel, but not as a target. Your own baseline matters more than any average.

Do you need a CRO agency at all?

This is the question before you choose an agency. CRO improves what already arrives. If hardly any visitors come, there is nothing to optimize. Check these four points honestly:

1. Traffic. Do you have enough visitors on the pages that matter (home, service, pricing, demo page)? We give a rough rule of thumb further down, with a worked example.
2. Clean tracking. Do you know today how many inquiries come from where? If the numbers are off, the first task is measurement, not optimization.
3. A clear offer. If visitors cannot understand in five seconds what you solve and for whom, no A/B test on button position will help. That is a positioning or messaging problem. Our guide to the SaaS hero that explains software instantly shows what that can look like.
4. Someone who implements changes. An analysis is worthless if the changes sit in a ticket queue for six months.

If point 1 or 2 is missing, the better first step is usually SEO and AI visibility for more visitors, or a tracking setup. That is not a rejection of CRO, just the right order. This is part of good advice too: a provider who sells you a testing retainer at 800 visitors a month is working against you.

Which types of providers exist

Very different providers hide behind the term "CRO agency". The table helps you place who you are dealing with.

Provider type	Strength	Typical weakness	Fits when
CRO specialist agency	Deep methodology, lots of testing experience, statistics	Often does not implement changes on the website itself, dependent on your dev team	You have a lot of traffic and a team that implements changes
Web or design agency with CRO services	Analysis and implementation from one source, design and tech work hand in hand	Testing depth and statistics weaker than at specialists, quality varies widely	You want a website that keeps improving without coordinating three service providers
Performance or ads agency	Knows landing pages for campaigns, fast creative tests	Focus on paid traffic, the website as a whole is often out of view	Most of your leads come from ads
UX agency	Strong in user research, interviews, usability tests	Little quantitative testing, no ongoing optimization operation	You have a fundamental user guidance problem
Independent consultant or freelancer (market alternative)	Flexible, often experienced, cheaper hourly rate	Capacity, vacation and downtime, rarely design and development from one source	You need targeted help and have implementers in-house
Tool vendor with consulting	Platform and analysis in one	Method follows the tool, not your problem	You already have a team and are looking for tools
In-house team	Close to product and sales	Takes time to build, knowledge depends on individuals	You have traffic, budget for several roles and staying power

A combination of SEO and CRO at one provider, which many SaaS teams look for, makes sense if both feed the same goal. The SEO side brings qualified visitors, the CRO side turns them into inquiries. Important: report the two separately. Otherwise you cannot tell later which part delivered what. You will find more on the distinction in our article on SEO for B2B SaaS.

One more note on the platform: some providers only work with certain systems. Ask early whether the agency can work directly in your CMS (for example Webflow, WordPress or TYPO3) or whether every change has to go through your internal dev team. That affects speed more than any method does.

What a clean CRO method looks like

Whatever the agency calls it, a reliable process always has the same building blocks. If one is missing, be careful.

1. Clarify goals and measurement

At the start is the question of what a conversion means for you and how it is measured. In B2B, that usually means two levels:

• Micro-conversions: click on "Request a demo", starting the form, downloading a whitepaper.
• Macro-conversions: submitted inquiry, qualified inquiry (MQL or SQL), booked appointment.

The agency should propose how both become visible together in analytics and in the CRM. The conversion goal is the anchor for everything else.

2. Research: data and users

This is where craft separates from guessing. Good research combines:

• Quantitative data: Where do users drop off? Which pages bring traffic but no inquiries? Which devices and sources convert worse?
• Behavioral data: Heatmaps, scroll depth and session recordings show what visitors really do, for example whether they see the form but do not start it.
• Qualitative sources: Conversations with sales and customers, questions from demo calls, support tickets, interviews. In B2B, this is often the richest source because it shows the real objections.
• Heuristic review: Clarity of the message, trust elements, form length, mobile use, load time.

An agency that writes hypotheses without talking to your sales team is missing part of the answer.

3. Hypotheses instead of ideas

Research produces hypotheses in this form: "Because we observe X, we assume that change Y will improve metric K for audience Z." A hypothesis is reasoned and falsifiable. "Let's try a green button" is not one.

This includes prioritization, for example by expected effect, effort and confidence in the assumption. The agency should be able to explain why test A comes before test B.

4. Implement and test

Not every change needs an A/B test. Obvious errors (a broken form, missing contact details, an incomprehensible headline) you fix directly. You test where you are unsure and have enough visitors. With little traffic there are alternatives, which we show below.

5. Evaluate, learn, document

Every test ends with a decision and a learning note, even if the variant loses. Good agencies keep a test archive that you can take with you when the contract ends. Lost tests are not a defeat, they save future failed attempts.

How much traffic tests need

This is the point where many B2B projects tip over. A test needs a minimum number of visitors per variant so that a difference is not chance. You can do a rough calculation yourself. The table shows the number of visitors needed per variant for a standard test with a 5% significance level and 80% power (our own calculation using the two-sample formula for proportions).

Baseline conversion	Expected relative lift	Visitors per variant	Visitors in total (2 variants)
2%	+10%	about 80,700	about 161,000
2%	+20%	about 21,100	about 42,000
2%	+30%	about 9,800	about 19,600
3%	+20%	about 13,900	about 27,800
3%	+30%	about 6,500	about 13,000
5%	+20%	about 8,200	about 16,300
5%	+30%	about 3,800	about 7,600

How do you read this? If your demo page gets 2,000 visitors a month and 3% convert, you need about 14 months for a test aimed at a 20% effect. That is not practical. Two consequences:

• Only test big levers. Small changes produce small effects that you can never prove at your volume. Bigger interventions (new page structure, different offer, form cut in half) are testable.
• Start where most of the traffic is. Often that is the home page, not the demo page.

There is also the often-cited rule of "100 conversions per variant". According to Growth Method, it is a myth that goes back to old hints in the Google Optimize documentation, which spoke of "at least a few hundred conversions per variant". The recommendation there: calculate the required sample size in advance and then estimate the run time (Growth Method, 100 Conversions Rule).

Equally important: do not peek in the middle and stop as soon as it looks good. Repeated checking and stopping early significantly increases the number of false winners, and a risk reported as 1% can in reality be around 5%. The safe approach is to fix the sample size beforehand and wait for the end (Evan Miller, How Not To Run An A/B Test). Ask the agency how it handles this. A good answer mentions run times calculated in advance, full weekly cycles and clear stopping rules.

What to do when traffic is low

Low traffic does not mean "do nothing". It means: other methods.

• Qualitative analysis first: User interviews, session recordings and conversations with sales deliver insights without a sample size problem.
• Implement best-practice changes directly and compare before and after, knowing that it is not proof. Seasonal effects and campaigns quickly distort a before-and-after comparison.
• Direct more traffic to the important pages before testing, for example through search and AI answers.
• Use micro-conversions as the test goal when they are a reliable precursor of the inquiry (for example a click on the appointment link). Here you have to check whether the link to the real inquiry holds.

Measurement: What counts as a result?

Most bad CRO projects fail not because of tests but because of the yardstick. Three traps:

Trap 1: Form submissions without a quality check. If you remove form fields, you get more inquiries, but often worse ones. So ask whether results are tracked all the way into the CRM: How many of the new inquiries became appointments, how many became proposals?

Trap 2: Multiple touchpoints. In B2B, someone visits the site five times before they inquire. The inquiry then counts for the last channel. If the agency only evaluates the last click, some things look bad that in fact built trust.

Trap 3: Averages that hide differences. Mobile and desktop, new and returning visitors, organic and paid all behave differently. Good analyses segment.

Demand a measurement definition in one sentence up front, for example: "Success is the number of qualified inquiries per 1,000 visitors to the service pages, measured in the CRM." Everything else is decoration.

Tracking, heatmaps and consent

CRO needs data, and under German law data needs a clean legal basis. Under § 25 TDDDG, storing information on the end device or reading it is generally only allowed with consent. Exceptions exist for technically strictly necessary access (§ 25 TDDDG, Gesetze im Internet). Heatmap and recording tools usually use scripts or cookies that can fall into this area. On top of that comes the GDPR (DSGVO) for processing personal data.

What does that mean in practice, without being legal advice (that comes from data protection officers or lawyers)?

• Before the start, the agency should explain which tools it uses, what they capture and when they load (after consent).
• Visitors who have not consented are missing from your data. How many that is depends on the banner and the audience. The numbers are therefore a sample of those who agree. That is no drama, but a bias you need to know about.
• The consent banner itself is a conversion factor. Good agencies do not test it for "more consent at any price" but for clarity and legal correctness.

How this works cleanly on the technical side is described in our guide to cookie consent and tracking under the TDDDG. Ask agencies specifically: "How does heatmap tracking after consent work at your company?" Anyone who hesitates there does not have the topic under control.

Technology: Who implements the changes?

The difference between slides and results often lies in implementation. Three models:

1. The agency analyzes, your team rebuilds. Works if you have a fast dev team. Otherwise there are waiting times.
2. The agency analyzes and rebuilds itself, directly in the system. This is the fastest route, but it requires the agency to master your CMS.
3. A testing tool serves variants via script. That is fast, but has limits with bigger rebuilds and can lead to visible flicker if poorly configured.

If your site runs on Webflow, Webflow Optimize is a built-in tool there for A/B testing and personalization, which according to Webflow also works with other systems through a JavaScript snippet (Webflow Optimize). Webflow does not state prices on this page. An agency blog (Digidop, so with a vested interest) lists monthly tiers from 299 USD at 25,000 page views up to 1,199 USD at 500,000 page views and a maximum of five simultaneous optimizations in the standard tiers (Digidop, Optimize by Webflow). Check current prices directly with Webflow. The same source, by the way, recommends having an internal or external expert support campaign and variant setup. The tool does not replace the method.

Other options are specialized testing platforms or custom solutions. For choosing an agency, only this counts: Does the agency have experience with the setup that makes sense for you, and can it explain why it recommends it, even if it is not its favorite tool?

The 9 vetting questions as a checklist

Ask these questions in the first conversation. The right-hand columns show how to recognize good answers.

No.	Question	Good answer	Red flag
1	How does the analysis work before you change anything?	Data analysis, behavioral data, conversations with sales or customers, prioritization	"We have proven templates"
2	How much traffic do you need, and what do you do when there is little?	Calculates sample sizes and run time in advance, suggests qualitative methods and big levers	Sells tests regardless of traffic
3	How do you measure success?	Qualified inquiries, CRM integration, a metric defined in advance	Only clicks or just the conversion rate
4	How do you handle lost tests?	Documentation, learning note, follow-up hypothesis	Losers are not mentioned
5	Who implements changes, and how fast?	Clear roles, realistic implementation time, access to your system	Unclear, "your team does that then"
6	How is tracking legally clean?	Consent before scripts, tool list, reference to the data protection officer	"It's all anonymous" with no detail
7	What examples are there from similar B2B situations?	Concrete cases with starting point, hypothesis, result and limits	Only percentages without context
8	What happens when the contract ends?	Test archive, access and documentation go to you	Knowledge and access stay with the agency
9	What results do you promise?	No fixed uplifts, but clear goals and reporting	Guaranteed increase in percent

A bonus question that reveals a lot: "What would you not do first at our company?" Anyone who has a service for every problem is looking for revenue. Anyone who says "we would fix the tracking first and not test yet" is thinking about your result.

Contracts, models and market ranges

CRO is offered in four models:

Model	Content	Advantage	Disadvantage
Audit (one-off)	Analysis, findings, prioritized list of measures	Manageable risk, good entry point	Impact only comes once someone implements
Project	One page or funnel is reworked and tested	Clear scope	No lasting learning effect
Retainer (ongoing)	Monthly cycle of analysis, hypotheses, tests, reporting	Continuous learning, needs traffic	Only worth it with enough volume
Success-based	Pay tied to results	Seemingly low risk	Definition of "success" is prone to disputes, incentive for short-term tricks

You can read more on ongoing versus one-off in the comparison guide Ongoing conversion optimization or one-off project.

For market ranges, there is hardly any reliable, independent data for Germany. A vendor site (ClicksGeek, so a vendor source, US dollars only, referring to the US market) gives the following for 2026 as a guide: a conversion audit between 3,000 and 8,000 USD one-off, a complete funnel optimization between 15,000 and 50,000 USD one-off and retainers between 3,000 and 20,000 USD a month, with a mid-range at 5,000 to 10,000 USD (ClicksGeek, CRO Agency Pricing). For German providers and euro quotes, this can differ significantly. So use the figures only to get a sense of the order of magnitude, and get at least two to three quotes.

What to watch for in the contract:

• Minimum term and notice period (for retainers, better to start short)
• What belongs in the monthly scope and what costs extra (for example designs, development, tools)
• Who owns test results, tool access and data accounts
• Who bears tool licenses
• Data protection: data processing agreement if the agency processes personal data

The first 90 days

This is what a sensible start should look like, regardless of provider. If your offer is structured very differently, ask for an explanation of why.

Period	Content	Result for you
Weeks 1 to 2	Goal clarification, access, tracking check, kick-off with sales	Measurement plan, list of data gaps
Weeks 3 to 6	Research: data, heatmaps, recordings, interviews	Findings report with evidence
Weeks 5 to 8	Hypotheses, prioritization, first direct fixes	Roadmap, obvious errors fixed
Weeks 8 to 12	First tests or bigger rebuilds running	Running experiments with a measurement plan
End of month 3	Evaluation, decision on continuing	Report with learnings, numbers, recommendation

If after 90 days neither findings nor running tests are visible, you should ask what is holding things up.

Red flags

• Fixed promises ("guaranteed plus 30%")
• No interest in your sales team or CRM
• Tests without a calculation of duration or sample size
• Only tool reports, no explanation of what they mean
• Winners are celebrated, losers are kept quiet
• Access and data sit exclusively with the agency
• One standard solution for every industry
• No mention of consent and data protection

When you do not need a CRO agency

To be fair: there are situations where the effort is not worth it.

• Very little traffic. Below a few thousand visitors a month on the important pages, tests deliver no insights. Invest in visibility first.
• The website is about to be relaunched. Then analysis insights flow better directly into the new concept instead of optimizing the old site.
• You have an in-house team with analysis, design and development. Then occasional consulting is often enough.
• You have a positioning problem. If your offer is not clear to the target audience, no optimization of details will help.
• A quick self-check is enough. With analytics, a free or cheap heatmap and conversations with three customers, you can often find the biggest problems yourself.

How we do it

Webwavers is an agency for branding, web design and web development. Our approach to conversion optimization follows the cycle of observing, forming a hypothesis, testing and implementing: analysis in the first month, first tests and landing pages in months two and three, then an ongoing testing cycle. We only use CRO when there is existing traffic, and with few visitors we recommend SEO or AI visibility first. Analyses run only with consent via the cookie banner, and every test ends with a decision. You can find more on the conversion optimization page.

FAQ

Which CRO agency fits a B2B website with many visitors but few qualified inquiries?

One that first clarifies why the inquiries are not qualified. Often the cause is an unclear message, the wrong audience targeting or forms without pre-qualification. Look for a provider that combines behavioral data with conversations from sales and tracks results all the way into the CRM. An agency that only wants to raise the form rate is more likely to make the problem worse.

What is better for demo requests: a specialist agency or a web agency with CRO?

That depends on your implementation team. A specialist agency brings deeper testing methodology but needs someone to build the changes. A web agency with CRO services implements directly but often has less statistical depth. With little traffic and frequent page changes, the web agency is usually more practical. With lots of traffic and your own dev team, the specialist pays off.

How many visitors do I need for A/B tests to make sense?

That depends on your conversion rate and the expected effect. With a 3% baseline rate and a hoped-for lift of 20%, you need about 14,000 visitors per variant, so about 28,000 in total (our own calculation, 5% significance, 80% power). With less traffic, qualitative methods and bigger rebuilds make more sense than small tests.

How do I tell whether a CRO agency is reputable?

By its method, not its promises. Reputable agencies start with analysis and a measurement plan, calculate sample sizes and run times, document lost tests, give you access and the test archive, and do not promise fixed percentages. The checklist with nine questions above is meant for exactly this.

Do I need a data protection check for heatmaps and tests?

Yes. Many heatmap and testing tools access the end device or process personal data. Under § 25 TDDDG, consent is usually required for that. Have the agency name the tools and the process in advance, and agree the details with your data protection officer.

Can the same agency handle SEO and CRO?

Yes, that can make sense because the two work together: SEO brings suitable visitors, CRO turns them into inquiries. Demand separate metrics and separate reports, otherwise the effect of each area cannot be told apart. It is also worth checking whether the agency has references in both areas and is not just selling one of them along the way.

What does a CRO agency cost?

For Germany, there is no reliable independent market overview. A vendor source from the US gives audits from about 3,000 USD and retainers from about 3,000 USD a month (see the section on market ranges for the source and its limitations). What matters is what is included in the price: analysis, design, development, tools, reporting. So compare quotes by scope of services, not by monthly total.

Conclusion

You recognize the right CRO agency for your B2B website by method, measurement and honesty. It calculates whether your traffic is enough for tests, judges results by qualified inquiries, keeps data protection clean and also tells you when you do not need it yet. Take the nine questions into your next conversation, compare at least two quotes, and start with a manageable audit rather than an annual contract.

If you would like to talk to us about your site, you will find details on conversion optimization and on the matching industry page for IT, software and SaaS, or for law firms and consultancies the page on consulting and agencies. We answer questions about your setup via contact.

Sources

• 6sense, B2B Buyer Experience Report 2025, press release (Nov. 2025)
• Gartner, Sales Survey: 67 Prozent der B2B-Käufer bevorzugen Kauf ohne Vertriebsmitarbeiter (March 2026) (67 percent of B2B buyers prefer a purchase without a sales rep)
• Unbounce, Conversion Benchmark Report SaaS
• Growth Method, The 100 Conversions Rule
• Evan Miller, How Not To Run An A/B Test
• § 25 TDDDG, Gesetze im Internet (German laws online)
• Webflow Optimize
• Digidop, Optimize by Webflow (agency blog, 2026)
• ClicksGeek, CRO Agency Pricing (vendor blog, 2026)