Conversion Optimization: Ongoing or One-Off, In-House, With Tools or With an Agency?

The key points in brief Ongoing conversion optimization pays off when your pages get enough visitors for tests and you can implement changes quickly. With low traffic or an obvious defect, a one-off project with analysis and rebuild is often enough. Whether you go in-house, with tools or with an agency depends less on company size than on whether you can cover analysis, design, development and statistics. Calculate in advance how many extra leads you need for the model to pay for itself.

• Two decisions: who does it (in-house, tools, agency) and how (ongoing or one-off)
• Tools deliver data, not hypotheses
• One-off fits a relaunch, an audit and clear defects, ongoing fits high traffic and a need to learn
• Extra leads per month = visitors times the difference in conversion rate
• You calculate the break-even point from lead value and close rate, using assumptions you enter yourself

Table of contents

• Two questions that often get mixed up
• Who does it: in-house, tools or agency?
• Which roles CRO actually needs
• Use tools yourself or outsource the analysis?
• How: ongoing or one-off?
• A four-step decision guide
• Calculator: extra leads per month
• Break-even: when does a retainer pay for itself?
• Cost comparison
• Common mistakes
• How we do it
• FAQ

Two questions that often get mixed up

Anyone looking into conversion optimization runs into two questions that tend to get thrown into one pot:

1. Who does it? An internal team, you yourself with tools, or an agency?
2. How does it run? As a one-off project with a clear end or as an ongoing process with a monthly rhythm?

You can combine the two questions. An internal team can handle a one-off project, an agency can run a permanent retainer, and there are hybrids in between. If you answer them separately, the decision gets easier. This guide goes through both questions, gives you calculation tools and tells you when no option makes sense.

Some background on the terms: The conversion rate describes the share of visitors who take a desired action. CRO is the ongoing effort to improve that rate and the quality of those actions. We describe how to pick a B2B agency separately in the guide How to choose a CRO agency for B2B websites.

Who does it: in-house, tools or agency?

There are four realistic paths. The table compares them side by side.

Criterion	In-house team	DIY with tools	Agency	Hybrid (agency plus internal)
Typical start	Takes months to build	Immediately	Weeks until kick-off	Weeks
Methodology and experience	Grows over time, depends on individuals	Depends on the person's knowledge	Proven across many projects	Good if roles are clearly split
Closeness to product and sales	Very high	High	Has to be built up	High
Implementation speed	Depends on dev capacity	Limited to what the tool can do	High if the agency works in your system	Depends on agreements
Statistical rigor	Often a weak spot	Often a weak spot	Usually better	Good
Cost structure	Salaries, tools, training	Licenses plus your own time	Monthly or project price	Both
Risk	Loss of key people	The tool becomes busywork	Dependency, knowledge leaves at contract end	Unclear responsibility
Fits	High traffic, several products, budget for roles	Small budget, first steps, learning goal	Missing capacity or experience	Teams with marketing and development that want guidance and speed

A complete comparison case comes from a question many startups ask: agency or in-house growth team? The honest answer is that each has its moment.

• Early stage, little traffic, little money: A growth team is a luxury you don't need. A one-off project (positioning, hero, pricing page) or targeted advice usually helps more. The foundation is a clear offer, not test operations.
• Growth stage, rising traffic, marketing budget: An agency can bring methodology and speed while you learn internally. A handover plan is important.
• Scaling, high traffic, several markets: An in-house team pays off, because closeness to product and data becomes more important. An agency can absorb peaks or take on specialist topics.

So the value of an agency lies not in a permanent solution but in a window of time in which it brings you more than building a team would. Plan this handover from the start.

Which roles CRO actually needs

One argument for agencies is the sheer range of skills that sound CRO requires. Rarely does everything sit in one person. You need:

1. Analysis. Read data from web analytics, behavior tools and CRM and turn it into findings.
2. Research. Interviews, review of support and sales conversations, competitor and heuristic review.
3. Concept and design. Turn findings into variants that also fit the brand visually.
4. Development. Build variants in the system without wrecking load time.
5. Statistics and test design. Sample size, run time, evaluation, handling of multiple tests.
6. Project management. Roadmap, alignment with sales, documentation.

A marketing lead who does CRO "on the side" usually covers two or three of these roles. That is enough for quick wins. For ongoing test operations with clean evaluation, it gets tight.

Use tools yourself or outsource the analysis?

The question often comes up in small form: heatmap tools are cheap, so use them yourself or leave the analysis to an agency? The answer lies in the difference between data and insight.

What a tool delivers: a heatmap shows where visitors click and how far they scroll. Recordings show individual sessions. Form analytics show which field makes users drop out. That is valuable and quick to set up.

What a tool does not deliver: the reason. The heatmap shows that hardly anyone clicks the button. Whether that is down to the position, the copy, the offer or missing trust only gets decided by interpretation. And that starts where knowledge of the target audience, sales and competitors comes together.

A pragmatic split for small teams:

• Do it yourself: Set up, observe, fix simple obvious errors. For example, a form that can't be submitted on a smartphone.
• Get help: Forming hypotheses, prioritization, test design and evaluation, when decisions affect a lot of revenue.

Another point is consent. Many of these tools access the end device or capture behavioral data. Under Section 25 TDDDG (German Telecommunications Digital Services Data Protection Act), this generally requires consent, unless the access is technically strictly necessary (§ 25 TDDDG). We explain the technical process in the guide on cookie consent and tracking under TDDDG. This applies to DIY users just as much as to agencies.

On testing tools: Webflow offers its own tool for A/B testing and personalization, called Optimize. According to Webflow, it runs on any CMS via a small JavaScript snippet; certain integrations such as Salesforce or HubSpot require the Enterprise plan (Webflow Optimize). The page does not state prices. An agency blog lists tiers from 299 to 1,199 USD per month, depending on monthly page views, and notes that an internal or external expert makes sense for setup and ongoing support (Digidop, an agency source with its own interest, please check current prices with Webflow). For your decision, this means: A tool costs money and does not yet produce insight. You have to get the knowledge separately, internally or externally.

How: ongoing or one-off?

The second question is about rhythm. Both models have a goal, but different mechanics.

One-off project

Here a clearly defined task gets worked on. Typical forms:

• Audit and action list: Analysis of the existing site, findings, prioritized recommendations. Implementation happens later, often by your own team.
• Landing page or funnel rework: A page or flow is newly conceived, built and possibly tested alongside.
• Relaunch with a CRO focus: The insights flow straight into the new site. This is often the most sensible moment, because you are not optimizing legacy problems.

Strengths: manageable budget, clear end, fast impact on obvious defects, no long-term commitment. Weaknesses: no lasting learning effect. User behavior changes, competitors catch up, campaigns rotate. Without upkeep, the site drifts back.

Ongoing optimization

Here there is a monthly cycle of analysis, hypothesis, test, evaluation and implementation. A fixed rhythm with reporting and a shared roadmap is typical.

Strengths: continuous learning, response to change, tests built on a growing knowledge base, better alignment with campaigns. Weaknesses: needs traffic, needs discipline in implementation, impact only shows over months, costs keep running even when there is little to test.

Direct comparison

Criterion	One-off	Ongoing
Goal	Solve a specific problem	Keep getting better
Minimum traffic	Low to medium	Medium to high
First results	Weeks	More reliable from about the third month
Learning effect	Low to medium	High
Risk	The impact fizzles out	Ongoing costs without tests
Fits	Relaunch, audit, clear defects, low traffic	High traffic, campaigns, several pages, high lead value
Example B2B software	New demo page with a reworked form	Monthly tests on the home, pricing and demo pages
Example consulting	New contact flow to the initial call	Continuous tests of trust elements and appointment booking

A third option is the switch: project first, then retainer. You start with an audit, fix the obvious defects, and only once the foundation is right and traffic is there does the test cycle begin. In many cases this is the most sensible order. We describe how this looks for SEO and GEO in the comparison SEO and GEO: retainer or one-off project.

A four-step decision guide

Go through the steps in order. As soon as an answer is "no", the recommendation is settled.

1. Do you have reliable tracking? If not: build measurement first, don't optimize yet.
2. Does your most important page get enough visitors for tests? Use the worked examples in the guide How to choose a CRO agency for B2B websites as a guide: with a 3% baseline rate and a hoped-for plus 20%, that is about 14,000 visitors per variant. If not: a one-off project with qualitative analysis, or more visibility first.
3. Can you implement changes within one to two weeks? If not: either buy the implementation too or create a faster path. A retainer stuck in a queue burns money.
4. Is there someone internally who understands results and makes decisions? If not: that is the role you need to fill first, whichever model you choose.

If all four questions get a "yes", a lot speaks for an ongoing process. Whether in-house or agency is then decided by capacity and experience.

Calculator: extra leads per month

The central calculation is simple. It answers: how many additional inquiries does a better conversion rate bring?

Formula:

Extra leads per month = visitors per month × (new conversion rate − old conversion rate)

Inputs:

• Visitors per month on the affected page or flow
• Current conversion rate (as a share, for example 2% = 0.02)
• Expected new conversion rate or relative increase

Assumptions you should disclose:

• The additional leads are of the same quality as the previous ones. (That is not always the case. Check it in your CRM.)
• Traffic stays the same.
• The effect lasts.

The following example table shows what comes out, purely arithmetically. All numbers are assumed example values, not a forecast and not a promise.

Visitors per month	Baseline conversion	Relative increase	New conversion	Extra leads per month
2,000	2.0%	+20%	2.4%	8
2,000	2.0%	+50%	3.0%	20
5,000	2.0%	+20%	2.4%	20
5,000	3.0%	+20%	3.6%	30
10,000	2.0%	+20%	2.4%	40
10,000	3.0%	+30%	3.9%	90
25,000	1.5%	+20%	1.8%	75
25,000	3.0%	+20%	3.6%	150

What the table shows: at 2,000 visitors, even a strong improvement brings only a handful of inquiries. That is why ongoing programs rarely pay off with small traffic. At 10,000 visitors and more, the picture changes.

Uplift for ongoing programs. An ongoing process doesn't improve once but several times. For that you need two more assumptions, which you have to enter honestly yourself:

• How many tests run per year?
• How many of them bring a real, lasting gain? (Hit rate)

A worked example: Say you run 6 tests a year, and 2 of them each bring a +10% relative improvement. The conversion rate then rises by about 21% after a year (1.1 × 1.1 = 1.21). At 10,000 visitors and a 2% baseline rate, that is 200 leads before and about 242 after, so roughly 42 extra leads a month. If only one test wins, it is about 20. Enter your own values. A retainer provider who never tells you a hit rate is not calculating honestly.

Important: Published win rates for A/B tests vary widely by provider and by which tests are selected. We therefore deliberately give no market value and leave the assumption open.

Break-even: when does a retainer pay for itself?

Extra leads are only half the calculation. What matters is what they are worth. For that you need:

• Lead value: How many leads turn into appointments, quotes, orders? And what is the average order worth (contribution margin)?
• Cost of the model: Monthly price plus internal time plus tools.

Formula:

Value per extra lead = close rate × average contribution margin per customer Break-even extra leads per month = total monthly cost ÷ value per extra lead

Worked example (freely chosen assumptions):

Assumption	Value
Close rate, lead to customer	10%
Average contribution margin per customer	€12,000
Value per extra lead	€1,200
Total monthly cost of the model	€4,000
Break-even extra leads per month	about 3.3

In this example, the optimization has to bring a good three additional leads per month on average for the amount to pay for itself. With the numbers from the table above, that would already be reachable at 2,000 visitors and a 20% increase (8 extra leads), if the improvement actually materializes. Flip the assumptions (close rate 3%, contribution margin €4,000) and the value per extra lead is €120 and the break-even point is above 33 extra leads a month. Those are completely different conditions.

The lesson: retainer models fit high lead value and a lot of traffic better. With low lead value and few visitors, a manageable project pays off better.

A warning: This break-even ignores the time lag. In B2B, months often pass between inquiry and close. According to 6sense, the average B2B buying cycle in 2025 was around 10 months (6sense, press release 2025). A retainer therefore pays off only with a delay. That argues for minimum terms of several months, but against long, restrictive contracts.

Cost comparison

For Germany there is no reliable, independent market overview of CRO prices. As a rough guide, there is a US vendor page (ClicksGeek, 2026, vendor source, USD only). It lists conversion audits between 3,000 and 8,000 USD one-time, full funnel optimizations between 15,000 and 50,000 USD one-time, and ongoing CRO retainers between 3,000 and 20,000 USD per month, with the middle range at 5,000 to 10,000 USD (ClicksGeek, CRO Agency Pricing). For a landing page redesign including testing, the same source cites 10,000 to 30,000 USD. German offers may differ, so please read this as an order of magnitude.

With the in-house model, you calculate differently. You budget salaries for the roles you want to cover, tool licenses and the time that departments invest. Hourly rates of agencies and freelancers as a market alternative give a reference value: for web agencies, the Adfera Website Cost Report 2026 cites roughly €80 to €120 per hour, for freelancers in web design and development €60 to €90 (Adfera, Website-Kosten-Report 2026). According to the source, the values vary widely, and other providers cite higher ranges. So calculate with a range, not a single point value.

To compare the models:

Model	Cost drivers	What is often forgotten
In-house	Salaries, recruiting, tools, training	Onboarding, vacation, turnover
Tools yourself	Licenses, your own time	Time of the departments, missing statistics
Agency ongoing	Monthly price, add-on services	Implementation, design, tool costs, internal coordination
Agency one-off	Project price	Upkeep afterwards

Common mistakes

• Retainer without traffic. The classic mistake: a test program for a page with 1,500 visitors a month.
• Tool before method. An expensive platform gets bought, but nobody knows which hypotheses to test.
• No path to implementation. Results sit in presentations, changes wait in the backlog.
• Wrong goal. The click rate on a button gets improved, not the number of qualified inquiries.
• Stopping too early. Tests get stopped as soon as one variant is ahead. That raises the risk of false winners significantly (Evan Miller, How Not To Run An A/B Test).
• Changing everything at once. Five changes at the same time, and afterward nobody knows which one works.
• Contract end without handover. Access, test archive and documentation belong to you.
• Buying once, then never touching it again. Even a good project ages. Plan at least an annual review.

How we do it

At Webwavers, conversion optimization runs as a cycle of observing, hypothesis, test and implementation. The start is an analysis in the first month, followed by first tests and landing pages, and from the fourth month an ongoing test cycle. One prerequisite is existing traffic. With few visitors, we recommend SEO or GEO first. You commission landing pages from us as needed, analysis as well as forms and funnels run continuously, and hypotheses, tests and evaluation run monthly. The evaluation is oriented toward goals rather than clicks, and analytics only run with consent through the cookie banner. You can find details on the Conversion Optimization page.

FAQ

Conversion optimization through an agency or an in-house growth team: which makes more sense for a SaaS startup?

For a young startup with little traffic, usually neither as a permanent setup. A clear offer, clean measurement and a good hero matter more. A one-off project or targeted advice helps here. As traffic grows, an agency can deliver speed and methodology until an in-house team makes sense. Plan the handover from the start.

Is an ongoing conversion package with heatmaps and A/B tests worth it compared to a one-off optimization of the landing pages?

That depends on traffic and lead value. Ongoing packages pay off with many visitors, a high value per lead and fast implementation. With low traffic or clearly visible defects, a one-off project is enough. A break-even calculation with your own assumptions (see above) makes the decision concrete.

Should I use heatmap tools myself or hand analysis and hypotheses to an agency?

You can set up and observe yourself. Interpretation, forming hypotheses and test design benefit from experience and statistics knowledge. A good split: you fix obvious errors yourself and get help with prioritization and tests. Keep consent under TDDDG in mind before tools capture data.

How long should an ongoing CRO package run at minimum?

At least three to six months, because analysis, first tests and reliable results take time. In B2B, add the time lag until close. Still, avoid long commitments at the start. A shorter starting term with clear interim goals and an extension option is better.

Can CRO agencies optimize directly in the Webflow CMS?

Some can, some can't. That is a capability of the individual agency, not a rule. Ask whether changes are implemented directly in your system and who is responsible for tests. Webflow has its own testing tool in Optimize. What stays decisive is that someone masters hypotheses, setup and evaluation.

How much traffic do I need for an ongoing program?

There is no fixed number. As an example: with a 3% baseline rate and a hoped-for plus 20%, about 14,000 visitors per variant are needed (our own calculation, 5% significance, 80% power). With less traffic, you only test big changes or use qualitative methods.

What is a one-off project worth when campaigns and offers change?

It stays valuable if it solves fundamental problems: messaging, structure, form, trust. For changing campaigns you additionally need a light upkeep process, for example an annual review or landing pages as needed. Continuous tests only pay off when there is enough volume.

Conclusion

You answer the question "ongoing or one-off" with three numbers: traffic, lead value and implementation speed. You answer the question "in-house, tools or agency" by asking which roles you are missing. Calculate the extra leads with your own assumptions, set them against the costs and, if doubts remain, start with an audit rather than an annual contract. The order of project first, retainer after, is the most honest one for many companies.

You can find help with sizing this up on our Conversion Optimization page. For the industry view, see IT, software and SaaS for software providers and Consulting and agencies for law firms. If you want to talk to us: Contact.

Sources

• § 25 TDDDG, Gesetze im Internet (German Federal Law Online)
• Webflow Optimize
• Digidop, Optimize by Webflow (agency blog, 2026)
• ClicksGeek, CRO Agency Pricing (vendor blog, 2026)
• Adfera, Website-Kosten-Report 2026 (Website Cost Report 2026)
• 6sense, B2B Buyer Experience Report 2025, press release (Nov. 2025)
• Evan Miller, How Not To Run An A/B Test