Conversion rate optimisation Switzerland.

The conversion rate is the share of visits that turn into a desired action: conversions divided by visits, times one hundred. Conversion rate optimisation, or CRO, is the work of raising that share without buying more visitors. The arithmetic behind it explains the popularity of the discipline: lifting the rate from 1.5 to 3 per cent halves your cost per enquiry without spending a franc more on advertising.

What CRO is not: a collection of tricks. The idea that a different button colour, a countdown or a note about supposedly limited availability moves the numbers persists stubbornly and is true only rarely. In Swiss B2B settings, urgency mechanics are usually harmful, because they read as unserious. The large movements come almost always from three sources: the offering became easier to understand, the interface became easier to use, or the visitors were the right ones from the start.

The third point gets overlooked routinely. A low conversion rate is frequently not a page problem but a traffic problem. A page receiving visits from searches that have little to do with the offering will never convert well, however it is designed. Before you work on the page, it is worth looking at which search terms and campaigns people arrive through — a question that belongs to search optimisation rather than to design.

The formal definition of the metric, with a worked example, is in our glossary entry on the conversion rate. For this page it is enough to know that the metric is a ratio, and that ratios built from small numbers fluctuate heavily. That is exactly where the problem in the next section comes from.

An A/B test only delivers a valid result once enough people have seen both variants — and at typical conversion rates, "enough" means tens of thousands. The values below apply to a two-sided test of two variants at a 95 per cent significance level and 80 per cent statistical power, starting from a conversion rate of 2 per cent:

• +10 per cent relative (from 2.0 to 2.2 per cent): around 80,700 visitors per variant, a good 161,000 in total.
• +20 per cent relative (from 2.0 to 2.4 per cent): around 21,100 per variant, a good 42,000 in total.
• +25 per cent relative (from 2.0 to 2.5 per cent): around 13,800 per variant, a good 27,600 in total.
• +50 per cent relative (from 2.0 to 3.0 per cent): around 3,800 per variant, a good 7,600 in total.

Now the counter-calculation. A typical Swiss SME landing page receives between 300 and 1,500 relevant visits a month. At 800 visits, proving a 20 per cent improvement takes around 52 months. Even the cheapest case in the list — an improvement of half, which only occurs on badly broken starting pages — needs close to ten months. In that time the season, the campaigns, the prices and the competition all change. The test is then no longer measuring the variant. It is measuring the year.

On top of that sits a mistake even more common than a runtime that is too short: repeated peeking. Anyone checking the tool daily and stopping the test the moment "significant" appears somewhere raises the chance of a false alarm considerably. Given enough looks, there is almost always a point at which one variant happens to be ahead. That is precisely how the success stories about button colours arise, and why they never reproduce.

The consequence is not that conversion work is pointless. The consequence is that the A/B test is the wrong tool for a small website — and that any provider selling it as a core service without first looking at your visitor numbers has either not done the arithmetic, or is counting on you not doing it.

The mechanics of the method itself, including what significance and power actually mean, are set out in our glossary entry on A/B testing.

Before any conversion work comes the question of whether the numbers are right at all — and in an alarming number of accounts they are not. The most common findings: the form submission is not recorded as a goal, the thank-you page view is, and that page can be opened directly. Phone calls are not counted at all. The consent banner blocks part of the measurement without anyone knowing the size of the gap. Internal visits from your own team run into the same figures.

As long as that is the case, every optimisation produces movement in a metric that represents nothing. Conversion work with us therefore always starts with the measurement: goal definitions, checking the triggers, splitting by channel and device, and handling the consent gap. How that is set up technically is described in detail on the page about our conversion tracking agency in Zurich. It is the precondition for everything that follows here, not an upsell.

One point deserves naming, because it appears in almost no quote: part of your conversions will never be measured. Visitors who decline the consent banner do not appear in the ad accounts. If you sell by telephone, you additionally lose everything decided in conversation. The right answer is not to argue the gap away, but to size it and read the numbers accordingly.

Only once the measurement stands does the second question become useful: where exactly do people drop out? No overall conversion rate answers that. It takes a breakdown by step, device and source. Almost always the largest loss sits at a single point — and it is rarely the point anyone expected.

When statistical proof is off the table, what remains are methods that do not need it — because they make problems visible instead of measuring effects. In our experience these six return the most, in this order:

• Watch real people use it. Five moderated tests with people from your audience expose the bulk of the coarse usability problems. That is not a statistical procedure, it is observation: if three out of five get stuck on the same field, you do not need a significance test.
• Review recordings and heatmaps. Session recordings show drop-out points, scroll depth and clicks on elements that are not clickable. Microsoft Clarity costs nothing and is enough for most SME websites. What the visualisation does and does not tell you is covered in our entry on the heatmap.
• Take friction out of the form. Every field not needed for a first contact should go. Salutation, company size and budget range can be settled in conversation. The same applies to mandatory fields that are hard to complete on a phone.
• Say what it costs. Missing prices are one of the largest conversion losses in Swiss service businesses. A stated range filters out unsuitable enquiries and raises the close rate on the rest — and both of those are wanted.
• Fix load time and mobile usability. Both are technically measurable without any test. Buttons that are too small, overlays that cover the content and a hero image that takes three seconds all cost enquiries. What helps here is a look at the UX and interface work, not at the statistics.
• One page per intent. Sending campaign traffic to the home page is the most expensive avoidable mistake there is. A dedicated landing page per offer, with exactly one possible action, beats any fine-tuning of an overloaded page.

What stands out about this list: not one point needs a testing tool, and most are one-off pieces of work rather than ongoing management. That is exactly why we rarely recommend a monthly retainer for conversion work on small websites.

The before-and-after comparison is the usual substitute. It is legitimate, but only if its weaknesses are stated alongside it. It compares two periods instead of two simultaneous groups. Everything else that changed between those periods flows into the result: the season, the campaign budget, a price change, a competitor starting to advertise, a public holiday.

Three rules make the comparison more defensible. First, equally long periods, at least four weeks and preferably eight. Second, evaluate by channel, because otherwise the composition of the traffic shifts unnoticed. Third, run an unchanged comparison page alongside. If enquiries rise on the changed page and stay flat on the unchanged one, that is a usable indication — not proof, but considerably more than a bare comparison over time.

And then honesty requires stating the order of magnitude. A before-and-after comparison over eight weeks with 1,200 visits can reveal a doubling. It cannot reveal a ten per cent improvement. Reading decimal places out of data like that sells chance as performance. We therefore write into reports which statements the volume of data supports and which it does not — including when that makes the result look less impressive.

For the same reason you will find no percentage figure on this page presented as a reference. DLM Digital runs no published conversion optimisation engagement, and a value such as "plus 38 per cent enquiries" would be a claim with nothing behind it. What we can evidence is websites, online shops and SEO engagements; they are listed under our work.

In four situations we advise against conversion work, because the money achieves more elsewhere. That costs us engagements and saves you disappointed expectations.

• You have too few visitors. Below roughly 300 relevant visits a month, even a marked improvement is barely noticeable in absolute terms: three enquiries become four. The money then belongs in visibility rather than in fine-tuning — through ongoing SEO or paid search.
• The website is at the end of its life technically or visually. If the site comes from a dated page builder, breaks on a phone and can only be changed with outside help, optimising is patchwork. A rebuild is then the more honest route; the price bands are set out under web design and prices.
• The offering itself does not convince. No copy and no design makes an offering attractive that is too expensive, too vague or too late for the market. A higher conversion rate would only mean more people declining after reading the quote.
• Enquiries sit unanswered. If several days pass between a form arriving and a call back, the website's conversion rate is not the bottleneck. No agency can solve that part for you.

Which of these applies to your company is settled faster in a conversation than in an analysis. That is what our digital marketing consulting is for: fixing the order, before budget flows into the first channel that comes to mind.

We deliberately name no monthly price for conversion optimisation, because the effort depends entirely on the starting position. Whether the measurement works, whether one page or twenty are affected, and whether only copy or also engineering has to change makes a difference of a factor of ten. A flat rate in that situation would be a bad deal either for us or for you — and in both cases a number nobody can name seriously before the analysis.

Where conversion work forms part of a larger engagement, that engagement's prices apply, and ours are public: a website sits between CHF 3,000 and CHF 25,000, ongoing SEO starts at CHF 800 a month, local SEO at CHF 300 a month. In those engagements the conversion view is included anyway — building a page so that it produces enquiries is not an add-on, it is the job.

For a standalone conversion analysis we estimate the effort after a first conversation and name a binding figure before work begins. The analysis covers a check of the measurement, an evaluation of drop-out points by channel and device, a review of session recordings, and a prioritised list of concrete changes with an effort estimate per item. What you do with it is your decision; the implementation does not have to sit with us.

To put your own numbers in context: work out what an enquiry costs you today by dividing your advertising budget by the number of enquiries. That single figure tells you what an improvement would be worth — and whether the effort pays. For the standard services, the range can be estimated with our project calculator.

The market for conversion optimisation is full of numbers with no origin, and few of them survive a follow-up question. Five questions separate the defensible providers from the rest:

• "How many visitors did the page have where this result was produced?" Without that number, any percentage is worthless. At 400 visitors a month, a 30 per cent improvement is noise.
• "Was that an A/B test or a before-and-after comparison?" Both are legitimate, but their weight differs considerably. Anyone dodging the question probably compared and claimed to have tested.
• "What else changed in the same period besides the tested change?" New campaigns, a price update or a website relaunch in the same month devalue the result.
• "Are you counting form submissions or actual orders?" Doubling the forms while halving the quality is not a success. Good providers talk about qualified enquiries.
• "What do you recommend when the data supports no statement?" The right answer is: do not test, observe and fix the obvious faults. Anyone selling a test plan instead is selling runtime.

The same questions apply to us. DLM Digital has no published CRO reference, and this page therefore claims no result. What you get is an analysis that says what the data supports — and what it does not. If that is too little for you, that is an understandable reason to hire somebody else; demand the same five answers there.