Conversion rate optimisation that starts with tracking you can trust

There is no single good figure, because conversion rates vary widely by industry, traffic source, price point and what counts as a conversion. A free consultation request and a high-value purchase are not comparable, and paid search visitors usually behave differently from social media traffic. The most useful benchmark is your own baseline, split by device, channel and landing page, which shows where you are weakest and where improvement is realistic.

Many smaller sites do not, so we calculate it before proposing any test. The sample needed depends on how many conversions the tested page produces and the size of change you want to detect, and low-volume pages may never reach a reliable result. In that case we rely on qualitative research, usability sessions and sequenced changes with careful before-and-after comparison, and we say plainly that the evidence is weaker.

Track the actions that represent business value first, then the steps that lead to them. For most sites that means enquiries, bookings, purchases or sign-ups as key events, plus supporting steps such as form starts, pricing page views, add to basket and checkout progress. We write each event into a measurement plan with consistent names and parameters, keep personal data out of event payloads, and avoid tracking every click, because noisy data hides the signals that matter.

We treat tracking as part of the codebase rather than tags bolted on afterwards. Pages push events into a documented data layer, and Tag Manager reads from it instead of relying on button text or styling classes that designers routinely change. Each release is checked against a tracking test list on staging, and alerts flag sudden drops in key events after deployment, so a broken form or tag is caught within days rather than at month end.

Yes, and how much depends on the mode. In basic mode, Google tags do not load until a visitor accepts, so declined visits are missing entirely. In advanced mode, tags send cookieless signals before consent, and Google can model part of the gap once your property meets its data thresholds. Google also expects consent signals for UK and EEA visitors before some advertising features work. We explain which reported figures are observed and which are modelled.

Server-side tagging routes tracking data through a server you control before it reaches analytics and ad platforms, instead of loading every vendor's script in the visitor's browser. It can improve page speed, give you control over what each platform receives and make conversion data more durable. It also adds hosting cost and maintenance. We recommend it when you spend meaningfully on advertising, run several ad platforms or handle sensitive data, not by default.

We rank ideas by expected impact, strength of evidence and effort. Impact reflects how much traffic and revenue pass through the page or step, evidence comes from funnel drop-off, recordings, survey answers and sales feedback, and effort covers design and build time. A broken mobile form usually outranks a new hero image. Clear bugs are fixed directly rather than tested, because there is nothing to learn from testing a fault.

A test should run until it reaches the sample size calculated before launch, and across full weekly cycles so weekday and weekend behaviour are both included. Stopping as soon as one variant looks ahead is a common cause of false winners. We set the duration, primary metric and stopping rule up front, and avoid launching tests during promotions or seasonal peaks that would distort the outcome.

We choose per site rather than by habit. For Next.js applications we can run experiments on the server or at the edge, which avoids page flicker and keeps pages fast. For WordPress and Shopify sites we pick a testing platform that suits your traffic and budget, or use the store's built-in options. Whichever tool runs the test, results are checked against Google Analytics or your back-end data, and winning variants are rebuilt properly in the site code.

Not if the test follows Google's published guidance on website testing. Search engines must see the same content as visitors, so we never serve Googlebot a special version. When variants live on separate URLs, they carry a canonical tag pointing to the original and any redirect is temporary rather than permanent. Tests run only as long as the sample size requires, and the winning version is then built into the page so search engines index one stable design.

Usually yes, if they are configured carefully and disclosed. We mask all form inputs and on-screen personal details by default, exclude account, payment and other sensitive pages, record a sample of sessions rather than every visit, and set the shortest retention period that still supports the research. Recording tools should be named in your privacy notice and, where they set non-essential cookies, covered by your consent banner. Your legal adviser makes the final call.

Cost depends on the state of your current tracking, how many sites, platforms and conversion types are involved, and whether you need ongoing testing or a one-off setup. Most engagements start with a fixed-price measurement audit and tracking repair, because optimising on unreliable data wastes money. After that we quote a monthly scope covering research, design, build and experiment analysis, with the number of tests set by what your traffic can support.

You do. The Google Analytics property, Tag Manager containers and any server-side tagging project are created in your organisation's accounts, with our team added as users you can remove. We set data retention to the longest standard option of 14 months and can enable the BigQuery export, so raw event data is kept in your own Google Cloud project. The measurement plan, event names and consent settings are documented so another team can maintain them.