Website & Conversion Optimization Playbook

A B2B website converts when a buyer can tell within seconds what you sell and who it is for, finds the proof and the price they came for, and can ask for a demo without a struggle. Fix those first, judge every change by the qualified pipeline it brings, and A/B test only where your traffic can give an answer.

Most B2B websites were built to look credible at launch, and nobody has checked since whether they turn their visits into sales conversations. I wrote this for B2B SaaS and high-consideration B2B service companies that get some traffic, fewer demo or contact requests than they would like, and website changes driven by opinion rather than data. The nine steps run in three phases over about 12 weeks: diagnose in weeks 1 to 3 (Steps 1 to 4), improve in weeks 4 to 8 (Steps 5 to 8), then set up the routine that keeps the gains in weeks 9 to 12 (Step 9). If you sell to consumers, the B2C edition covers product pages, checkout and order value instead.

Start by agreeing what the site is for. A B2B site has a few primary conversions, the actions that start a sales conversation: a demo request, a trial start, a pricing or proposal request, a contact request. Everything else is a secondary conversion, a signal of interest such as a newsletter sign-up, a content download, a webinar registration or a key video play. Count the two separately. A month with more downloads and fewer demo requests is a worse month, and a single “conversions” number hides that.

Write the list as a conversion map: each conversion, the page where it happens and the event that records it. In Google Analytics 4 (GA4), the events that matter most are marked as key events, the name Google now uses for what it used to call conversions. A standard property can mark up to 30, and marking an event does not change data already collected, so mark yours before the audit starts^[1],[2].

Decide now which segments you will read every number by: channel (paid, organic, direct, referral, email), new against returning visitors and, where you can tell, visitors from companies that fit your ideal customer profile against everyone else.

Before you judge any page, make sure the numbers are accurate enough to decide with. Three checks catch most problems:

• Events. Each form fires its event on a successful submission, not on a button click, and a visit that crosses between your domain and a subdomain, such as a separate booking or app domain, stays one session.
• Forms and CRM. Every submission reaches the CRM once, with its source, medium, campaign and landing page, and lands with the right owner. Compare a month of form events in analytics with the leads the CRM created from the same forms; the two should match closely.
• Baseline. Record the sitewide conversion rate, the rate on each key page (home, demo or trial, pricing, the top landing pages) and both split by desktop and mobile. Every later change is judged against this row.

The GA4 Tracking Plan and UTM Builder lists the events and source fields in one sheet, and AAJ’s Marketing Technology Stack Landscape shows where the website layer connects to the rest of the stack.

Add page speed to the baseline. Google’s Core Web Vitals judge a page at the 75th percentile of real visits, mobile and desktop separately. Largest Contentful Paint, which measures loading, should be 2.5 seconds or less; Interaction to Next Paint, which measures the delay after a tap or click, 200 milliseconds or less; Cumulative Layout Shift, which measures how much the page jumps, 0.1 or less. Above 4 seconds, 500 milliseconds and 0.25, a page is rated poor^[3]. Some 2026 articles report tighter thresholds; Google’s own documentation still gives these^[4].

Read real-visitor data, not one lab test on your office connection. PageSpeed Insights reports real Chrome visits over the previous 28 days, and when a page has too few of them, as many B2B pages do, it falls back to your whole site. Google does not disclose the minimum number of visitors^[5],[6].

Now look at the site the way a distracted prospect does. Jakob Nielsen, summarizing a Microsoft Research study of more than two billion page visits, found that people often leave a web page within 10 to 20 seconds, and that a page has to state its value proposition within 10 seconds to hold attention for longer^[7],[8].

Review five kinds of page: the homepage, the product or solution pages, the demo, trial or contact page, the pricing page, and the blog or resource pages with the most traffic. On each, check five things:

• Clarity. Can a new visitor say what you do within 10 seconds?
• Relevance. Is it obvious who it is for, and does it name a problem they have?
• Value and proof. Does it say what the buyer gets, and why they should believe it?
• Visual hierarchy. Is the path to the main call to action obvious?
• Friction. Confusing navigation, competing calls to action, long forms, jargon.

Test clarity on other people; you know your product too well to judge it. The Five-Second Message Test shows your page for five seconds, then asks four questions: what is it, who is it for, why choose it and what to do next. I use ten people from the roles you sell to: enough to see whether the message lands, not to measure it precisely. The Website Conversion Audit checks the rest of the list, from speed to forms, in one pass.

The audit shows what is wrong on each page; the journeys show where it costs you. Find your top entry pages, usually the homepage, a few blog posts and the paid landing pages, then trace the two or three most common paths from entry to a primary conversion, such as paid ad, landing page, pricing, demo. GA4’s funnel and path explorations build these from the events you fixed in Step 2.

Look for drop-offs where you would expect interest to continue: a pricing page most visitors leave, a demo form many open and few send. Then split each journey by device and by channel. Mobile and paid visitors often behave differently from desktop and organic ones, and an average hides it.

The numbers tell you where; recordings tell you why. Session recordings and heatmaps show what people did on the page that lost them, and Microsoft Clarity provides both free^[9]. Mask what visitors type, watch 20 or 30 recordings of people who left the page you are worried about, and write a hypothesis for each drop-off: what happens, why you think so, and what would show you are right.

Steps 3 and 4 will give you more ideas than you can ship. Score each with the ICE method from Sean Ellis and Morgan Brown’s Hacking Growth: impact, how much it could move a primary conversion; confidence, how sure you are, given what Steps 2 to 4 showed; and ease, how little time and help it needs^[10]. I score each from 1 to 10 and add them up. The score is a way to argue in the open, not a measurement.

Then sort the ranked list into three lanes:

• Quick wins (days). Small, low-risk fixes: a clearer headline or call to action, fewer form fields, a missing trust signal.
• Core improvements (weeks). Layout or messaging changes to key pages: a reworked hero, a rebuilt demo page, a clearer pricing page.
• Experiments (weeks to months). Changes you are unsure of, tested only where your traffic can answer (Step 8).

Give each item an owner and a week. The list is your roadmap for the next six to twelve weeks; ship it in order, a few changes at a time, so you can tell which one moved the numbers.

Quick wins are changes small enough to ship within days. Four kinds come up on almost every B2B site.

Say who it is for, what problem it solves and what the buyer gets

Every key page should answer those three questions on its first screen, in the buyer’s words rather than yours. Run the five-second test again after the rewrite, with new people from the same roles.

Make every call to action say what happens next

Replace “Submit” and “Learn more” with what the click does: “Book a 30-minute demo”, “Start a free trial”, “See pricing for your team”. Keep one main call to action per page, and give visitors earlier in their search one quieter alternative, such as a sample report.

Ask only for what sales uses

Remove every field sales does not use to qualify or route the lead; company size and industry can usually be looked up from a work email address. Accept phone numbers in any format, show errors next to the field, and say beside the button what happens after the visitor sends the form. In a lab study of forms taken from real company websites, versions rebuilt to 20 published guidelines on labels, input formats and error messages were completed faster, with fewer failed submissions, and satisfied people more^[11]. Split a form into several steps only if each step is short and you can measure the result.

Check the same form for accessibility: visible labels, errors that say what is wrong, and no repeated entry. The checklist is in Accessible Marketing for Startups (Step 3).

Add the proof you already have, and clear distractions

If a key page has no proof, add what you can stand behind: customer logos you have permission to use, a short quote with a name and a job title, a review rating, a security badge you have earned. Then remove competing calls to action from key pages and cut the menu down to the journeys that matter.

Core improvements rebuild the few pages where buyers decide: the demo, trial or contact page, the pricing page, the core product or solution page, and the homepage hero and the sections under it. Build each in the order a buyer’s doubts arrive: the problem, your solution, how it works, proof, then the call to action. On the way, give it:

• A first screen that states the value, who it is for and the main call to action.
• Use cases: short, concrete examples by type of customer.
• Proof where the doubt arises: the integration list next to the form, the customer quote next to the price.
• Risk reducers: trial terms, what a pilot involves, how data is stored and secured.
• One clear next step, plus a quieter option for visitors earlier in their search.

The pricing page. For four years running, transparent pricing has been the top item on technology buyers’ wish list for vendors in TrustRadius’s annual survey, most recently of 1,862 buyers in January 2026; the same survey found that 74% of buyers use reviews to inform a purchase^[12]. If your price depends on the deal, show a starting price or what the price depends on, and what each plan includes, and link to your reviews if you have them. The Pricing Page Teardown scores your page and up to three competitors’ on 21 checks.

The demo page. Say what the demo covers, who the buyer will meet and how long it takes, and let them pick a time as soon as they send the form rather than wait for an email. What happens after the form, and how fast, is the subject of the Lead Management Playbook.

Most changes don’t help. Ron Kohavi and Stefan Thomke report that at Microsoft about a third of the ideas tested improved the metric they targeted, a third made no difference and a third made it worse^[13]. That is the case for testing. The catch on a B2B site is traffic.

A test needs enough visitors to tell a real difference from noise, and the number climbs steeply as the conversion rate and the lift you hope to detect get smaller. Kohavi and Thomke put the entry point at “at least a few thousand daily active users”, more than most seed-stage B2B sites have. Before you build a test, put your traffic, baseline rate and the smallest lift worth having into the A/B Test Significance Calculator. If it needs more than a quarter, don’t run it. Instead:

• Test a busier step. Clicks on the main call to action happen far more often than demo requests.
• Test a bolder change. A new hero can show a large effect with far fewer visitors than a reworded button needs to show a small one.
• Ship and measure. For changes you would make anyway, set the win line in advance and compare the weeks after with the weeks before, as Example Co. did in Step 6.
• Learn from people. Five-second tests, recordings and sales-call notes need no traffic.

When a test is worth running, write a one-page brief before anyone builds it: the hypothesis, the control and one variant (more only if traffic allows), one primary metric, the sample size, a run time in whole weeks, and the rule for calling a winner, a loser or no result. Then leave it alone until the stop date. If you check every day and stop when the result looks significant, the p-values and confidence intervals become, in the words of the researchers who studied it, “wholly unreliable”^[14].

Gains fade unless someone keeps watching and the next change can’t quietly undo them. Three things keep them.

A dashboard with four views

Traffic by channel and device; conversion rates sitewide and for each key page; the funnel from visit to lead, qualified opportunity and won deal, joined to the CRM; and the change and experiment log, with what changed, when, and what happened. GA4 with Looker Studio, or your own BI tool, is enough.

A monthly review and a quarterly deep-dive

Spend 30 to 45 minutes each month on the numbers and their trend, the results of recent changes and tests, any channel or device falling behind, and the next one to three changes to start or stop. Each quarter, spend 60 to 90 minutes reading the numbers by segment (ideal customers against the rest, region, industry) and by offer (demo, trial, content), and reset the roadmap to the company’s current goals.

Rules that protect the gains

Write down who can edit copy, change layouts, change tracking and launch tests; a change process (brief, review, build, measure, record) with a rule for when a change should be a test; and standards for UTM tags, event names, required and optional form fields, and lead routing.

What changes by stage?

What mistakes should I avoid?

• Reading one “conversions” number. Count primary and secondary conversions separately (Step 1).
• Trusting the analytics count. Check form events against the leads your CRM created (Step 2).
• Redesigning everything at once. Ship in ranked order, so you can tell what worked (Step 5).
• Running tests your traffic can’t answer. Size every test before you build it (Step 8).
• Judging by demo requests alone. Follow each change through to qualified opportunities (Step 9).