Background
The operator was not in trouble. Marketing spend was efficient, CRM was running multiple lifecycle campaigns, retention was healthy, and the brand was performing well across its three regulated markets. But internal benchmarks against competitors suggested the registration funnel was underperforming. Specifically, the funnel from landing-page arrival to completed registration was converting in the low single digits where comparable operators in the same markets were doing two or three percentage points better, and the funnel from completed registration to first deposit was running at sixty-one per cent against an industry benchmark closer to seventy.
The CMO suspected the issue was on-site rather than upstream - paid acquisition was bringing the right traffic, the offer was competitive, and the brand was strong - but nobody on the team had the time to do a proper systematic audit of the journey. The work had been done piecemeal over years, with various people optimising various pieces, and the cumulative result was a flow that nobody owned end-to-end and that nobody had stress-tested in twelve months.
What the diagnostic surfaced
The first three weeks were a disciplined audit of every step in the journey from paid-acquisition click to first deposit. Heatmap and session-recording tools captured user behaviour. Funnel analytics quantified the drop-off at each step. Twenty mobile and twenty desktop user tests put real users through the flow with think-aloud commentary. The audit produced thirty-one specific friction points across the journey, of which fourteen were judged material enough to be in scope for the engagement.
The biggest losses were not where the team had expected them. The CMO had assumed the registration form was the leak. It was, but only the third-largest of the leaks. The largest leak was earlier in the journey - the landing pages were producing a measurable drop-off at the moment of clicking the primary CTA, because the CTA opened a registration modal that loaded slowly and that on slower mobile connections appeared to do nothing for one to two seconds. Users were tapping again, getting confused, and bouncing. The second-largest leak was in the registration form itself - eleven required fields, three of which were not legally required at registration but had been added because the data team wanted them, and one of which (postcode) was failing on certain mobile autofill setups because the field validation was rejecting valid formats. The third-largest was on the deposit page, where the payment method selector defaulted to the operator's preferred method rather than the player's previous method or the most popular method in the player's geography.
The approach
Phase 1: Page-load and CTA responsiveness
The registration modal was rebuilt to render instantly on tap with skeleton loaders for any content that needed to fetch from the server. The CTA itself was given visual feedback on tap so users had unambiguous confirmation that their action had registered. Largest contentful paint on the landing pages was pulled from an average of 3.4 seconds to under 2.0 seconds through image optimisation and critical CSS work. The single-largest behaviour change was the disappearance of the "double-tap then bounce" pattern that had been responsible for an estimated five per cent of total CTA-click loss.
Phase 2: Registration form reduction
Eleven required fields became five. The three fields that were not legally required at registration were moved to a profile-completion step that the player completed during their first session, rewarded with a small game-credit incentive that produced higher completion than the at-registration approach had achieved. The postcode autofill issue was fixed by accepting any of three common formatting variants and validating server-side rather than rejecting client-side. Email and password requirements were simplified, with the password complexity rules relaxed to industry-standard minimums and the email-confirmation field removed in favour of email-correction logic that prompted users only when the address looked typo'd. The shortened form lifted form-completion rates by an estimated nine per cent on its own.
Phase 3: Progressive KYC instead of upfront KYC
The operator had previously required full KYC documentation upload before allowing first deposit. The legal interpretation that drove this was correct under the operator's licence framework but stricter than necessary. A revised approach allowed deposit and first session play before full KYC, with the documentation requested at the player's first attempted withdrawal. This is standard practice in the operator's primary market and required only minor compliance documentation updates. The change removed a substantial pre-deposit friction point and shifted the operator's NDP-to-FTD conversion measurably upward without any change in compliance outcomes - KYC completion rates ended up slightly higher under the new model than under the old, because players had a stronger reason to complete it (their first withdrawal was waiting for it) than they had under the previous model (they were complying with a request before they had even funded their account).
Phase 4: Deposit page UX
The deposit page was rebuilt around three principles. Most popular payment methods in the player's detected geography surfaced first with clear visual hierarchy. The operator's preferred method was no longer the default; the player's most-likely method based on geography and device was. Trust signals - payment processor logos, security badges, the operator's licence indicator - were given more prominence. The deposit amount input accepted both keyboard input and tap-to-select on common amounts, with the tap-to-select options sized larger for thumb interaction on mobile. The form validation was made forgiving, accepting deposits with currency symbols, spaces, and various common formatting quirks rather than rejecting them.
Phase 5: Mobile-first redesign of the entire flow
Throughout the engagement, every change was mobile-first. The operator's traffic mix was eighty-two per cent mobile, but the previous design work had been done with desktop as the primary canvas and mobile as an adapted version. Reversing that priority - designing for mobile first and adapting to desktop - surfaced friction points that had been invisible in desktop-first design reviews and that compounded the impact of the other changes.
The result
Four months from project start, comparing matched cohorts of paid-acquisition traffic with consistent campaign settings:
Registration rate from landing-page arrival to completed registration: up seventeen per cent against the pre-engagement baseline.
NDP-to-FTD conversion from completed registration to first deposit: up nine per cent.
Mobile-only registration rate: up twenty-three per cent. The mobile flow had been carrying a disproportionate share of the friction.
Marketing spend: unchanged. Every percentage point of the lift came from on-site improvements, not from increased acquisition investment.
The compounding effect through to retention metrics took longer to read but trended positive. Day-30 retention on the post-engagement cohort was running about two points above the pre-engagement cohort by month four, and day-90 retention was running about a point above. The hypothesis is that lower-friction registration produces more committed players because the players who clear the (now-lower) friction bar are still entertainment-motivated rather than friction-tolerant; that hypothesis remains under measurement but the early signal is in the right direction.
What carried over
Three observations that now anchor my approach to on-site journey work generally. First, friction in the registration funnel often does not live where teams intuitively believe it lives; it lives where nobody has looked recently because nobody owns end-to-end responsibility for the flow. Systematic audit beats intuition. Second, every legally non-required form field is a tax on conversion; operators routinely add fields because internal teams want the data, without measuring the conversion cost; that cost is almost always larger than the data utility. Third, mobile-first design is non-negotiable in 2026; operators still designing desktop-first and adapting to mobile are leaving meaningful conversion on the table by default.
The pattern works in any operator with mature acquisition where the on-site journey has not been audited end-to-end in the last twelve to eighteen months. It does not work for operators with weak product-market fit; in those cases, journey work polishes a pipeline that the wrong players are trying to enter, and the lift is smaller and less durable.