First Deposit

Amplitude flagged something that didn't add up: users cleared identity verification, the hardest part of onboarding, then never deposited.

Deposit is where a signup turns into revenue. Until a user funds an account, ESPN BET earns nothing from them no matter how far into onboarding they got, which makes a verified non-depositor acquisition spend with nothing behind it.

Company data put a number on the size of it: only a minority of users who create an account go on to make a deposit within a year, and of the ones who don't deposit right away, the vast majority never come back to do it at all.

These weren't casual signups. They'd already cleared the biggest friction point in the funnel, which made the drop-off both costly and confusing. Working alongside my mentor and a cross-functional group spanning design, customer experience, product, and marketing, I spent 4 months at theScore finding out why.

Design, customer experience, product, and marketing each had their own theory about this drop-off, and their own stake in the answer.

Before running anything, I held a kickoff with all four to hear what each team already knew, what they could contribute, and what they each needed out of the project, so the diagnosis wouldn't hit four separate blind spots later, once it was too late to fold them in. From there I kept a standing biweekly with CX specifically, since new support tickets kept surfacing complaints in real time and that channel was the fastest way to catch a new pattern before the diagnosis locked.

A single data source wasn't going to explain a drop-off this specific, so I built the diagnosis from four angles. Behavioral analytics in Amplitude mapped exactly where in the flow users stalled. A targeted survey, sent to 80,000 registered non-depositors (150 responded), surfaced their stated hesitations. A thematic read of support tickets caught the technical complaints users were already volunteering: deposit errors alone accounted for a significant share of all payment-related tickets that year. And competitive market research checked whether our offer was simply weaker than what users could get elsewhere.

Recruiting fresh test users to simulate the flow wouldn't have worked.

This group had lived through the real sign-up experience, including third-party failures on their bank's own site that ESPN BET's analytics could never see, because they happened off-platform. There was no substitute for hearing from people who'd actually been through it, and people who'd already walked away were the only group who could answer the question at all.

The risk was that reaching them could backfire. A large blast to users who'd already disengaged, with no ongoing stake in helping the company, ran a real risk of reading as spam and damaging the brand rather than informing it. That risk showed up immediately: the first send-out, framed around helping ESPN BET improve, got almost no response. Reworking the messaging to lead with the monetary incentive instead brought the response rate back up, to 150 replies out of 80,000.

The drop-off wasn't one problem. It was three, each hitting a different kind of user.

Users already loyal to a competing sportsbook were window-shopping, not stalling. Experienced bettors were markedly more likely than brand-new ones to say they'd compared us to other sportsbooks before deciding, and more likely again to say they simply preferred another book. Lining up our welcome offer against theirs made it easy to see why: our $100 bonus was the smallest of the four, despite asking users to risk one of the higher minimum bets to unlock it.

Users who did attempt to deposit sometimes hit real technical errors, made worse by vague messaging that turned a fixable bug into a dead end and a support ticket. Only a small fraction of users who hit an error were able to resolve it themselves; nearly everyone else gave up or ran into the same wall a second time.

Most of those errors were simply users trying to deposit less than our $10 minimum, a floor twice as high as two of our direct competitors.

A third group, simply cautious and new to depositing, hesitated over unclear terms: many said they'd intended to deposit but got distracted, and several specifically cited not understanding how withdrawals worked as the reason they changed their mind.

The competitor-loyalty finding was the one I expected pushback on.

Telling marketing that our own welcome offer was the weakest of the four was never going to land well if it read as "just raise the bonus": that's a budget conversation marketing couldn't win on our terms, competing dollar-for-dollar against books with deeper promotional budgets.

That reframed the conversation from "our offer is worse" to "our offer is aimed at the wrong lever," which was a much easier finding for marketing to act on.

Four months of research only matters if the room that has to act on it actually buys in. At the end, I brought design, CX, product, and marketing back together for a single shareout, and framed it less as a report and more as a walkthrough: here's the customer's actual journey from sign-up to deposit, here's each place they fell off, and here's why. I'd learned by then that this particular room moved on numbers, not narrative, so I led with the data at every stop rather than the emotional case, and paired every barrier with a recommendation rather than stopping at diagnosis.

The detail that actually moved the room wasn't one of the three barriers. It was proof the fix was worth funding: a majority of the very users who'd dropped off said they were still likely to come back and deposit on ESPN BET. That turned the diagnosis from a list of problems into a specific, fundable opportunity, and made the case for investing in the fixes far more convincing than the barriers on their own.

The first send-out to the 80,000 non-depositors almost sank the survey before it started.

Framed around helping ESPN BET improve, it landed as corporate spam to a group with no reason to open it, and the response rate showed it. I recovered by reworking the messaging to lead with the incentive instead, and that worked, but if I ran this again, I'd pilot the messaging on a few hundred users first, before committing the full list. A cheap test up front would have caught the framing problem before it cost real reach, instead of after.

I'd also be more careful about how much weight the return-intent finding carried. It was the number that unlocked the room, and it was self-reported intent from 150 people who had already walked away once. Intent isn't behavior, and I'd want a behavioral read alongside it before treating that number as the size of the opportunity.