Why Operators Know FTD Is Wrong and Still Can't Replace It
- Kevin Jones
- 8 hours ago
- 6 min read
Allan Stone, CEO of Intelitics, argues that FTD-led acquisition has persisted less because it works and more because replacing it is operationally expensive. The diagnosis is not particularly contested; what most operators lack is the first-party data pipeline, the attribution-to-CRM continuity, and the BI latency to act on it. Google, Meta and the major affiliate channels, which between them account for the bulk of most operators' marketing mix, are moving toward value-based optimisation regardless of whether operators are ready to feed them clean LTV signals. The performance gap between those who can and those who cannot is measurable at the campaign level, as Stone demonstrates with a $6M spend that produced $100M in deposits. What follows is an assessment of why the shift has stalled, where it is already producing measurable returns, and how long operators have before the platforms begin optimising without them.
Gaming Eminence: What is actually preventing operators from shifting away from FTD-led acquisition in practice?
Allan Stone: "The single biggest factor is data complexity. FTD is sticky because it's simple to track across systems. Every platform can report it. Every team can agree on the number. The moment you try to move beyond it, you're into per-player revenue calculations that are heavy to build, harder to maintain, and change as player behavior evolves. That complexity alone kills most initiatives before they start.
But there's a second layer that doesn't get talked about enough. FTD-led acquisition makes it reasonably straightforward to evaluate performance at the channel or partner level. Move beyond FTD, and you're immediately confronted with the reality that a single partner might be running six different channels across thousands of placements. Suddenly "how is this partner performing?" becomes "which cohorts, from which channels, across which placements, are actually working?" That is an infinitely harder question to answer. And most operators don't have the infrastructure to answer it. So they don't try.
The metric didn't survive because it's right. It survived because replacing it is genuinely hard."
Gaming Eminence: Where has predictive modelling genuinely changed acquisition decisions rather than just informed reporting?
AS: " There's a difference between knowing and acting. Knowing the predicted value of a partner or channel is useful. Knowing which specific segments inside that partner's traffic are pulling that value up or dragging it down is actionable. That's where decisions actually change.
The $6M spend, $100M deposits example is a good illustration. The aggregate number looked fine. The model let us get underneath it. We could identify, at a cohort level, which segments of a partner's traffic were trending toward low predicted value early, and cut them faster than we ever could waiting for 30-day or 90-day actuals. That's not reporting. That's a decision with a financial consequence.
The other side of it was equally important. Once you can pinpoint the underperforming segments, you can go back to the partner and show them exactly which parts of their traffic aren't working. That gives them something to act on. The ones who respond well to that conversation scale back the underperforming placements, their overall numbers improve, and they make more money. It stops being adversarial. The model creates alignment, not just insight."
Gaming Eminence: Where does the CPA and FTD affiliate model break when operators try to optimise for player value?
AS: "It breaks at the point where you try to hold a partner accountable for something they have no visibility into.
Here's a concrete example. One partner might have a top-five placement on their website, a dedicated brand review, an email newsletter, a Discord community, and a Reddit presence. Under current structures, an operator tracks that partner by a single ID and looks at their overall performance. Not the placement. Not the channel. Not the campaign. The whole thing, averaged together. That means the high-performing email newsletter is subsidising the low-performing Reddit traffic, and nobody knows it.
The reason operators default to this approach is mostly technical. Getting clean, granular tracking data from partner to operator across all those individual sources is genuinely hard. It requires a level of integration that most operators haven't built. So they default to aggregate partner performance, and the optimization potential inside that partnership stays invisible.
Revenue share addresses some of this in theory. In practice, the model doesn't hold up in markets like North America, where partners have real content costs and paid traffic is dominant. A long-tail revenue share doesn't cover the cost of running the business. What we're starting to see instead is a hybrid model: a lower upfront CPA for new depositors, paired with a smaller, direct incentive tied to subsequent player activity. A small payout when a referred player returns and places a bet, completes a session, or engages with the product again. Not tied to outcome. Tied to action. That structure starts to align the partner's incentive with player quality over time, without asking them to absorb all the risk on a pure revenue share basis."
Gaming Eminence: Using the $6M spend and $100M deposits example, what did the modelling change in practical terms?
AS: "The aggregate numbers told one story. The cohort-level data told a different one.
On channel mix: we identified partner segments that were consistently trending low on predicted LTV within 72 hours of acquisition. Those segments looked acceptable on an FTD basis. The model made them impossible to ignore. Budget came out of those segments fast, before 30-day actuals would have flagged them.
On partner selection: getting to cohort-level performance visibility changed the partner conversation entirely. Instead of a general discussion about whether a partner was performing, we could go to them with specifics. Here are the placements that are working. Here are the ones that aren't. Scale back the underperformers and your overall numbers improve. Partners who engaged with that conversation got better outcomes. Partners who couldn't provide that level of transparency got less budget.
On targeting: by understanding which cohorts were producing the expected post-acquisition engagement and LTV, we could better align incentives around the segments that were actually working. That meant fewer generic acquisition targets and more focus on the traffic profiles that had a track record of value.
The $100M headline is the output. The real value was in being able to see which part of the $6M was doing the work, and which part was carrying dead weight."
Gaming Eminence: What data or infrastructure gaps limit operators from implementing value-based bidding properly?
AS: "Three gaps. None of them are glamorous, but all three will kill the implementation if they're not solved.
First, no clean first-party data pipeline. Most operators have event-level data that's inconsistent across web and app, built across different tag management systems, with no single player ID that carries through from click to first deposit to lifetime activity and back into media channels. You cannot model a value you cannot observe. And you cannot observe value without identity continuity across the full player lifecycle. Most operators have pieces of this. Few have the closed loop.
Second, attribution and player lifecycle run on different timelines in different systems. Attribution historically closes at first deposit. CRM opens after first deposit. Those two systems rarely communicate in a way that allows player value to flow back into acquisition decisions. The result is two versions of truth owned by two different teams with two different tools. Connecting them requires cross-functional accountability that most organizations haven't structured.
Third, the BI layer is too slow. Operators typically have solid historical cohort data. What they don't have is the ability to surface a performance signal fast enough to act on it. Digital media campaigns need to be optimised within 72 hours at the outside. Waiting for a weekly data pull or a monthly cohort report means you're making decisions against data that's already stale. The model might be right. The latency kills the use case."
Gaming Eminence: Will the industry realistically move towards value-based acquisition, or will FTDs anchor the market?
AS: "It will move. But not because the industry collectively decides it should.
The forcing function isn't internal conviction. It's the platforms. Google, Meta, and affiliate and partnership channels together represent at least 70% of most operators' marketing mix. Every one of those platforms is already moving toward value-based optimization. They have to. Consumer attention is fragmenting. They need to justify ad spend in dollar terms that are relevant to the business, not vanity metrics that just prove a player registered. They're not going to wait for operators to figure out the LTV question. They'll start inferring it themselves unless you give them the actual data.
When the platforms shift the optimization model, operators who have the infrastructure to feed them real LTV signals will outperform operators who can't. That performance gap compounds fast. That is what actually moves the market.
The operators who move first will have a structural advantage for years. The long tail will hold on to FTDs until margin compression or consolidation forces the issue. But the direction is already set. The platforms decided that. The question for operators is just how long they want to wait before they catch up."
