PlayAIO's Joshua Gibbs: The problem is not how much data operators have, but what they choose to measure
Joshua Gibbs, Co-Founder and CEO of PlayAIO, examines how disconnected KPIs, poor retention and fragmented reporting can distort commercial decisions, and why AI cannot compensate for a lack of business-specific context.
Rising customer acquisition figures can conceal a more costly problem: players who leave before operators recover the expense of attracting them. A marketing campaign may appear to be underperforming when the real issue is a failed deposit journey. Both illustrate the risks of measuring individual departments without understanding how their performance affects the wider business.
As operators apply AI to their data, the question becomes whether additional technology can improve decisions when the underlying information and objectives remain poorly defined.
Joshua Gibbs, Co-Founder and CEO of PlayAIO, argues that operators need to look beyond acquisition targets and examine what happens throughout the player journey. In this interview with Gaming Eminence, he discusses how disconnected reporting can distort commercial decisions, where external market benchmarks provide useful context, and why AI models need a clear understanding of an operator's product and strategy.
His central argument is that collecting more data matters less than knowing what the business is trying to improve and why.
Gaming Eminence: Operators collect enormous amounts of data across their businesses, but having more data does not necessarily mean having a clearer view of what is happening. Where do you think businesses most commonly lose sight of the bigger picture?
Joshua Gibbs: "Every operator I speak to is collecting more data than they were three years ago. Very few of them would say they have a clearer view of their business than they did three years ago. That gap is the whole problem, and buying more tooling does not close it.
Most data warehouses are fragmented, and for good reason. Security and compliance demand it. The practical result is that the only people looking at metrics across the entire business tend to sit at C level. Everyone below that is looking at the metrics their own department has elected to define success by.
Take FTDs and new registered customers. Both are good metrics. Neither is wrong. But a team measured on them will optimise for them, and a business that optimises for acquisition alone stops asking what happens to a player on day two.
We have all seen where that ends. In new markets it becomes a race to the bottom. The US is the clearest example: operators bought volume at any cost, and some eventually could not afford to stay in the market at all.
What gets forgotten in that race is that a new player is a real individual, and they want to feel the product was built for them. If the only thing you measure is the first experience, the first experience is the only thing you will ever design for."
Gaming Eminence: Customer behaviour, product performance, payments, marketing and safer gambling can all generate their own sets of data. What important relationships or warning signals can be missed when those areas are analysed separately rather than together?
Joshua Gibbs: "A player's decision-making is part conscious and part unconscious. Depositing and withdrawing sit firmly in the conscious half. If your core payment flows have a problem, the player consciously registers that something about this site is not right. That is why so many negative reviews trace back to payments. The correlation between broken payment journeys and reviews calling a site a scam is heavy and consistent.
Now put marketing alongside that. You use AI to generate a personalised bonus for every player and drive them to deposit. If half of those deposits fail, that effort and that money was spent producing negative publicity. More app store reviews calling you a scam.
Product and marketing are usually treated as separate areas with separate reporting. Analysed separately, that campaign reads as a marketing underperformance. Analysed together, it is a payments defect that marketing spend amplified. Same data, completely different decision."
Gaming Eminence: How far can an operator realistically get by looking only at its own internal data? What can wider market behaviour add to that picture, and where does external benchmarking risk becoming misleading?
Joshua Gibbs: "External research is extremely valuable when it is used for the right job. Your internal data will tell you the majority of your problems and, with them, the majority of your solutions. External data is there for the macro factors.
The clearest use of it is the final check before you commit. You stand up a new operator, the platform is tested, everyone says go live. The question that should be asked before that button is pressed is whether the offering is actually competitive. Sports or casino, do we have what this market requires in order to compete? That check stops a business pushing a product that is not market ready.
It can also become a problem if you become too engrossed in the macro. The balance comes from choosing external KPIs that sit against your own strategy and your own USPs. If your business intends to offer the best bonuses in the market, then benchmark bonuses properly and hold yourself to it. You do not need to benchmark everything.
Gaming Eminence: Are there particular KPIs that you think operators can place too much confidence in when they are viewed in isolation? What needs to sit alongside them to give management a more accurate picture of performance?
Joshua Gibbs: "The easiest metrics to lose money on are FTDs and new registered customers viewed in isolation.
Registration does not guarantee revenue. A player who arrives and leaves within a day cost you money. Newly registered players only mean something when they are read against the player journey and the micro goals underneath it: getting a player from day one to day two, from day two to day ten, and onwards. Acquisition volume paired with days retained is a real number. Acquisition volume on its own is a vanity number that pays a bonus to someone."
Gaming Eminence: AI and machine learning are increasingly being applied to operator data, but not every data problem is necessarily an AI problem. Where do you think these technologies genuinely add value, and where is the underlying issue still more likely to be data quality, ownership or process?
Joshua Gibbs: "AI and machine learning genuinely add value for operators who already understand their data and their product's player journey. It is an excellent automation engine. What is often forgotten is how much context it needs before it understands anything worth acting on.
The models have largely been trained on public gambling data, which comes mainly from the larger listed organisations. Left unchecked, that produces strategies and assumptions built for a business that is not yours.
The way through is to keep the tasks small and manageable. The failure mode is handing over an entire database and asking for the solution. Two questions come before that. First, do you know this data set is correct? Second, does the model understand your product, your offering and your context? Without focus, you can spend more time narrowing the approach than a data engineer would have spent narrowing the tables in the first place.
Once you have focus, the narrative has to be your KPIs. What does good look like in your business? What is a bad metric that will always be bad, simply because of the market you operate in? That context is what makes AI useful, and most people underestimate the time and cost of reaching it.
So proceed with AI, but proceed with caution. You are paying for a model with access to a hundred billion parameters when you need one. If you are not careful, it will overcomplicate your solution rather than simplify it.
Gaming Eminence: If you look ahead a few years, what will distinguish an operator that is genuinely data-led from one that has simply accumulated more dashboards, models and AI tools?
Joshua Gibbs: "An operator that is truly data-led is one that leads with data. Everyone in that business, regardless of their area, understands what success looks like. There is no room for interpretation. This is the goal, these are the KPIs, now how do we get there?
AI and everything else are tools in the ecosystem. The businesses that get it right make it look seamless and simple. Everything rolls up into an alert and an automation, the system works with the system, and the controls and the strategy stay human-led.
The future looks bright, but it means moving away from hype trends. Less of what the industry says I should be doing, and more of what my own data tells me I should be focused on. No two operators are the same, and no two markets are the same, no matter how often that is claimed.
The operators that get there will not be the ones with the most dashboards. They will be the ones who can tell you, in a single sentence, what they are trying to move and why."
