With a particularly skewed customer value curves – a broad industry benchmark is that the top 5% of users will deliver roughly 95% of revenue – iGaming brands stand to gain a significant improvement in ROI from adopting a value-based bidding strategy. However, the vertical also faces more challenges than most in doing so.
Here are the top 4 challenges limiting Gaming brands to inefficient CPA bidding strategies (and their solutions).
1. Long LTV Curves
LTV matures over time as players deposit and play on-site, fluctuating based on wins, losses and subsequent deposits. Most operators don’t see a positive return on players until months after acquisition, while typical pixel-based attribution windows are around 7 days. This disconnect means many advertisers rely on optimisation towards a first deposit or deposit value, which holds a limited correlation with customer LTV.
In this case, prediction is the solution, because it allows the modeler to accurately predict future value in accordance with early journey indicators.
2. LTV Customers are Tricky to Predict
Prediction itself is challenging when the business’ revenue model is built on outcomes of games of chance. Additionally, players rarely monetise in a linear fashion, they follow Eric Bradlow’s model of ‘clumpiness’. It can be hard for marketers and data analysts to spot causations between customer behaviours and LTV. Machine learning models will allow causation to be found and exploited.
3. Customer Value Distribution
Bidding algorithms work best when fed with high volumes of high-quality data at a consistent rate. Unfortunately, iGaming customer value patterns rarely follow this model. For example, the very highest-value customers are few in number, but these make up a high share of total revenue. A bidding algorithm with insufficient learning data inevitably ends up over and under-valuing customers at the extreme ends of the value distribution.
Modelling against large, historical datasets provides the volumes needed for reliability. Bidding against early journey proxy events provides the volumes needed for effective algorithmic bidding.
4. Slow Signals
The most effective VBB occurs when activated against real-time signals. Most LTV prediction takes weeks or months, and even immediately available predictions are not typically accessible as real-time bidding signals. These predictions need to be accurate, fast, and easily available for bidding activation in marketing platforms – and that requires technology. Real-time prediction and passback of a value signal are crucial to an optimised data and media strategy. In VBB, fast prediction trumps absolute precision.
VBB can have its challenges, but it’s a hugely valuable tool for performance marketers looking to optimise their bottom line. Making Science’s Gauss Smart Advertising solution is an AI-driven, predictive value modelling tool. It allows advertisers to bid in real time on predicted LTV value. It’s channel and vertical-agnostic and proven to drive 10-30% stronger ROI in activation.
To reach better bidding, reach out to our team of experts to learn more about effective bidding strategies or to kick-off your campaign!