CRM budgets in iGaming have grown faster than the ability to prove they work. Most operators we audit can show that players who receive bonuses deposit mo
CRM budgets in iGaming have grown faster than the ability to prove they work. Most operators we audit can show that players who receive bonuses deposit more than those who do not, then assume the bonuses caused the deposits. That assumption is what incrementality testing was built to break. Without it, CRM teams keep spending on campaigns that would have happened anyway, and finance keeps approving budgets that are mostly redistributing existing revenue.
This article explains how to set up, run and interpret incrementality tests inside an iGaming CRM program in 2026. It is written for retention managers, CRM leads and analytics teams at casino and sportsbook operators who already use platforms like Optimove, Smartico, Solitics or Fast Track and want to move past last-touch reporting.
We will focus on practical mechanics: control groups, holdouts, lift calculation and the operational rules that protect the integrity of the test.
Why CRM teams overstate impact by default
When a player receives a free spin offer and then deposits, attribution tools typically credit the campaign with the deposit. The problem is that a significant share of those players, often forty to sixty percent in active casino segments, would have deposited regardless. That share is the natural behaviour of the segment, and it is what incrementality testing isolates.
If you do not measure it, you over-credit CRM and under-invest in real growth levers like product, payments and acquisition quality. Our [casino player LTV calculation formula](/article/casino-player-ltv-calculation-formula) goes deeper into how this distorts LTV reporting.
The core method: control versus exposed
The simplest incrementality test in CRM has two groups. The exposed group receives the campaign. The control group, randomly selected from the same eligible audience and held back from the campaign, receives nothing. After the campaign window, you compare deposits, gross gaming revenue and active days between the two groups. The difference is the incremental lift.
Control groups should typically be ten to twenty percent of the eligible audience. Smaller than ten percent and you lose statistical power. Larger than twenty percent and finance starts asking why you are deliberately not marketing to revenue.
Picking what to test first
Operators with no incrementality history should not start by testing everything. Start with the campaigns that consume the most budget or operational time. Usually that is weekly reload bonuses for active casino players, reactivation campaigns for dormant sportsbook users and VIP-tier free bet drops. Those three categories represent most CRM spend at a typical mid-size operator and are the easiest to clean up if the test shows weak incrementality.
Setting up the test inside your CRM platform
In Optimove, you can configure a control group at the strategy or campaign level. Smartico exposes holdouts through its segmentation engine. Solitics and Fast Track support similar functionality with slightly different UI. The mechanic is the same: at the moment the audience is built, the platform randomly assigns a portion to a "do not contact" bucket and tags those users for downstream analysis.
The two operational rules that matter most: the control group must be excluded from all touchpoints of the campaign, including the bonus, the email, the SMS and the in-app banner. And the control group must not be reused across overlapping campaigns within the same week, otherwise the contamination makes the lift unreadable.
Calculating lift correctly
Incremental lift is the difference in the target metric between exposed and control, divided by the control rate. If the exposed group has a deposit rate of twenty-two percent and the control group has eighteen percent, the absolute lift is four percentage points and the relative lift is around twenty-two percent.
You should report both. Absolute lift drives revenue. Relative lift drives strategic decisions. A campaign with twenty-two percent relative lift on a small audience may matter less than a campaign with eight percent lift on a large one.
Always include a confidence interval. Tools like a basic z-test for proportions or, better, your data team's preferred Bayesian framework will tell you whether the lift is real or noise. For most CRM tests, aim for ninety-five percent confidence and at least one thousand users per arm.
What to do when lift is negative
Negative lift is not a bug. It happens regularly in iGaming, especially with low-value reload offers sent to high-value players. The signal is that the campaign cannibalised behaviour that would have occurred at higher margin. Pause the campaign, redesign the offer or restrict the eligible audience.
We have seen multiple operators discover that their weekly cashback campaigns produced negative incremental GGR for VIP segments because the cashback was being paid against play that was already happening.
Building a campaign-level scorecard
After ninety days of consistent testing, build a scorecard that maps each campaign type to its incremental GGR, incremental deposits and bonus cost. Categories will naturally split into three: clearly incremental, clearly cannibalising, and ambiguous. The cannibalising campaigns are the budget you reclaim. The ambiguous ones need redesigned offers or tighter targeting.
This scorecard is what gives CRM credibility with finance. Without it, the conversation is about creative volume. With it, the conversation is about contribution.
Common failures in incrementality testing
The most common failure is contamination. A control user receives the campaign through another channel, usually push notifications, because the CRM platform exclusion was not configured across every touchpoint. Always validate exclusions across email, SMS, push and in-app before the test starts.
The second is short windows. A reload campaign tested over three days will look weaker than the same campaign tested over fourteen days, because deposit decisions are not always immediate. Pick a window that matches the campaign's intended decision horizon.
The third is reusing the same control users across multiple campaigns. After two or three weeks, those users are systematically under-marketed and start drifting in behaviour. Rotate control assignments monthly.
Where this connects to player value
Incrementality testing feeds directly into LTV models. If your CRM contribution is overstated by thirty percent, your LTV by acquisition channel is also overstated. Tightening incrementality improves the accuracy of acquisition decisions, which is why we treat CRM testing as part of the broader retention and acquisition system, not an isolated analytics exercise. See our [iGaming CRM platforms comparison 2026](/article/igaming-crm-platforms-comparison-2026) for a view on which tools handle holdouts well.
Quarterly cadence
Run a structured testing quarter at least twice a year. Pick six to ten campaigns, define hypotheses, run the tests with proper controls, then present the findings to leadership. The output is a refined CRM calendar where each campaign has a known lift expectation. Over time, that calendar becomes one of the most valuable internal artefacts the operator owns.
FAQs
**Is ten percent control group enough?**
For most large campaigns, yes. For VIP segments with small audiences, push to twenty percent and consider running the test over a longer window to gather enough events.
**What metric should I use as the primary outcome?**
For casino, incremental GGR is the strongest. For sportsbook, incremental net gaming revenue after bonus cost. Deposits are useful as a secondary metric but can be inflated by bonus stacking.
**Can I use uplift modelling instead of holdouts?**
Yes, once you have enough historical test data to train it. Uplift models are a refinement, not a replacement, and they still need periodic holdout validation.
**How long should a typical CRM incrementality test run?**
Seven to fourteen days for transactional campaigns, twenty-one to twenty-eight days for reactivation and lifecycle campaigns.
**What if my CRM platform does not support automatic control groups?**
Build them manually in your data warehouse before each campaign and pass the exclusion list to the platform. It is slower but works.
**How do I explain negative lift to stakeholders?**
Frame it as cost recovery. A campaign with negative lift was paying for behaviour that would have happened anyway. Pausing it returns margin to the business.
**Should affiliates and bonus abuse segments be excluded from tests?**
Yes. They distort lift in both directions. Remove them from both arms before randomisation.
How Basher helps
We design and operate CRM testing programs for casino and sportsbook operators across LatAm and Europe, including holdout setup, lift analysis and quarterly scorecards inside Optimove, Smartico and similar platforms. See our [CRM and retention services](/services/crm) or [contact us](/contact) to talk through a test plan.