What is Incremental Lift? Formula and Control Groups

Incremental lift is the part of a campaign result that would not have happened without the campaign. A control group that receives nothing is how you measure it.

Updated 8 November 2026 · 2 min read · Gamble Grip

Incremental lift is the extra result a campaign caused: the gap between players who were contacted and a comparable group who were not. A raw count of players who deposited after a message includes players who would have deposited anyway. Only the gap to the control group belongs to the campaign.

Blueshift describes holdout testing the same way. One group receives the messages, the control group is withheld, and the difference in their results is attributed to the campaign.

Formula

Absolute lift = treated rate − control rate
Relative lift = (treated rate − control rate) ÷ control rate
Incremental NGR = (NGR per treated player − NGR per control player) × treated players
Net incremental NGR = incremental NGR − campaign cost

Both groups come from one frozen audience, are assigned at random and are measured over the same window.

Example. A frozen list holds 10,000 dormant players. 1,000 go to the control group and 9,000 are treated. In 30 days 900 treated players deposit (10%) and 60 control players deposit (6%). Absolute lift is 4 points. Relative lift is 67%. NGR per player is €6 treated and €3 control, so incremental NGR is (6 − 3) × 9,000 = €27,000. The campaign costs €9,000, so net incremental NGR is €18,000.

Where the reported NGR really came from

Example. Treated group, 30 days.

How operators use it

Teams use lift to decide which campaigns to repeat, which channels to pay for and where to cut bonus spend. Lift also sets the fair base for a performance fee: a fee tied to incremental NGR pays for the effect and not for players who returned alone.

A persistent global holdout, a small group that no programme contacts, keeps measurement honest when several programmes run at once. Gutierrez and Gerardy define uplift modelling as the set of techniques that model the incremental impact of an action on a customer outcome. You do not need such a model to start. A random holdout is enough.

Players who are self-excluded, opted out or flagged under safer gambling rules stay out of both groups.

Common mistakes

Using all returning players as the result. This credits the campaign with natural returns.
Choosing the control group by hand. Assign at random from the same list.
A control group that other campaigns still contact.
Measuring GGR. Use NGR so bonus cost comes out.
Stopping at 30 days. A bonus can pull a deposit forward without adding one, so check a longer window.
Reading a small gap on a small control group as a result.

Reactivation rate is the most common metric tested this way. The guide to player reactivation campaigns shows the full holdout set-up, and player churn in iGaming covers holdouts for at-risk players.

CALCULATOR

Calculate returns for your player base

Pick a market, a campaign goal and the number of players. The calculator models NGR, ROMI and payback over 12 months.

Open the calculator

LIFECYCLE FACTORY

We run your first campaign with a control group

One player list, one campaign and a control group. Your CRM stays in place.

Discuss a campaign

Frequently asked questions

What is incremental lift?

Incremental lift is the difference in outcome between players who received a campaign and a comparable control group that did not. It separates what the campaign caused from what would have happened anyway.

What is the incremental lift formula?

Absolute lift is the treated rate minus the control rate. Relative lift is that difference divided by the control rate. Incremental NGR is the NGR gap per player multiplied by the number of treated players.

How large should a control group be?

Large enough that the difference is larger than random noise. Small bases need a larger share of players held out. Test the gap with a significance test or a confidence interval.

Can a before-and-after comparison replace a control group?

It is a weaker substitute. Seasons, events and other campaigns change results over time. Label any such estimate as an estimate.