Why retention budgets leak

A large share of every retention budget is spent on players who would have come back without it.

Take a standard reactivation campaign. It goes to everyone who has not deposited in 10 days, with a 20% reload offer. A good part of that group deposits on payday every month. The offer did not change what they did. It only made the deposit they were going to make cost more.

Say an operator sends that offer to 50,000 lapsed players and the report shows 9,000 returned. Marketing reads that as a strong result. Then someone adds a holdout of 5,000 players who got nothing, and 13% of them came back anyway. That is 6,500 returns in a group of 50,000. The campaign produced 2,500 extra players, not 9,000. Every cost and every bonus paid to the other 6,500 was a gift.

Without a holdout, nobody can say which number is true. Most retention reporting has no holdout.

The second leak is segmentation by recency alone. Recency is easy to pull from the data, so it becomes the audience rule. A player who lapsed 10 days ago after losing a month of salary and a player who lapsed after a quiet fortnight get the same message and the same offer. Value, risk and intent are all different, and the offer treats them as one group.

The third leak is the bonus-only cohort. Some players deposit only when an offer is live. They show up in reactivation reports as wins every time. Their net revenue after bonus cost is often close to zero, and sometimes below it. If your retention system rewards them every month, it teaches the rest of the base to wait for the offer too.

The fourth leak is the measurement window. A campaign judged on seven-day returns will look better than one judged on ninety-day net revenue, because a discount pulls deposits forward. A player who would have deposited on day 12 deposits on day 3 with a bonus attached. Week one looks great. Week two is quiet.

The fifth leak is channel cost. SMS, push and email each have a price, and each also has a fatigue cost. Players who stop opening messages stop being reachable at all. Nobody budgets for that, so nobody notices it until open rates fall below the level where the channel pays for itself.

Size the holdout with some care. It has to be random, drawn from the same audience at the same moment, and large enough that a few players either way do not move the result. For an audience of 50,000, a holdout of 5,000 is usually enough to see a difference of a couple of points in return rate. Smaller audiences need a larger share. Keep the holdout out of every other promotion running at the same time, or the comparison is contaminated.

The bonus-only cohort can be found with one query. List players whose deposits in the last six months all came within 48 hours of an offer being sent. Look at their net revenue after bonus cost. In most bases this is a surprisingly large group, and the result is rarely flattering.

A sixth leak sits outside commercial reporting entirely. In regulated markets, players showing signs of harm have to be taken out of promotional activity. If the retention platform does not receive that flag from the player protection system, offers go to people who should not get them. That is a compliance problem and also wasted spend. We cover it in the post on responsible gambling.

Put one person in charge of the retention P&L, with a single definition of success. Net revenue after bonus and channel cost, measured against a holdout, over a window long enough to catch pulled-forward deposits. Agree that window in advance, not after the results come in.

Set a standing holdout, between 5% and 10% of every campaign audience. It costs a small amount of revenue each month. It tells you what every other number in the report means.

Split the budget by incremental value. Fund the campaigns that produce players who would not have returned, and cut the ones that mostly pay people to do what they were already doing.

Decide who owns the suppression list. It should be one list, fed by player protection, fraud and finance, and applied to every channel without exceptions.

Questions worth putting to the team this week:

• What share of last quarter's retention spend went to campaigns with a holdout?
• What is net revenue per reactivated player after bonus cost, at 30 and 90 days?
• How many players received an offer in every month of the last six?
• How does the suppression list reach the CRM tool, and how often does it refresh?
• Which campaigns would we stop running if we could see only incremental results?