Referral programmes are attractive because the cost is paid only on a result. They work extremely well in some categories and produce almost nothing in others, and the difference is predictable in advance.
The programme amplifies conversation, it does not create it
A referral scheme rewards an action customers are already inclined to take, making it slightly more likely and slightly more frequent.
Where nobody talks about the product, there is no behaviour to amplify, and the incentive is not large enough to start a conversation that would not otherwise happen.
This is why the first question is whether customers mention the product unprompted, not whether the reward is generous.
Recommendation carries a social cost
Recommending something puts the recommender's judgement at risk, since a poor experience reflects on them as well as on the company.
That cost rises with the price and consequence of the purchase, which is why casual referral works better for low-risk products than for major commitments.
Rewards can offset the effort of referring and cannot offset the reputational risk, which is why increasing the payment often fails to increase participation.
Some categories are actively private
Purchases connected to money trouble, health, personal appearance or professional insecurity are ones customers may value highly and discuss with nobody.
Satisfaction in those categories can be very high while referral volume stays near zero, and the two facts are entirely consistent.
Companies that read low referral rates as a product problem in these markets are usually misreading a category characteristic.
The reward changes what the referral means
Paying the referrer converts an endorsement into a transaction, and the recipient's response depends on whether they know a payment was involved.
Schemes that reward both sides tend to be received better, because the referrer is passing on a benefit rather than collecting one.
Where the reward is large enough to motivate people who do not use the product, the scheme attracts low-quality volume and the economics deteriorate quickly.
Timing determines whether anyone acts
Customers are most willing to recommend immediately after a good experience, and that window is short.
Programmes that ask at an arbitrary time, such as in a monthly newsletter, reach people at the wrong moment and perform accordingly.
The mechanics that matter most are therefore about placement rather than reward size, which is the opposite of where most programme design effort goes.