Referral programmes are copied from companies where they worked, into companies where they do not. The variable that separates the two is whether the product's use is visible to others.
A referral needs an occasion, not just an incentive
Recommending something requires a moment where the subject arises naturally. Without that moment, a customer has to manufacture a conversation in order to collect a reward.
Most people will not do that, whatever the incentive, because raising a product unprompted carries a small social cost they are unwilling to pay.
So the constraint is not motivation but opportunity, and incentives cannot create opportunities that the product's use does not generate.
Visible products generate occasions continuously
Products used in front of others, or that require a second person to participate, produce conversations about themselves as a normal part of being used.
In those cases a referral scheme is adding a small reward to something that was already happening, which is why the results appear so strong.
The programme gets the credit, but the underlying driver is the product's visibility, and that is what fails to transfer when the scheme is copied elsewhere.
Private products need the occasion built into them
Where a product is used alone, the referral has to be prompted at a moment the customer is already in it, ideally just after something went well.
Timing matters more than the reward, because a request arriving after a good outcome is asking someone to share an experience they are currently having.
Requests sent on a schedule unrelated to usage arrive at arbitrary moments and are ignored, which is usually read as the incentive being too small.
Rewards can undermine the recommendation
A recommendation carries weight because the recommender is thought to have no interest in it. A visible payment removes exactly that quality.
Where the recipient learns a reward was involved, the referral becomes an advertisement, and it persuades approximately as well as one.
This is why schemes that reward the recipient rather than the referrer often perform better, since the referrer is passing on a benefit rather than earning one.
Measurement usually overstates the effect
Referral tracking captures customers who used a link, including many who would have arrived anyway through a conversation that needed no scheme.
Attributing all of that volume to the programme makes it look effective and makes the reward budget look justified, without establishing what was incremental.
Comparing acquisition in periods with and without the scheme is a cruder measure and answers the question the tracking cannot.