Which referral growth produces qualified participation?

A growing referral cohort can look unusually engaged because referrers choose people who already understand the product. That selection may be valuable recruitment, but it tells us little about what the referral mechanism itself changes. We would separate the people selected, the invitation delivered and the evidence contributed afterward.

The public DXAP points page describes referral benefits as a share of the referred trader's points and DXAP builder fees. A program with those incentives can influence whom a referrer approaches. A quality study should therefore start before the selected people sign up.

Save the candidate list first

For a proposed six-week study, ask 30 consenting referrers to nominate four eligible candidates each under a fixed eligibility definition. Freeze the 120-person list before assignment. Within each referrer, randomly select two candidates for an authorized invitation now and two for an invitation after the four-week measurement window. This delayed-invitation comparison requires permitted access timing; it cannot override an existing access entitlement.

Define qualified participation in advance as completing an interpretation exercise and submitting one independently reviewable case within 28 days. The exercise checks a documented boundary such as the distinction between a submitted order and a fill in Activity. A case may establish correct behavior while exposing an unclear explanation. Qualification depends on usable evidence, rather than trade count or reward claims.

In an illustrative comparison, 18 of 60 immediate candidates and six of 60 delayed candidates qualify through available access routes. The assignment difference is 30% minus 10%, or 20 percentage points. If 36 immediate candidates sign up, their conditional qualification rate is 18/36, or 50%. That second rate describes the recruited users. The 18/60 rate preserves the invitation experiment.

Some delayed candidates may hear about the product from another friend. Record that exposure and retain original assignment in the primary estimate. Estimate uncertainty with referrer grouping, since four nominees from one person can share experience and communication. Where spillovers are extensive, randomizing whole referral networks may provide a cleaner next design.

Selection deserves its own comparison

The randomized result applies to nominated candidates. Generalizing it to all eligible people requires evidence about nomination. Compare the frozen candidate list with a contemporaneous eligible population on prior agent experience, stated research interests and existing access. These are descriptive comparisons unless recruitment itself was randomized.

For example, a 70% prior-experience rate among nominees versus 25% in the broader population would help explain why a referral cohort produces sophisticated reports. It would also reveal a coverage gap: novice misunderstanding may remain underrepresented in the feedback queue.

A useful growth report consequently carries three counts: nominated candidates, assigned invitations and qualified contributors. It also shows the distinct issue families those contributors reveal. We can value a referrer's selection skill while reserving the causal claim for the randomized invitation contrast. That distinction guides whether the next investment belongs in recruitment, onboarding or the review task itself.