Targeting Cohorts

Internal, opt-in, then wider

Early stages target specific cohorts: employees (dogfooding), customers who opted into a beta, specific plans or regions, or accounts with friendly contacts. Exclude groups where risk is higher or the feature is not ready (regions with specific legal requirements, languages that failed evaluation, regulated customers) until they are ready. Targeting rules should be explicit, versioned and visible to support teams, so they know who has the feature.

Cohort rules for a beta stage, run

I ran this with Python 3 (scipy 1.18.1 where imported) on example numbers, not data from a real product. In the beta stage the feature is on for an employee and for an opted-in user in India, but off for an opted-in user in an excluded region and for a user who did not opt in.

users = [
    {"id": "u1", "email": "asha@ourco.com", "plan": "enterprise", "region": "IN", "beta_opt_in": False},
    {"id": "u2", "email": "raj@gmail.com", "plan": "free", "region": "IN", "beta_opt_in": True},
    {"id": "u3", "email": "li@corp.cn", "plan": "pro", "region": "CN", "beta_opt_in": True},
    {"id": "u4", "email": "sam@yahoo.com", "plan": "pro", "region": "US", "beta_opt_in": False},
]
stage = "beta"
rules = {
    "internal": lambda u: u["email"].endswith("@ourco.com"),
    "beta": lambda u: u["email"].endswith("@ourco.com") or (u["beta_opt_in"] and u["region"] in {"IN", "US"}),
}
for u in users:
    print(u["id"], u["email"], "->", "ON" if rules[stage](u) else "off")

Output:

u1 asha@ourco.com -> ON
u2 raj@gmail.com -> ON
u3 li@corp.cn -> off
u4 sam@yahoo.com -> off

Give support a lookup

Support staff should be able to see whether a given customer has the AI feature enabled when handling a complaint.

त्वरित जाँच: Who usually sees an AI feature first?

• Only regulated customers
• All users at once
• Internal employees (dogfooding)
• Nobody until it is perfect