The discount is the least useful thing a promo code does. The useful thing is that it tells you which partner, which post and which channel actually produced a sale.
One code per source
Give every influencer, every media partner and every co-host their own code. At the end of the campaign you will know exactly who delivered and who merely posted. That knowledge is worth more than the margin the codes cost you.
Keep public codes rare and time-boxed
A code that anyone can find, that works indefinitely, is just a lower price with extra steps. If it is public, give it a hard expiry and a quantity cap.
Codes that do not discount
An access code that unlocks a hidden ticket type is often more useful than money off. Use it for a members' presale or a sponsor's guest allocation, where the value is exclusivity rather than price.
Watch the stacking
Decide explicitly whether a promo code can combine with an early bird price. Usually it should not, or you will discover you sold a third of the room at half price.
Name codes so you can read the report
Use the partner's name, not a clever word. TOLA20 tells you who delivered at a glance; SUMMER tells you nothing three months later when you are deciding who to work with again. Keep the format consistent across a campaign so the report sorts sensibly.
Cap every code
A redemption limit protects you when a code escapes into a group chat with four thousand members. Set it slightly above what you expect the partner to deliver, so a genuine success is not blocked but a leak is contained.
Recovery codes are the highest-value use
Send a time-limited code to people who reached checkout and did not complete, or who attended last time and have not bought yet. These are warm buyers with a specific, identifiable hesitation, and a small, clearly temporary discount aimed only at them converts far better than any public offer.
Review the report afterwards
Redemptions per code, revenue per code, and which codes were shared beyond their intended audience. Take that into next campaign's partner list.