Decisions · Risk and fraud
Should you design loyalty controls for employee fraud
Yes, when your programme has staff who can adjust balances, issue points, or process redemptions, design employee fraud controls before launch. Treat point issuance as a cash-like liability and build separation of duties, approval limits, and audit trails into the core, not as a post-incident patch.
The decision
The decision is whether employee fraud controls are a first-class design requirement for a loyalty programme or an optional compliance item added later. The mechanism underneath is that loyalty points behave like a private currency with a delayed redemption cost. An employee who can create, move, or redeem points without a second person checking is effectively holding the keys to a cash drawer. Designing for employee fraud means putting a second pair of eyes on high-value point movements, limiting how many points any one employee can issue in a day, and keeping an immutable log of who changed what and why. It is not about suspecting every staff member. It is about making the programme's economics depend on controls rather than on individual honesty.
What it costs you
The recommended path costs engineering time, operating capacity, and some member friction when a legitimate point adjustment needs approval. It also carries a hidden cost: every control you design for employees is a control members will occasionally feel, such as a delayed redemption while a supervisor reviews a manual credit. You pay with margin if you set approval limits too low and with trust if you set them too high.
An illustrative calculation with a point liability of 1 cent per point: without a daily issuance cap, a single employee with access to a service tool could issue 50,000 points per day for 30 days before a dormant-account audit notices, a total of 1,500,000 points, or 15,000 dollars in redemption liability. A control that caps manual issuance at 5,000 points per day allows the same employee to issue at most 150,000 points over 30 days, or 1,500 dollars. The control costs one engineer for one month at 12,000 dollars plus 150 dollars per month for the audit script, which is 3,600 dollars over two years. Total two-year control cost is 15,600 dollars, slightly more than one uncontrolled month but far less than the uncontrolled path if the behaviour repeats. The real trade-off is not the one-off build cost; it is the permanent monthly overhead of running the control.
How to decide
Run a simple test on your own programme. List every role that can create points, transfer points, or approve a redemption without a second sign-off. For each role, ask two questions. Can this person move enough points in one week to create a liability greater than the annual salary of the person who would audit them? If the answer is yes, you need a preventive control, not just a detective one. Can this person alter or delete the audit log of their own actions? If the answer is yes, the control does not exist for that role. If you find three or more roles with both answers yes, employee fraud controls must be designed into the programme now. If you find none, you have already designed them, whether you called them fraud controls or not. The deciding factor is not the size of the programme; it is the concentration of point-moving authority in a single pair of hands.
What would change our mind
The recommendation flips if point issuance and redemption authority is fully automated with no manual override accessible to any employee. In that case, employee fraud is not a design problem because there is no employee action to control. It would also flip if the programme's points are non-transferable, expire within days, and cannot be redeemed for anything with resale value, because the incentive for employee fraud disappears. A third condition that changes the recommendation is a regulation that requires a specific external audit of point liability, because that shifts the control burden from preventive design to statutory oversight. Absent those conditions, a programme that gives employees point-moving power without designed controls is paying for honesty it has not verified.