Decisions · Tiers and status
Should you model the cost of tier benefits before launch
Yes, model the cost of tier benefits before launch if you already hold earning and redemption data from a comparable programme or pilot. If you lack that data, do not pretend to model; launch a limited pilot first, then model against observed behaviour.
The decision
The decision is whether to commit to specific tier benefits before launch based on modelled redemption cost, or to delay commitment until real earning and redemption data exists. The underlying mechanism is a liability. Each tier benefit has a value, an eligible population, and an unknown redemption rate. The product of those three quantities is the expected cost to the programme. Modelling before launch forces you to name that product and use it to set qualification thresholds, point prices, or accrual rates. If you model without any observed data, you are not modelling cost; you are modelling assumptions about member behaviour. That distinction separates a decision tool from a spreadsheet that flatters your launch plan.
What it costs you
The recommended path, modelling with observed data before launch, costs engineering time and a short delay while you gather comparable data or run a controlled pilot. It also costs operating capacity because someone must collect redemption records, clean them, and maintain the model over time. The alternative, launching without a model, costs margin and member trust if actual redemption exceeds the budgeted rate. A model built on real data transfers some of that risk from the launch date to the design phase.
An illustrative calculation: suppose you plan a tier benefit of a hotel credit worth 100 dollars for 20,000 qualifying members. If 30 percent of members redeem, the benefit liability is 20,000 multiplied by 100 dollars multiplied by 0.30, which equals 600,000 dollars. If the actual redemption rate is 45 percent, the same benefit costs 20,000 multiplied by 100 dollars multiplied by 0.45, which equals 900,000 dollars. The 15 percentage point miss is an unbudgeted liability of 300,000 dollars. Pre-launch modelling with observed redemption data narrows that miss before you commit, and the engineering cost is typically a few weeks of one analyst's time, far less than 300,000 dollars.
How to decide
Apply two tests to your own programme. First, do you have earning and redemption data from a comparable tier, a previous programme, or a pilot that ran for at least one full redemption cycle? If yes, build a simple model: expected cost equals eligible members multiplied by benefit value multiplied by the observed redemption rate, and add an upper bound from the highest redemption period in your data. Second, can you state the maximum liability you would accept for each benefit? If you cannot state that maximum, modelling will not help because you have no budget line to compare against. If both tests pass, model before launch. If the first test fails, do not model from assumptions; launch a small pilot first and then model against that pilot's redemption behaviour before expanding the tier.
What would change our mind
We would reverse this recommendation if the tier benefit carries a contractual cap that limits total redemption cost to a known amount, for example a fixed number of hotel nights per member per year that cannot be exceeded. In that case the cost is driven by the cap, not by redemption behaviour, and pre-launch behavioural modelling adds little. You should still model the capped liability against your margin, but you no longer need redemption rate modelling. We would also reverse if the programme operates on a fully prepaid model where members buy tier benefits with cash at the time of earning them. Then the tier benefit cost is offset immediately and there is no future redemption liability to model before launch.