Showback and Chargeback
Showing teams their costs, or actually billing them — two different mechanisms with different incentive effects and different failure modes.
Definition
- Showback reports each team's spend without moving money. Visibility and social pressure.
- Chargeback bills the cost to the team's own budget. Real financial consequence.
What each achieves
Showback is easy to introduce, creates no adversarial dynamics, and works well where teams genuinely want to be efficient and simply lacked visibility — which is more common than cynics expect. Its limit is that when a team must choose between shipping a feature and reducing cost, visibility alone rarely wins.
Chargeback creates a real incentive and makes cost a genuine engineering constraint. It also creates predictable pathologies:
- Gaming. Teams avoid tagged resources, use shared infrastructure to hide cost, or resist adopting a platform whose cost is charged to them.
- Under-investment in shared platforms, because nobody wants to fund something others use.
- Perverse reliability trade-offs, where a team removes a replica to reduce their bill and creates an outage that costs more than the saving.
- Substantial administrative overhead, arguing about allocation keys.
The pragmatic middle
Most organisations land on showback with accountability: teams see their spend and their unit cost, have a target, and explain variance — without money actually moving. This captures most of the incentive with little of the dysfunction.
Where chargeback is genuinely right: when internal teams are cost centres with real budgets, when usage varies enormously between teams, or when cost must be passed to external customers.
The critical refinement
Report unit cost, not just total. A team whose spend rose 40% while volume rose 60% has improved efficiency, and penalising them teaches the organisation that growth is punished. Unit cost is the only fair basis for comparison across teams of different sizes and growth rates.
Failure scenarios
- Chargeback without accurate allocation, so the numbers are disputed and the mechanism is discredited.
- Shared platform costs charged fully to early adopters, killing adoption.
- Cost targets without reliability targets, so teams optimise cost at the expense of availability.
- Reported monthly in arrears, too late to connect a change to its effect.
Interview question
"A team reduces its cloud bill by 30% by removing a standby replica, then has an outage. What was wrong with the incentive structure?"