When would you add a cache?

Direct answer

What a strong answer covers

How it works

A cache pays off when the same result is read far more often than it changes and a slightly stale answer is acceptable. Before adding one, name the source of truth, the key that identifies a result, how long staleness is tolerable, how invalidation happens on write, and what the service does on a miss or when the cache is down. Without those answers a cache turns a slow system into a wrong one. Often the better first step is fixing the query or adding an index.

Example

Cache a product catalogue with a TTL and invalidate on publication; do not cache a balance without an explicit consistency rule. A catalogue page cached for five minutes in Redis and refreshed on the publish event cut database load by 90 percent, while account balances stayed on the primary because a customer seeing a stale balance is a support ticket.

Common mistakes

• Caching before profiling and hiding a missing index behind a Redis dependency.
• Using a TTL as the only invalidation and accepting silently wrong data after writes.
• Letting a cache outage take the service down instead of degrading to the source of truth.

A useful answer structure

Answer from your own work.

Do not memorise a script: use only examples you can explain honestly in follow-up questions.