Write-through vs write-behind caching: understanding the consistency trade-offs

We're trying to improve the consistency and performance of an inventory system, and caching is a big part of that. We're currently debating write-through versus write-behind caching strategies and the consistency trade-offs involved. For an inventory system, accuracy is paramount; we can't accidentally show wrong stock levels. Write-through caching ensures strong consistency because the write goes directly to the cache and the database in the same transaction. The downside is increased write latency. Write-behind caching improves write performance because the cache acknowledges the write immediately, and the data is asynchronously written to the database. However, this introduces a risk of data loss if the cache fails before the data is persisted. We've also seen issues with write-around, where updates bypass the cache, leading to cache miss storms after bulk updates. What's your preferred pattern for critical data where low latency reads and high consistency writes are both important?