TechnicalBeginner2 min read
Why does one small detail blow up your metrics bill?
Add or remove criteria on a metric and watch the number of rows to store, memory and cost jump orders of magnitude.
For: engineers and SREs · architectsPrerequisites: Know what a metric is. The page "Cardinality explained" is enough to get started.
An illustrative scenario: a developer adds the user ID to a metric to chase a bug. A few weeks later, the observability bill has multiplied while traffic has not moved. The CFO asks what happened. This simulator lets you show it with figures in hand. It also lets you check, before a release, that a new criterion will not send the bill off the rails.
A metric works like a tracking spreadsheet. Each criterion is a column (the server, the page requested, the outcome) and each combination of values opens a new row. In the tools’ vocabulary, columns are called labels, rows are active series and the number of rows is the cardinality. Rows do not add up, they multiply: that is why a single criterion too many costs so much. If the idea is new to you, the page Cardinality explained walks through it step by step, with a video.
Storage tools are always sized per row and many providers also bill per row. Prices change fast and vary by contract: this guide gives none. The price field is therefore empty. Enter the one from your provider’s dated price list to get a monthly cost.
Metric: HTTP latency histogram
| Keep | Labels | Distinct values |
|---|---|---|
service | ||
endpoint | ||
method | ||
status_code | ||
pod | ||
user_idunbounded |
Assumptions
Each label combination produces one series per bucket, plus the histogram’s _sum and _count series. Orders of magnitude. Memory per series depends heavily on the backend (Prometheus, VictoriaMetrics, SaaS offerings): replace the assumption with your own value. The price is empty by default: enter the one from your provider’s dated price list to get the cost.
Things to try
Section titled “Things to try”- Start from the defaults: 268,800 active series, in other words 268,800 rows in the spreadsheet, with about 525 MB of memory.
- Tick
user_id, the user’s identifier. Its values have no limit: every new user opens new rows. The row count reaches 5.4 billion and memory 10 TB. With your price entered, cost is multiplied by 20,000, just like the rows. - Untick
pod(withuser_idunticked). A pod is one instance of a service; counting per service rather than per instance divides rows by six (44,800) without losing what you actually look at. - Cut buckets from 12 to 8. A metric that measures durations sorts them into ranges, like the age brackets of a census: each range adds rows. Going down to 8 ranges gives 192,000 rows instead of 268,800, an often forgotten lever.
The rule to remember
Section titled “The rule to remember”An identifier with no limit (user, request, session, full web address) belongs in a trace or a log, never on a metric. To tie a metric to a specific request, use exemplars, sample requests attached to the metric. On Monday morning, I recommend listing the labels of your most expensive metrics and checking that each one has a bounded number of values.
For technical teams: the simulator’s formula
A metric’s cardinality is the product of the number of values of each label. For a classic histogram, multiply it by the number of buckets plus two, because each combination also produces a _sum and a _count series. The simulator applies this formula.
Going further: Cardinality explained, the business plane and the glossary.
Revised on 2 October 2026: prices removed; cost only shows with your price.