Skip to content

How to query for total number of inquiries by status (at the given time)? #9552

Open
@ivan-wolf

Description

@ivan-wolf

Problem

We're developing a system in which users submit inquiries (tickets, issues), for which the status is tracked (so very much like GitHub issues).
We store the following in the database:

  • Inquiries
  • Status changes for each inquiry

Now, the idea is to present a graph in the user interface which shows how many inquiries there were in system for each status at a specific time. Here's an example:

Image

For this purpose, we built a table with a snapshot for each day (which represents the finest granularity that is required) and inquiry, which is exposed as cube inquiry_snapshots.

day inquiry status
2025-01-01 1 open
2025-01-01 2 open
2025-01-02 1 open
2025-01-02 2 resolved

With the following query, the results are as expected:

{
  "measures": [
    "inquiry_snapshots.count"
  ],
  "dimensions": [
    "inquiry_snapshots.status"
  ],
  "timeDimensions": [
    {
      "dimension": "inquiry_snapshots.day",
      "dateRange": [
        "2025-01-01",
        "2025-01-31"
      ],
      "granularity": "day"
    }
  ]
}

Our problem is how to accomplish the same with higher granularities (week, month, etc.), as now the inquiry could've been in more than one status over the span of each time frame (as seen in the above example: inquiry 2), but only the last status (in each time frame) is relevant.

Related schema

cubes:
  - name: inquiry_snapshots
    sql_table: inquiry_snapshots

    dimensions:
      - name: day
        sql: day
        type: time
        primary_key: true

      - name: inquiry
        sql: inquiry
        type: string
        primary_key: true

      - name: status
        sql: status
        type: string

    measures:
      - name: count
        sql: inquiry
        type: count_distinct

Practical example

Granularity day ✅:
Image

Granularity week ❌:
Image

As we can see, in week 10 there were many status changes (to resolved), which causes inflated numbers, because a single inquiry is counted more than once (i.e. once for each status it was in during the given week).

Did someone face a similar challenge and could point us in the right direction about how to model this? Any help would be much appreciated!

Metadata

Metadata

Assignees

No one assigned

    Labels

    data modelingquestionThe issue is a question. Please use Stack Overflow for questions.

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions