Work with dates

Get trends and comparisons that cover exactly the dates you meant, counted by the right date and compared fairly.

Before you start

• Permissions: anyone who’s signed in can ask the Librarian. Building a trend or a chart in a project needs edit access to that project.
• Data you need: a table with a date or time column. The examples use Curio’s shopify_orders, where created_at is when an order was placed, and shipstation_shipments, with ship_date and delivered_date.
• Credits this uses: each question to the Librarian, or in a project, uses a credit.

Steps

1. Find out what dates the data covers

Before you ask about a period, ask what’s there: “What are the earliest and latest created_at dates in shopify_orders?”

Curio’s data runs from January 2024 to September 2025. Ask about anything later and there’s nothing to find.

2. Pick the date that matters

Which date does an order belong to? Many tables have more than one. An order has the day it was placed. A shipment has the day it shipped and the day it arrived. Sales by order date and sales by ship date can put the same order in different months.

Name the column you mean: “Shopify revenue by month of created_at”. If your team has already settled which date counts, write it down so nobody has to ask again.

3. Give exact dates

“Last 30 days” and “this quarter” count back from today, which can land where there’s no data. Say “January to September 2025” or “Q3 2025” instead. Asking good questions covers this and other wording.

If your year doesn’t start in January, say so, as in “by fiscal quarter, with the fiscal year starting February 1”. Put the rule in the Organization Glossary too.

4. Choose the grain

Grain	Good for	Needs at least
Day	The last few weeks, in detail	A few weeks of data
Week	Operations trends	A few months
Month	Business reporting	A year, to see a pattern
Quarter	Summaries for leadership	Two years

Weeks can start on Sunday or Monday, so say which: “weekly Shopify orders, with weeks starting Monday”.

5. Compare fair periods

Compare periods of the same length, and only complete ones. With Curio, compare January to September 2025 with January to September 2024, not all of 2024 with part of 2025.

For year over year by month, ask for the two years side by side: “Shopify revenue by month for 2024 and 2025, one column per year”.

6. Measure the time between dates

Durations answer questions totals can’t:

Average days between ship_date and delivered_date, by service level, for 2025.

Shipments with no delivered_date can’t be measured, so ask how many were left out. To smooth out a noisy daily number, ask for a rolling average, like “the 7-day moving average of daily Shopify orders”.

7. Show trends as charts

A trend reads better as a line than as a table. Ask in a project, like “Chart monthly Shopify revenue for 2024 and 2025 as a line”, and it adds a Visualization step.

To look ahead instead of back, ask for a forecast: “Forecast weekly Shopify revenue for 12 weeks.” See Forecasting.

8. Fix dates stored as text

On a column’s panel, a date column that shows Top values instead of a chart is stored as text. Mixed formats like 03/04/25 and 2025-04-03 put rows in the wrong month. Curio’s marketplace export writes its dates seven different ways. Clean up messy data shows how to fix it. For a spreadsheet, format the column as a date before you upload.

Check your work

• Show the query filters on the date column you meant, over the range you meant.
• Every period in the result is complete, or you know which one isn’t.
• Periods you compare cover the same number of days or months.
• A trend has one row per period, with no gaps you didn’t expect.

Troubleshooting

Nothing comes back for “this month”. The data may stop earlier. Ask for the date range first. Curio’s ends in September 2025.

The latest period looks low. It may not be complete. Leave it out, or compare the same days in each period.

Dates show a time, like 2025-09-30T00:00:00. Rows in a chat answer show raw values. The written reply formats them.

Rows land in the wrong month. The query may have used a different date column, or the dates may be text in mixed formats. Check Show the query, then the column’s panel.

Weekly numbers don’t match someone else’s. Your weeks may start on different days. Say which day your weeks start.

Next steps

• Aggregate data: totals, counts, averages and splits.
• Clean up messy data: fix dates and other values stored the wrong way.
• Asking good questions: word time windows so they’re read the way you meant.
• Forecasting: predict what comes next from a date and a number.