Periodicity

Set consistent time interval for all project datasets e.g. Minute, Hour, Day, Week, Month, Quarter, Year.

Overview

Data periodicity defines the consistent time interval at which observations are recorded across all data sources in a project. This ensures alignment and coherence for modeling, analysis, and insights in Alviss AI. All datasets within a project must adhere to the same periodicity (e.g., daily, weekly, monthly), which is set during project creation and cannot be changed later.

For example, if sales are tracked weekly, every data point (e.g., media investments, brand metrics) must also aggregate to weekly intervals. This prevents mismatches that could lead to inaccurate attributions, simulations, or predictions. The Date field always represents the first day of the period, even if the value (e.g., total sales) spans the entire interval.

Supported Periodicities

The currently supported periodicity options are:

Minute
Hour
Day
Week
Month
Quarter
Year

Key Requirements

Coherent Intervals: All data sources must have exactly one observation per period for each unique combination of Country, Region, Product, Grouping, etc. Gaps or duplicates will cause upload errors or model inaccuracies.
Date Placement: The Date must fall on the first day of the period:
- Minute: The start of the minute (use timestamp if needed for precision).
- Hour: The start of the hour.
- Day: The exact date.
- Week: Monday (following ISO 8601 standard, where weeks start on Monday).
- Month: The 1st day of the month.
- Quarter: The 1st day of the starting month (e.g., January 1 for Q1, April 1 for Q2, July 1 for Q3, October 1 for Q4).
- Year: January 1.
Aggregation: If raw data is more granular (e.g., daily sales for a weekly project), aggregate values (e.g., sum units sold) and assign to the period's start date.

Alviss AI follows the ISO 8601 standard for weeks (starting on Monday), regardless of local calendar conventions (e.g., Sunday starts in some countries).

Example: Weekly Periodicity

Consider ISO Week 32 in 2022, starting Monday, August 8, and ending Sunday, August 14. Total sales for this week would be recorded with the date "2022-08-08" (the first day), even if sales occurred throughout the week.

┌────────────────────┐
│ August ▒▒▒▒▒▒ 2022 │
├──┬──┬──┬──┬──┬──┬──┤
│Su│Mo│Tu│We│Th│Fr│Sa│
├──┼──┼──┼──┼──┼──┼──┤
│▒▒│01│02│03│04│05│06│
├──╔══╗──┼──┼──┼──┼──┤
│07║08║09│10│11│12│13│ <--- Week 32 starts here (Monday, August 8)
├──╚══╝──┼──┼──┼──┼──┤
│14│15│16│17│18│19│20│
├──┼──┼──┼──┼──┼──┼──┤
│21│22│23│24│25│26│27│
├──┼──┼──┼──┼──┼──┼──┤
│28│29│30│31│▒▒│▒▒│▒▒│
└──┴──┴──┴──┴──┴──┴──┘

In a Sales file for this week:

Date: 2022-08-08
Value: Sum of units sold from August 8–14

This approach applies to all data types, ensuring the date references the full period.

Best Practices

Choose Periodicity Wisely: Select based on your data availability and business needs (e.g., weekly for fast-moving consumer goods, monthly for B2B). Finer granularity (e.g., daily) provides more detail but requires more data.
Data Preparation: Aggregate raw data externally (e.g., using tools like Python's pandas) to match the project's periodicity before upload.
Consistency Across Sources: Align all files (e.g., Media investments, Weather averages, ...) to the same intervals to avoid errors during dataset creation or modeling.
Time Zones and Holidays: Use UTC or consistent local time for dates. Account for holidays or irregular periods in Events if they disrupt standard intervals.
Validation: After upload, review in the Activities dashboard for gaps or misalignments. Test with a small dataset to confirm periodicity before full import.

Common Issues and Troubleshooting

Mismatched Dates: If dates don't align with period starts (e.g., weekly on Tuesday), uploads will fail. Adjust externally to the first day.
Gaps in Data: Missing periods can bias models, missing data will be interpolated and extrapolated according to the Interpolation and Extrapolation rules.
Granularity Mismatch: Uploading daily data to a weekly project will error; aggregate first.
ISO Week Conflicts: If your local week starts on Sunday, remap dates to ISO Mondays to comply.

For more on date formatting, see Date. If issues persist during upload, refer to Upload Data or contact support.