Article Contents:
Overview
The data description tables provide qualitative context for Hatched Analytics’ data feeds. They help users understand both the nature of the signals being captured and the degree of alignment between those signals and the KPIs they are mapped to.
The descriptions provide transparency into dataset quality and limitations. They help explain variations in backtest accuracy and give quant users a consistent framework for weighing confidence when incorporating the data into their models.
Two description files are available:
Index Descriptions
Each index group* is scored along two dimensions:
- Relevancy (0–5): Measures how closely the signal corresponds to a true transactional event.
- A score of 5 indicates that the signal is generated at the exact moment of purchase (e.g., checkout confirmation).
- Lower scores indicate less direct proxies (e.g., signals tied to shopping carts or shipments that may not reflect every transaction).
- Completeness (0–5): Measures how much of the company’s transaction flow is captured by the signal.
- A score of 5 indicates full coverage of transactions.
- Lower scores reflect blind spots (e.g., orders processed through third-party platforms like the App Store, or regions not covered by the data).
Each score is supplemented by a narrative explanation describing why the rating was assigned, highlighting known limitations and strengths.
*An index group may consist solely of the primary order/transaction index, or it may also include indices that break the signal down further by geography or brand.
Example (FIGS):
- Index: Order Index
- Relevancy: 5/5 – ID is generated at purchase.
- Completeness: 4/5 – App Store transactions are not included in the sequential ID stream.
KPI Descriptions
KPI descriptions assess how well the modelled KPI aligns with the indices.
- KPI Alignment (0–5)
- A score of 5 indicates the index directly and consistently measures the KPI (e.g., order counts when the company reports Number of Orders).
- Lower scores reflect partial or indirect alignment (e.g., when the KPI is Revenue, but the index captures Orders and average order value is not directly observed).
Again, each score includes a narrative explanation, clarifying whether limitations are structural (e.g., KPI definition differences) or data-related (e.g., incomplete flows).
Example (FIGS):
- KPI: Number of Orders
- Alignment: 4/5 – Index strongly reflects online order volume, but does not capture in-store purchases, a small segment of total orders.
Access
Both Index and KPI descriptions are provided in the SUPPLEMENTAL files of Snowflake and S3:
DESCRIPTIONS_INDEX– Relevancy, completeness, and narrative.
DESCRIPTIONS_KPI– KPI alignment and narrative.
Each row is tied to a ticker and index/KPI combination, ensuring descriptions can be joined directly to the core data feeds.
Made with Bullet