How-to: Edit historical data

​

The need for editing historical data

1. Error correction: Mistakes in data entry or processing can lead to inaccuracies. Correcting these errors ensures the data reflects true events or values.
2. Compliance and regulations: Certain industries have strict regulations that may require updates to historical data to remain compliant, especially if the data impacts financial reporting or legal obligations.
3. Improved analysis: As new insights or methodologies emerge, historical data may need to be adjusted to ensure analyses are relevant and accurate, allowing for better decision-making.
4. Data quality enhancement: Over time, organizations may refine their data quality standards. Updating historical records can help align them with current standards.
5. Business strategy changes: Shifts in business strategy might necessitate alterations in historical data for better reporting or strategic analysis.

🚧 Warning Editing historical data in a database should be approached with caution, as it can affect data integrity and accuracy.

​

How to edit historical data

• If you used usage event APIs to send data to Zenskar, refer to the section Edit historical data using usage-event APIs.
• If you used data-source connectors, refer to the following sections conditionally:
  • Edit historical data when data source supports remote querying
  • Edit historical data when data source does not support remote querying

​

Edit historical data using usage-event APIs

​

Edit historical data when billable metrics are derived using the COUNT SQL aggregate function

{
  "data": {
    "Id": "string",
    "was_office_space_used": "string"
  },
  "timestamp": "timestamp",
  "customer_id": "string"
}

SELECT 
		COALESCE(COUNT(CASE WHEN data.was_office_space_used = 'Yes' THEN 1 ELSE null END)) AS "quantity"FROM 
    your_table
WHERE 
    DATE("timestamp") >= DATE({{start_date}}) AND
    DATE("timestamp") <= DATE({{end_date}}) AND
    "customer_id" = CAST({{customer.external_id}} AS String)

​

Breakdown of the SQL query

COALESCE(COUNT(CASE WHEN data.was_office_space_used = 'Yes' THEN 1 ELSE null END)) AS "quantity"

• COUNT: This function counts non-null values.
• CASE statement: checks if the column data.was_office_space_used has the value Yes. If it does, it returns 1; otherwise, it returns null.
• COALESCE: this function returns the first non-null value in the list. If there are no instances where data.was_office_space_used is Yes, COUNT would return 0, and COALESCE ensures that it is treated as such.

FROM your_table

WHERE  
    DATE("timestamp") >= DATE({{start_date}}) AND  
    DATE("timestamp") \<= DATE({{end_date}}) AND  
    "customer_id" = CAST({{customer.external_id}} AS String)

• DATE("timestamp"): This converts the timestamp field to a date format, ensuring the comparison works with date values.
• DATE({{start_date}}) and DATE({{end_date}}): These placeholders represent the start and end dates for filtering.
• customer_id = CAST({{customer.external_id}} AS String): This checks if the customer_id in the table matches the provided external_id, which is also cast to a string to ensure type compatibility.

​

Method 1: modify the data schema by introducing a new data field

🔖 Note The examples given in this document are for reference only. The ideas given herein are guidelines and not rules. You must adapt the ideas given in this document as per your use case.

{
  "data": {
    "Id": "string",
    "was_office_space_used": "float",
    "deduct_from_count": "bool"
  },
  "timestamp": "timestamp",
  "customer_id": "string"
}

SELECT 
		COALESCE(COUNT(CASE WHEN data.was_office_space_used = 'Yes' THEN 1 ELSE null END)) - COUNT(NULLIF(data.deduct_from_count,'')) AS "quantity"
FROM 
    your_table
WHERE 
    DATE("timestamp") >= DATE({{start_date}}) AND
    DATE("timestamp") <= DATE({{end_date}}) AND
    "customer_id" = CAST({{customer.external_id}} AS String)

​

Breakdown of the SQL query

• COALESCE(...): This function returns the first non-null value in its list of arguments. Here, it is used to ensure that if the result of the subtraction is null, which can happen if there are no rows, it will return 0 instead.
• COUNT(data.was_office_space_used): This counts the number of non-null entries in the data.was_office_space_used column.
• COUNT(NULLIF(data.deduct_from_count, '')): The NULLIF function returns NULL if data.deduct_from_count is an empty string (”), effectively counting only non-empty entries. So, this counts the number of non-empty entries in data.deduct_from_count. The entire expression calculates the difference between the count of data.was_office_space_used and the count of non-empty data.deduct_from_count.

• DATE("timestamp") >= DATE({{start_date}}) AND DATE("timestamp") <= DATE({{end_date}}): this filters the results to include only those records where the timestamp is within the specified date range, defined by the variables {{start_date}} and {{end_date}}.
• "customer_id" = CAST({{customer.external_id}} AS String): This filters the results to include only rows where the customer_id matches the external_id, after converting external_id to a string.

​

Method 2: introduce a new usage event for deduction

{
  "data": {
    "Id": "string",
    "Premises_used": "string"
  },
  "timestamp": "timestamp",
  "customer_id": "string"
}

{
  "data": {
    "Id": "string",
    "Error": "string"
  },
  "timestamp": "timestamp",
  "customer_id": "string"
}

SELECT 
    (SELECT COUNT(data.was_office_space_used) FROM your_table) - 
    (SELECT COUNT(data.Error) FROM your_error_table) AS "quantity";

​

Edit historical data when billable metrics are derived using the SUM aggregate function

{
  "data": {
    "Id": "string",
    "Usage": "float"
  },
  "timestamp": "timestamp",
  "customer_id": "string"
}

SELECT 
    SUM(data.Usage) AS "quantity"
FROM 
    your_table
WHERE 
    DATE("timestamp") >= DATE({{start_date}}) AND
    DATE("timestamp") <= DATE({{end_date}}) AND
    "customer_id" = CAST({{customer.external_id}} AS String)

​

Edit historical data when billable metrics are derived using the MAX, AVG, MIN, and UNIQUE COUNT aggregate functions

​

Editing historical data using ROW_NUMBER() OVER (PARTITION BY) SQL function

​

Breakdown of ROW_NUMBER() OVER (PARTITION BY)

• ROW_NUMBER(): generates a unique number for each row in the result set, starting from 1 for the first row in each partition.
• OVER: specifies the window over which the function operates. It can include:
• PARTITION BY: defines the groups (partitions) within the data. Each partition is treated independently when generating row numbers.

WITH DeduplicatedReadings AS (
    SELECT 
        data.Id,
        data.Usage,
        timestamp,
        customer_id,
        ROW_NUMBER() OVER (PARTITION BY data.Id ORDER BY timestamp DESC) AS rn
    FROM 
        device_readings
)
SELECT 
 				data.Id,
        data.Usage,
        timestamp,
        customer_id,
FROM 
    DeduplicatedReadings
WHERE 
    rn = 1;

​

Edit historical data using data-source connectors

​

Edit historical data when data source supports remote querying

​

Edit historical data when data source does not support remote querying