SQL Reloaded: The WINDOW Clause in SQL Server 2022 (Part I: Simple Aggregations)

Between the many new features introduced in SQL Server 2022, Microsoft brings the WINDOW clause for defining the partitioning and ordering of a dataset which uses window functions. But before unveiling the change, let's take a walk down the memory lane.

In the early ages of SQL Server, many of descriptive statistics techniques were resuming in using the averages over a dataset that contained more or less complex logic. For example, to get the total and average of sales per Month, the query based on AdventureWorks database would look something like:

-- aggregated sales orders - detailed (SQL Server 2000+)
SELECT SOL.ProductId
, Year(SOH.OrderDate) [Year]
, Month(SOH.OrderDate) [Month]
, SUM(SOL.OrderQty) TotalQty
, AVG(SOL.OrderQty) AverageQty
FROM Sales.SalesOrderDetail SOL
     JOIN Sales.SalesOrderHeader SOH
	   ON SOL.SalesOrderId = SOH.SalesOrderId
WHERE SOL.ProductId IN (745)
  AND Year(SOH.OrderDate) = 2012
  AND Month(SOH.OrderDate) BETWEEN 1 AND 3
GROUP BY SOL.ProductId
, Year(SOH.OrderDate) 
, Month(SOH.OrderDate)

When possible, the base logic was encapsulated within a view, hiding query's complexity from the users, allowing also reuse and better maintenance, just to mention a few of the benefits of this approach:

-- encapsulating the logic into a view (SQL Server 2000+)
CREATE OR ALTER VIEW Sales.vSalesOrders
AS
-- sales orders details
SELECT SOL.SalesOrderId 
, SOL.ProductId
, Cast(SOH.OrderDate as Date) OrderDate
, Year(SOH.OrderDate) [Year]
, Month(SOH.OrderDate) [Month]
, Cast(SOL.OrderQty as decimal(18,2)) OrderQty
-- many more columns 
FROM Sales.SalesOrderDetail SOL
     JOIN Sales.SalesOrderHeader SOH
	   ON SOL.SalesOrderId = SOH.SalesOrderId
/* -- some constraints can be brought directly into the view
WHERE SOL.ProductId IN (745)
  AND Year(SOH.OrderDate) = 2012
  AND Month(SOH.OrderDate) BETWEEN 1 AND 3
*/

The above query now becomes:

 

-- aggregated sales orders - simplified (SQL Server 2000+)
SELECT SOL.ProductId
, SOL.[Year]
, SOl.[Month]
, SUM(SOL.OrderQty) TotalQty
, AVG(SOL.OrderQty) AverageQty
FROM Sales.vSalesOrders SOL
WHERE SOL.ProductId IN (745)
  AND SOL.[Year] = 2012
  AND SOL.[Month] BETWEEN 1 AND 3
GROUP BY SOL.ProductId
, SOL.[Year]
, SOl.[Month]

In many cases, performing this kind of aggregations was all the users wanted. However, quite often it's useful to look and the individual sales and consider them as part of the broader picture. For example, it would be interesting to know, what's the percentage of an individual value from the total (see OrderQtyPct) or any similar aggregate information. In SQL Server 2000 addressing such requirements would involve a join between the same view - one query with the detailed records, while the other contains the aggregated data:

-- sales orders from the totals (SQL Server 2000+)
SELECT SOL.*
, ASO.TotalQty
, Cast(CASE WHEN ASO.TotalQty <> 0 THEN 100*SOL.OrderQty/ASO.TotalQty ELSE 0 END as decimal(18,2)) OrderQtyPct
, ASO.AverageQty
FROM Sales.vSalesOrders SOL
     JOIN ( -- aggregated sales orders
		SELECT SOL.ProductId
		, SOL.[Year]
		, SOl.[Month]
		, SUM(SOL.OrderQty) TotalQty
		, AVG(SOL.OrderQty) AverageQty
		FROM Sales.vSalesOrders SOL
		WHERE SOL.ProductId IN (745)
		  AND SOL.[Year] = 2012
		  AND SOL.[Month] BETWEEN 1 AND 3
		GROUP BY SOL.ProductId
		, SOL.[Year]
		, SOl.[Month]
	) ASO
	ON SOL.ProductId = ASO.ProductId 
   AND SOL.[Year] = ASO.[Year]
   AND SOL.[Month] = ASO.[Month]
ORDER BY SOL.ProductId
, SOL.OrderDate

Now, just imagine that you can't create a view and having to duplicate the logic each time the view is used! Aggregating the data at different levels would require similar joins to the same view or piece of logic. 

Fortunately, SQL Server 2005 came with two long-awaited features, common table expressions (CTEs) and window functions, which made a big difference. First, CTEs allowed defining inline views that can be referenced multiple times. Secondly, the window functions allowed aggregations within a partition.

 

-- sales orders with CTE SQL Server 2005+ query 
WITH CTE
AS (--sales orders in scope
SELECT SOL.SalesOrderId 
, SOL.ProductId
, Cast(SOH.OrderDate as Date) OrderDate
, Year(SOH.OrderDate) [Year]
, Month(SOH.OrderDate) [Month]
, SOL.OrderQty
FROM Sales.SalesOrderDetail SOL
     JOIN Sales.SalesOrderHeader SOH
	   ON SOL.SalesOrderId = SOH.SalesOrderId
WHERE SOL.ProductId IN (745)
  AND Year(SOH.OrderDate) = 2012
  AND Month(SOH.OrderDate) BETWEEN 1 AND 3
)
SELECT SOL.SalesOrderId 
, SOL.ProductId
, SOL. OrderDate
, SOL.[Year]
, SOL.[Month]
, SOL.OrderQty
, SUM(SOL.OrderQty) OVER(PARTITION BY SOL.[Year], SOL.[Month]) TotalQty
, AVG(SOL.OrderQty)  OVER(PARTITION BY SOL.[Year], SOL.[Month]) AverageQty
FROM CTE SOL
WHERE SOL.ProductId IN (745)
  AND SOL.[Year] = 2012
  AND SOL.[Month] BETWEEN 1 AND 3

This kind of structuring the queries allows to separate the base logic for better maintainability, readability and fast prototyping. Once the logic from CTE becomes stable, it can be moved within a view, following to replace the CTE reference with view's name in the final query. On the other side, the window functions allow writing more flexible and complex code, even if the statements can become occasionally complex (see OrderQtyPct):

 

-- aggregated sales orders (SQL Server 2005+)
SELECT SOL.SalesOrderId 
, SOL.ProductId
, SOL. OrderDate
, SOL.[Year]
, SOL.[Month]
, SOL.OrderQty
, SUM(SOL.OrderQty) OVER(PARTITION BY SOL.[Year], SOL.[Month]) TotalQty
, AVG(SOL.OrderQty)  OVER(PARTITION BY SOL.[Year], SOL.[Month]) AverageQty
, Cast(CASE WHEN SUM(SOL.OrderQty) OVER(PARTITION BY SOL.[Year], SOL.[Month]) <> 0 THEN 100*SOL.OrderQty/SUM(SOL.OrderQty) OVER(PARTITION BY SOL.[Year], SOL.[Month]) ELSE 0 END as decimal(18,2)) OrderQtyPct
FROM Sales.vSalesOrders SOL
WHERE SOL.ProductId IN (745)
  AND SOL.[Year] = 2012
  AND SOL.[Month] BETWEEN 1 AND 3

Now, SQL Server 2022 allows to further simplify the logic by allowing to define with a name the window at the end of the statement and reference the name in several window functions (see last line of the query):

-- Aggregations per month (SQL Server 2022+)
SELECT SOL.SalesOrderId 
, SOL.ProductId
, SOL. OrderDate
, SOL.[Year]
, SOL.[Month]
, SOL.OrderQty
, SUM(SOL.OrderQty) OVER SalesByMonth AS TotalQty
, AVG(SOL.OrderQty) OVER SalesByMonth AS AverageQty
, Cast(CASE WHEN SUM(SOL.OrderQty) OVER SalesByMonth <> 0 THEN 100*SOL.OrderQty/SUM(SOL.OrderQty) OVER SalesByMonth ELSE 0 END as decimal(18,2)) OrderQtyPct
FROM Sales.vSalesOrders SOL
WHERE SOL.ProductId IN (745)
  AND SOL.[Year] = 2012
  AND SOL.[Month] BETWEEN 1 AND 3
WINDOW SalesByMonth AS (PARTITION BY SOL.[Year], SOL.[Month])

Moreover, several windows can be defined. For example, aggregating the data also per year:

-- Aggregations per month and year (SQL Server 2022+) 
SELECT SOL.SalesOrderId 
, SOL.ProductId
, SOL. OrderDate
, SOL.[Year]
, SOL.[Month]
, SOL.OrderQty
, SUM(SOL.OrderQty) OVER SalesByMonth AS TotalQtyByMonth
, AVG(SOL.OrderQty) OVER SalesByMonth AS AverageQtyByMonth
, Cast(CASE WHEN SUM(SOL.OrderQty) OVER SalesByMonth <> 0 THEN 100*SOL.OrderQty/SUM(SOL.OrderQty) OVER SalesByMonth ELSE 0 END as decimal(18,2)) OrderQtyPctByMonth
, SUM(SOL.OrderQty) OVER SalesByYear AS TotalQtyByYear
, AVG(SOL.OrderQty) OVER SalesByYear AS AverageQtyByYear
, Cast(CASE WHEN SUM(SOL.OrderQty) OVER SalesByYear <> 0 THEN 100*SOL.OrderQty/SUM(SOL.OrderQty) OVER SalesByYear ELSE 0 END as decimal(18,2)) OrderQtyPctByYear
FROM Sales.vSalesOrders SOL
WHERE SOL.ProductId IN (745)
  AND SOL.[Year] = 2012
  AND SOL.[Month] BETWEEN 1 AND 3
WINDOW SalesByMonth AS (PARTITION BY SOL.[Year], SOL.[Month])
, SalesByYear AS  (PARTITION BY SOL.[Year])

Isn't that cool? It will take probably some time to get used in writing and reading this kind of queries, though using descriptive names for the windows should facilitate the process! 

As a side note, even if there's no product function like in Excel, there's a mathematical trick to transform a product into a sum of elements by applying the Exp (exponential) and Log (logarithm) functions (see example).

Happy coding!

Data Analytics: Data Lakes/Lakehouses (Just the Quotes)

"If you think of a Data Mart as a store of bottled water, cleansed and packaged and structured for easy consumption, the Data Lake is a large body of water in a more natural state. [...] The contents of the Data Lake stream in from a source to fill the lake, and various users of the lake can come to examine, dive in, or take samples." (James Dixon, "Pentaho, Hadoop, and Data Lakes", 2010) [sorce] [first known usage]

"A data lake represents an environment that collects and stores large volumes of structured and unstructured datasets, typically in their original, unaltered forms. More than a data depository, the data lake architecture enables the various users and data science teams to conduct data exploration and related analytical activities." (EMC Education Services, "Data Science & Big Data Analytics", 2015)

"A data lake strategy supports the introduction of a separate analytics environment that off-loads the analytics being done today on your overly expensive data warehouse. This separate analytics environment provides the data science team an on-demand, fail-fast environment for quickly ingesting and analyzing a wide variety of data sources in an attempt to address immediate business opportunities independent of the data warehouse's production schedule and service level agreement (SLA) rules." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"At its core, it is a data storage and processing repository in which all of the data in an organization can be placed so that every internal and external systems', partners', and collaborators' data flows into it and insights spring out. [...] Data Lake is a huge repository that holds every kind of data in its raw format until it is needed by anyone in the organization to analyze." (Beulah S Purra & Pradeep Pasupuleti, "Data Lake Development with Big Data", 2015) 

"Having multiple data lakes replicates the same problems that were created with multiple data warehouses - disparate data siloes and data fiefdoms that don't facilitate sharing of the corporate data assets across the organization. Organizations need to have a single data lake from which they can source the data for their BI/data warehousing and analytic needs. The data lake may never become the 'single version of the truth' for the organization, but then again, neither will the data warehouse. Instead, the data lake becomes the 'single or central repository for all the organization's data' from which all the organization's reporting and analytic needs are sourced." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"[...] the real power of the data lake is to enable advanced analytics or data science on the detailed and complete history of data in an attempt to uncover new variables and metrics that are better predictors of business performance." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"The data lake is not an incremental enhancement to the data warehouse, and it is NOT data warehouse 2.0. The data lake enables entirely new capabilities that allow your organization to address data and analytic challenges that the data warehouse could not address." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"Unfortunately, some organizations are replicating the bad data warehouse practice by creating special-purpose data lakes - data lakes to address a specific business need. Resist that urge! Instead, source the data that is needed for that specific business need into an 'analytic sandbox' where the data scientists and the business users can collaborate to find those data variables and analytic models that are better predictors of the business performance. Within the 'analytic sandbox', the organization can bring together (ingest and integrate) the data that it wants to test, build the analytic models, test the model's goodness of fit, acquire new data, refine the analytic models, and retest the goodness of fit." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"A data lake is a storage repository that holds a very large amount of data, often from diverse sources, in native format until needed. In some respects, a data lake can be compared to a staging area of a data warehouse, but there are key differences. Just like a staging area, a data lake is a conglomeration point for raw data from diverse sources. However, a staging area only stores new data needed for addition to the data warehouse and is a transient data store. In contrast, a data lake typically stores all possible data that might be needed for an undefined amount of analysis and reporting, allowing analysts to explore new data relationships. In addition, a data lake is usually built on commodity hardware and software such as Hadoop, whereas traditional staging areas typically reside in structured databases that require specialized servers." (Mike Fleckenstein & Lorraine Fellows, "Modern Data Strategy", 2018)

"A data warehouse follows a pre-built static structure to model source data. Any changes at the structural and configuration level must go through a stringent business review process and impact analysis. Data lakes are very agile. Consumption or analytical layer can be modified to fit in the model requirements. Consumers of a data lake are not constant; therefore, schema and modeling lies at the liberty of analysts and scientists." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data in the data lake should never get disposed. Data driven strategy must define steps to version the data and handle deletes and updates from the source systems." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data governance policies must not enforce constraints on data - Data governance intends to control the level of democracy within the data lake. Its sole purpose of existence is to maintain the quality level through audits, compliance, and timely checks. Data flow, either by its size or quality, must not be constrained through governance norms. [...] Effective data governance elevates confidence in data lake quality and stability, which is a critical factor to data lake success story. Data compliance, data sharing, risk and privacy evaluation, access management, and data security are all factors that impact regulation." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data Lake induces accessibility and catalyzes availability. It warrants data discovery platforms to soak the data trends at a horizontal scale and produce visual insights. It largely cuts down the time that goes into data preparation and exhaustive data analysis." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data Lake is a single window snapshot of all enterprise data in its raw format, be it structured, semi-structured, or unstructured. Starting from curating the data ingestion pipeline to the transformation layer for analytical consumption, every aspect of data gets addressed in a data lake ecosystem. It is supposed to hold enormous volumes of data of varied structures." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data swamp, on the other hand, presents the devil side of a lake. A data lake in a state of anarchy is nothing but turns into a data swamp. It lacks stable data governance practices, lacks metadata management, and plays weak on ingestion framework. Uncontrolled and untracked access to source data may produce duplicate copies of data and impose pressure on storage systems." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data warehousing, as we are aware, is the traditional approach of consolidating data from multiple source systems and combining into one store that would serve as the source for analytical and business intelligence reporting. The concept of data warehousing resolved the problems of data heterogeneity and low-level integration. In terms of objectives, a data lake is no different from a data warehouse. Both are primary advocates of terms like 'single source of truth' and 'central data repository'." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"A data lakehouse is an amalgamation of the best components from both data lakes and data warehouses. A data lakehouse implements data structure and data management features from data warehouses into a cost-effective storage like a data lake. It tries to combine the best from both worlds - data lake - based Big Data analytics and a data warehouse." (Bhadresh Shiyal, "Beginning Azure Synapse Analytics: Transition from Data Warehouse to Data Lakehouse", 2021) 

"A defining characteristic of the data lakehouse architecture is allowing direct access to data as files while retaining the valuable properties of a data warehouse. Just do both!" (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"At first, we threw all of this data into a pit called the 'data lake'. But we soon discovered that merely throwing data into a pit was a pointless exercise. To be useful - to be analyzed - data needed to (1) be related to each other and (2) have its analytical infrastructure carefully arranged and made available to the end user. Unless we meet these two conditions, the data lake turns into a swamp, and swamps start to smell after a while. [...] In a data swamp, data just sits there are no one uses it. In the data swamp, data just rots over time." (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"Data lake architecture suffers from complexity and deterioration. It creates complex and unwieldy pipelines of batch or streaming jobs operated by a central team of hyper-specialized data engineers. It deteriorates over time. Its unmanaged datasets, which are often untrusted and inaccessible, provide little value. The data lineage and dependencies are obscured and hard to track." (Zhamak Dehghani, "Data Mesh: Delivering Data-Driven Value at Scale", 2021)

"Once you combine the data lake along with analytical infrastructure, the entire infrastructure can be called a data lakehouse. [...] The data lake without the analytical infrastructure simply becomes a data swamp. And a data swamp does no one any good." (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"The data lakehouse architecture presents an opportunity comparable to the one seen during the early years of the data warehouse market. The unique ability of the lakehouse to manage data in an open environment, blend all varieties of data from all parts of the enterprise, and combine the data science focus of the data lake with the end user analytics of the data warehouse will unlock incredible value for organizations. [...] The lakehouse architecture equally makes it natural to manage and apply models where the data lives." (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"With the data lakehouse, it is possible to achieve a level of analytics and machine learning that is not feasible or possible any other way. But like all architectural structures, the data lakehouse requires an understanding of architecture and an ability to plan and create a blueprint." (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"Delta Lake is a transactional storage software layer that runs on top of an existing data lake and adds RDW-like features that improve the lake’s reliability, security, and performance. Delta Lake itself is not storage. In most cases, it’s easy to turn a data lake into a Delta Lake; all you need to do is specify, when you are storing data to your data lake, that you want to save it in Delta Lake format (as opposed to other formats, like CSV or JSON)." (James Serra, "Deciphering Data Architectures", 2024)

DBMS: Data Warehouse/Mart (Just the Quotes)

"Unfortunately, just collecting the data in one place and making it easily available isn’t enough. When operational data from transactions is loaded into the data warehouse, it often contains missing or inaccurate data. How good or bad the data is a function of the amount of input checking done in the application that generates the transaction. Unfortunately, many deployed applications are less than stellar when it comes to validating the inputs. To overcome this problem, the operational data must go through a 'cleansing' process, which takes care of missing or out-of-range values. If this cleansing step is not done before the data is loaded into the data warehouse, it will have to be performed repeatedly whenever that data is used in a data mining operation." (Joseph P Bigus,"Data Mining with Neural Networks: Solving business problems from application development to decision support", 1996)

"There are four levels of data in the architected environment - the operational level, the atomic (or the data warehouse) level, the departmental (or the data mart) level, and the individual level. These different levels of data are the basis of a larger architecture called the corporate information factory (CIF). The operational level of data holds application-oriented primitive data only and primarily serves the high-performance transaction-processing community. The data-warehouse level of data holds integrated, historical primitive data that cannot be updated. In addition, some derived data is found there. The departmental or data mart level of data contains derived data almost exclusively. The departmental or data mart level of data is shaped by end-user requirements into a form specifically suited to the needs of the department. And the individual level of data is where much heuristic analysis is done." (William H Inmon, "Building the Data Warehouse" 4th Ed., 2005)

"Having a purposeless or poorly performing dashboard is more common than not. This happens when the underlying architecture is not designed properly to support the needs of dashboard interaction. There is an obvious disconnect between the design of the data warehouse and the design of the dashboards. The people who design the data warehouse do not know what the dashboard will do; and the people who design the dashboards do not know how the data warehouse was designed, resulting in a lack of cohesion between the two. A similar disconnect can also exist between the dashboard designer and the business analyst, resulting in a dashboard that may look beautiful and dazzling but brings very little business value." (Nils H Rasmussen et al, "Business Dashboards: A visual catalog for design and deployment", 2009)

"Having multiple data lakes replicates the same problems that were created with multiple data warehouses - disparate data siloes and data fiefdoms that don't facilitate sharing of the corporate data assets across the organization. Organizations need to have a single data lake from which they can source the data for their BI/data warehousing and analytic needs. The data lake may never become the 'single version of the truth' for the organization, but then again, neither will the data warehouse. Instead, the data lake becomes the 'single or central repository for all the organization's data' from which all the organization's reporting and analytic needs are sourced." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"There are, however, many problems with independent data marts. Independent data marts: (1) Do not have data that can be reconciled with other data marts (2) Require their own independent integration of raw data (3) Do not provide a foundation that can be built on whenever there are future analytical needs." (William H Inmon & Daniel Linstedt, "Data Architecture: A Primer for the Data Scientist: Big Data, Data Warehouse and Data Vault", 2015)

"Unfortunately, some organizations are replicating the bad data warehouse practice by creating special-purpose data lakes - data lakes to address a specific business need. Resist that urge! Instead, source the data that is needed for that specific business need into an 'analytic sandbox' where the data scientists and the business users can collaborate to find those data variables and analytic models that are better predictors of the business performance. Within the 'analytic sandbox', the organization can bring together (ingest and integrate) the data that it wants to test, build the analytic models, test the model's goodness of fit, acquire new data, refine the analytic models, and retest the goodness of fit." (Billl Schmarzo, "Driving Business Strategies with Data Science: Big Data MBA" 1st Ed., 2015)

"Data quality in warehousing and BI is typically defined in terms of the 4 C’s - is the data clean, correct, consistent, and complete? When it comes to big data, there are two schools of thought that have different views and expectations of data quality. The first school believes that the gold standard of the 4 C’s must apply to all data (big and little) used for clinical care and performance metrics. The second school believes that in big data environments, a stringent data quality standard is impossible, too costly, or not required. While diametrically opposite opinions may play well in panel discussions, they do little to reconcile the realities of healthcare data quality." (Prashant Natarajan et al, "Demystifying Big Data and Machine Learning for Healthcare", 2017) 

"Data warehousing has always been difficult, because leaders within an organization want to approach warehousing and analytics as just another technology or application buy. Viewed in this light, they fail to understand the complexity and interdependent nature of building an enterprise reporting environment." (Prashant Natarajan et al, "Demystifying Big Data and Machine Learning for Healthcare", 2017)

"A data lake is a storage repository that holds a very large amount of data, often from diverse sources, in native format until needed. In some respects, a data lake can be compared to a staging area of a data warehouse, but there are key differences. Just like a staging area, a data lake is a conglomeration point for raw data from diverse sources. However, a staging area only stores new data needed for addition to the data warehouse and is a transient data store. In contrast, a data lake typically stores all possible data that might be needed for an undefined amount of analysis and reporting, allowing analysts to explore new data relationships. In addition, a data lake is usually built on commodity hardware and software such as Hadoop, whereas traditional staging areas typically reside in structured databases that require specialized servers." (Mike Fleckenstein & Lorraine Fellows, "Modern Data Strategy", 2018)

"A data warehouse follows a pre-built static structure to model source data. Any changes at the structural and configuration level must go through a stringent business review process and impact analysis. Data lakes are very agile. Consumption or analytical layer can be modified to fit in the model requirements. Consumers of a data lake are not constant; therefore, schema and modeling lies at the liberty of analysts and scientists." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"Data warehousing, as we are aware, is the traditional approach of consolidating data from multiple source systems and combining into one store that would serve as the source for analytical and business intelligence reporting. The concept of data warehousing resolved the problems of data heterogeneity and low-level integration. In terms of objectives, a data lake is no different from a data warehouse. Both are primary advocates of terms like 'single source of truth' and 'central data repository'." (Saurabh Gupta et al, "Practical Enterprise Data Lake Insights", 2018)

"A defining characteristic of the data lakehouse architecture is allowing direct access to data as files while retaining the valuable properties of a data warehouse. Just do both!" (Bill Inmon et al, "Building the Data Lakehouse", 2021)

"The data lakehouse architecture presents an opportunity comparable to the one seen during the early years of the data warehouse market. The unique ability of the lakehouse to manage data in an open environment, blend all varieties of data from all parts of the enterprise, and combine the data science focus of the data lake with the end user analytics of the data warehouse will unlock incredible value for organizations. [...] "The lakehouse architecture equally makes it natural to manage and apply models where the data lives." (Bill Inmon et al, "Building the Data Lakehouse", 2021)

Performance Management: First Time Right (The Aim toward Operational Excellence)

 

Rooted in Six Sigma methodology as a step toward operational excellence, First Time Right (FTR) implies that any procedure is performed in the right manner the first time and every time. It equates to minimizing the waste in its various forms (inventory, motion, overprocessing, overproduction, waiting, transportation, defects). Like many quality concepts from the manufacturing industry, the concept was transported in the software development process as principle, process, goal and/or metric. Thus, it became part of Software Engineering, Project Management, Data Science, and any other similar endeavors whose outcome results in software products. 

Besides the quality aspect, FTR is rooted also in the economic imperative – the need to achieve something in the minimum amount of time with the minimum of effort. It’s about being efficient in delivering a product or achieving a given target. It can be associated with continuous improvement, learning and mastery, the aim being to encompass FTR as part of organization’s culture. 

Even if not explicitly declared, FTR lurks in each task planned. It seems that it became common practice to plan with the FTR in mind, however between this theoretical aim and practice there’s as usual an important gap. Unfortunately, planners, managers and even tasks' performers often forget that mistakes are made, that several iterations are needed to get the job done. It starts with the communication between people in clarifying the requirements and ends with the formal sign off. All the deviations from the FTR add up in the deviations between expected and actual effort, though probably more important are the deviations from the plan and all the consequences deriving from it. Especially in complex projects this adds up into a spiral of issues that can easily reinforce themselves. 

Many of the jobs that imply creativity, innovation, research or exploration require at least several iterations to get the job done and this is independent of participants’ professionalism and experience. Moreover, the more quality one needs, the higher the effort, the 80/20 being sometimes a good approximation of the effort needed. In extremis, aiming for perfection instead of excellence can make certain tasks a never-ending story. 

Achieving FTR requires practice - the more novelty, the higher the complexity, the communication or the synchronization needs, the more practice is needed. It starts with the individual to master the individual tasks and ends with the team, where communication, synchronization and other aspects need to be considered. The practice is usually achieved on hands-on work as part of the daily duties, project work, and so on. Unfortunately, it’s based primarily on individual experience, and seldom groomed in advance, as preparation for future tasks. That’s why sometimes when efficiency is needed in performing critical complex tasks, one also needs to consider the learning curve in achieving the required quality. 

Of course, many organizations demand from job applicants experience and, when possible, they hire people with experience, however the diversity, complexity and changing nature of tasks require further practice. This aspect is somehow recognized in the implementation in organizations of the various forms of DevOps, though how many organizations adopt it and enforce it on a regular basis? Moreover, a major requirement of nowadays businesses is to be agile, and besides the mere application of methodologies, being agile means to have also a FTR mindset. 

FTR starts with the wish for mastery at individual and team level and, with the right management attention, by allocating time for learning, self-development in the important areas, providing relevant feedback and building an infrastructure for knowledge sharing and harnessing, FTR can become part of organization’s culture. It’s up to each of us to do it!

SQL Reloaded: Successive Price Increases/Discounts via Windowing Functions and CTEs

I was trying today to solve a problem that apparently requires recursive common table expressions, though they are not (yet) available in Azure Synapse serverless SQL pool. The problem can be summarized in the below table definition, in which given a set of Products with an initial Sales price, is needed to apply Price Increases successively for each Cycle. The cumulated increase is simulated in the last column for each line. 

Unfortunately, there is no SQL Server windowing function that allows multiplying incrementally the values of a column (similar as the running total works). However, there’s a mathematical trick that can be used to transform a product into a sum of elements by applying the Exp (exponential) and Log (logarithm) functions (see Solution 1), and which frankly is more elegant than applying CTEs (see Solution 2). 

-- create table with test data
SELECT *
INTO dbo.ItemPrices
FROM (VALUES ('ID001', 1000, 1, 1.02, '1.02')
, ('ID001', 1000, 2, 1.03, '1.02*1.03')
, ('ID001', 1000, 3, 1.03, '1.02*1.03*1.03')
, ('ID001', 1000, 4, 1.04, '1.02*1.03*1.03*1.04')
, ('ID002', 100, 1, 1.02, '1.02')
, ('ID002', 100, 2, 1.03, '1.02*1.03')
, ('ID002', 100, 3, 1.04, '1.02*1.03*1.04')
, ('ID002', 100, 4, 1.05, '1.02*1.03*1.04*1.05')
) DAT (ItemId, SalesPrice, Cycle, PriceIncrease, CumulatedIncrease)

-- reviewing the data
SELECT *
FROM dbo.ItemPrices

-- Solution 1: new sales prices with log & exp
SELECT ItemId
, SalesPrice
, Cycle
, PriceIncrease
, EXP(SUM(Log(PriceIncrease)) OVER(PARTITION BY Itemid ORDER BY Cycle)) CumulatedIncrease
, SalesPrice * EXP(SUM(Log(PriceIncrease)) OVER(PARTITION BY Itemid ORDER BY Cycle)) NewSalesPrice
FROM dbo.ItemPrices

-- Solution 2: new sales prices with recursive CTE
;WITH CTE 
AS (
-- initial record
SELECT ITP.ItemId
, ITP.SalesPrice
, ITP.Cycle
, ITP.PriceIncrease
, cast(ITP.PriceIncrease as decimal(38,6)) CumulatedIncrease
FROM dbo.ItemPrices ITP
WHERE ITP.Cycle = 1
UNION ALL
-- recursice part
SELECT ITP.ItemId
, ITP.SalesPrice
, ITP.Cycle
, ITP.PriceIncrease
, Cast(ITP.PriceIncrease * ITO.CumulatedIncrease as decimal(38,6))  CumulatedIncrease
FROM dbo.ItemPrices ITP
    JOIN CTE ITO
	  ON ITP.ItemId = ITO.ItemId
	 AND ITP.Cycle-1 = ITO.Cycle
)
-- final result
SELECT ItemId
, SalesPrice
, Cycle
, PriceIncrease
, CumulatedIncrease
, SalesPrice * CumulatedIncrease NewSalesPrice
FROM CTE
ORDER BY ItemId
, Cycle


-- validating the cumulated price increases (only last ones)
SELECT 1.02*1.03*1.03*1.04 
, 1.02*1.03*1.04*1.05

-- cleaning up
DROP TABLE IF EXISTS dbo.ItemPrices

Notes:
1. The logarithm in SQL Server’s implementation works only with positive numbers!
2. For simplification I transformed percentages (e.g. 1%) in values that are easier to multitply with (e.g. 1.01). The solution can be easily modified to consider discounts.
3. When CTEs are not available, one is forced to return to the methods used in SQL Server 2000 (I've been there) and thus use temporary tables or table variables with loops. Moreover, the logic can be encapsulated in multi-statement table-valued functions (see example), unfortunately, another feature not (yet) supported by serverless SQL pools. 
4. Unfortunately, STRING_AGG, which concatenates values across rows, works only with a GROUP BY clause. Anyway, its result is useless without the availability of a Eval function in SQL (see example), however the Expr function available in data flows could be used as workaround.
4. Even if I thought about the use of logarithms for transforming the product into a sum, I initially ignored the idea, thinking that the solution would be too complex to implement. So, the credit goes to another blogpost. Kudos!

Happy coding!

ERP Implementations: It’s all about Partnership II

When starting an ERP implementation project an organization needs to fill the existing knowledge gaps in respect to whatever it takes to achieve the goals associated with the respective project. Therefore, it makes sense to work with a implementer that can help cover the gaps directly or indirectly. Moreover, it makes sense to establish a long-term relationship that would allow to harness ERP system’s capabilities after project’s end, increase the ROI and, why not, find other areas of cooperation. It’s in theory what a partner does, and a strategic technology partnership is about – providing any kind of technological expertise the customer doesn't have in-house. 

Unfortunately, from being a ‘service provider’ to becoming a ‘partner’ is a challenging road for many organizations, especially when this type of relationship is not understood and managed accordingly. Partnership’s management may resume in defining common goals, principles, values and processes, establishing a communication strategy and a common understanding of the challenges and the steps ahead, providing visibility into the cost estimates, billing, resources’ availability and utilization. Addressing these aspects would offer a framework on which the partnerships can nourish. Without considering these topics, the implementer remains just a 'service provider', no matter of the names used to characterize the relationship. 

Now, the use of the word ‘partner’ would make someone think that only one partner is considered, typically a big to middle-sized organization that would have this kind of resources. The main reason behind this reasoning is that the number of functional areas and volume of skillset required for filling the requirements of an implementation are high compared with other projects, the resources needing to be available on-demand without affecting the other constraints: costs, quality, time. This can be challenging, therefore can be met scenarios in which two or more external organizations are involved in the partnership, ideally organizations that complement each other. 

It is common in ERP implementations to appeal also to individual consultants for specific areas or the whole project. The principles and values of a partnership, as well the framework behind, can be applied to individual consultants as well. Independently of resources’ provenience more important is the partnership ‘mindset’ - being together in the same boat, working together on a shared and understood strategy, with clear goals and objectives.

Moreover, the people participating in the project must have a ‘partner's mindset’ as well. Without this, the project will likely get different impulses in the wrong direction(s), as a group’s interests will take priority over the ones of the organization. Ideally, this mindset should extend to the whole organization as topics like Data Quality and Process Improvement must be an organization’s effort, deep imprinted in organization’s culture.

More like ever, it’s important for the business to see and treat the IT department as a ‘partner’ and not as a ‘service provider’ by providing the needed level of transparency in requirements, issues, practices and processes, by treating the IT department as equal party in the decision-making and addressing its current and future strategical requirements. Ideally, this partnership should happen long before the implementation starts, given that it takes time for mentalities and practices to change, for knowledge to be acquired and used appropriately. 

Building a partnership takes time, effort and strategic thinking, this on top of the actual implementation, increasing thus the overall complexity, at least at the beginning. Does it pay off? Like in a marriage, it’s useful to have somebody you can trust, who knows you, whom you can rely upon, and talk with to find solutions. However, only time will tell whether such expectations are met and kept till the end. 

Previous <||> Next

Business Intelligence: Power BI’s Learning Curve - Part I

A learning curve attempts depicting the (average) time it takes a person to learn how to use a method, tool, or technique, tracing the path from newbie to mastery. A common definition of the learning curve is based on the correlation between a learner’s performance on a task or activity and the number of attempts or amount of time required to complete it.

There are several diagrams in circulation which depict the correlation between the difficulty of Power BI concepts and probably their implementation as functionality. Even if they reflect to some degree the rate of learning, their simplicity and fuzziness can easily make one question their accuracy in reflecting the reality.

Researchers tend to categorize the curves associated with the learning process in simple idealized patterns like S-curve (aka sigmoid), exponential growth, exponential rise and fall to limit, or power law, however the learning process in IT-based endeavors is seldom characterized by a linear or exponential curve, given that the tasks seldom allow a steady path. The jumps of knowledge between tasks can be wide enough to appear insurmountable, and they can prove to be quite of a challenge without some help.

Like a baby’s first steps, we, as learners, must learn first to crawl, before making some unsteady steps, and it can take long time until visible progress is made. It’s a slow progress until we suddenly hit a (tipping) point from which everything seems easy, fact that increases our confidence in us. On the other side, when we find that we make no visible progress for a long period, it’s easy to arrive to the opposite, a critical zone, which in extremis could make one lose interest.

As beginners, after the first tipping point on the learning journey, it’s easy to arrive at a plateau in which there seem no need to learn new things, the current knowledge allowing to handle a range of tasks of small to average complexity. This can last for a long time, and then, a big thing comes our way – a hard problem to solve or a concept hard to understand. It’s the point where we stagnate, and the deeper we go, and the more such challenges are thrown in our way, the more difficult the learning seems to be. However, with new understanding, small steps are made, one step after the other, the pace makes us to evolve faster until we reach again a critical point from which the process increases smoothly until we seem to stagnate again. We meet again a hard limit to growth, which seems to be more solid than the previous one.

Both limits to growth can appear to be hard, however, considering that the knowledge in the field expands, more opportunity for growth appear, thus, the limits are apparent. Even if knowledge tends to increase ‘indefinitely’, the limits are there in terms of complexity, time available, knowledge quality (incl. availability) or any other dimension of the learning process. Moreover, these successions of tipping points, growth limits, plateaus, critical, steady and fast progress zones can occur in several iterations in the learning path. Thus, the path seems to resemble a snakelike curve with many ups and downs.

For the learner is important to be aware of this last aspect, there are always ups and downs, taking effort, patience and maybe an expert’s help to bridge the gaps in between. The chances are high that the gap between what we think we know and what we know is considerable, therefore a reality check is useful from time to time. A new problem to tackle will provide that occasion!

Previous Post <<||>> Next Post

Strategic Management: Between Value and Waste I (Introduction)

 

Independently on whether Lean Management is considered in the context of Manufacturing, Software Development (SD), Project Management (PM) or any other business-related areas, there are three fundamental business concepts on which the whole scaffolding of the Lean philosophies is built upon, namely the ones of value, value stream and waste. 

From an economic standpoint, value refers to the monetary worth of a product, asset or service (further referred as product) to an organization, while from a qualitative perspective, it refers to the perceived benefit associated with its usage. The value is thus reflected in the costs associated with a product’s delivery (producer’s perspective), respectively the price paid on acquiring it and the degree to which the product can fulfill a demand (customer’s perspective).

Without diving too deep into theory of product valuation, the challenges revolve around reducing the costs associated with a product’s delivery, respectively selling it to a price the customer is willing to pay for, typically to address a given set of needs. Moreover, the customer is willing to pay only for the functions that satisfy the needs a product is thought to cover. From this friction of opposing driving forces, a product is designed and valued.

The value stream is the sequence of activities (also steps or processes) needed to deliver a product to customers. This formulation includes value-added and non-value-added activities, internal and external customers, respectively covers the full lifecycle of products and/or services in whatever form it occurs, either if is or not perceived by the customers.  

Waste is any activity that consumes resources but creates no value for the customers or, generally, for the stakeholders, be it internal or external. The waste is typically associated with the non-added value activities, activities that don’t produce value for stakeholders, and can increase directly or indirectly the costs of products especially when no attention is given to it and/or not recognized as such. Therefore, eliminating the waste can have an important impact on products’ costs and become one of the goals of Lean Management. Moreover, eliminating the waste is an incremental process that, when put in the context of continuous improvement, can lead to processes redesign and re-engineering.

Taiichi Ohno, the ‘father’ of the Toyota Production System (TPS), originally identified seven forms of waste (Japanese: muda): overproduction, waiting, transporting, inappropriate processing, unnecessary inventory, unnecessary/excess motion, and defects. Within the context of SD and PM, Tom and Marry Poppendieck [1] translated the types of wastes in concepts closer to the language of software developers: partially done work, extra processes, extra features, task switching, waiting, motion and, of course, defects. To this list were added later further types of waste associated with resources, confusion and work conditions.

Defects in form of errors and bugs, ineffective communication, rework and overwork, waiting, repetitive activities like handoffs or even unnecessary meetings are usually the visible part of products and projects and important from the perspective of stakeholders, which in extremis can become sensitive when their volume increases out of proportion.

Unfortunately, lurking in the deep waters of projects and wrecking everything that stands in their way are the other forms of waste less perceivable from stakeholders’ side: unclear requirements/goals, code not released or not tested, specifications not implemented, scrapped code, overutilized/underutilized resources, bureaucracy, suboptimal processes, unnecessary optimization, searching for information, mismanagement, task switching, improper work condition, confusion, to mention just the important activities associated to waste.

Through their elusive nature, independently on whether they are or not visible to stakeholders, they all impact the costs of projects and products when the proper attention is not given to them and not handled accordingly.

References:
[1] Mary Poppendieck & Tom Poppendieck (2003) Lean Software Development: An Agile Toolkit, Addison Wesley, ISBN: 0-321-15078-3