Select Page

Evolution of Leadership Models

Leadership is talked quite often in modern day work places. What is meant by leadership is quite often reliant on speaker’s mental model of it. There is a high likelyhood that speaker(or the leader) may not have a structured mental model for what she means by leadership. In addition, the listener may have an entirely different understanding of leadership. The following leadership models summarize characteristic of a few leadership models and attempts to review how leadership models have evolved over time.

Heroic Leadership

Standard mode of thinking in individualistic culture revolves around the pattern of thinking where Leader is a hero and Saviour. A superman who saves the day.

Theory Theme Contributing theorists Notes to self
Heroic Leadership Leader is born, not made. Leader possesses extra ordinary capabilities with which she can perform wonders which others cannot. Thomas Carlyle in 1840 originally phrased the term, Plato,  Lao-tzu, Aristotle , Machiavelli have been contributing over time to this pattern of thinking. This theory is still dominant mode of thinking in workplace today. Recent HBR article tries to identify a few of its caveats here.

Charismatic Leadership

Slightly modified version of the heroic leader. Charismatic leader may have inborn traits, but they also cquires the skills like charisma, integrity, motivation etc.

In the words of Riggio:

C”harismatic leaders are essentially very skilled communicators – individuals who are both verbally eloquent, but also able to communicate to followers on a deep, emptional level. They are able to articulate a compelling or captivating vision and are able to arouse strong emotions in followers.”

Theory Theme Contributing theorists Notes to self
Charismatic Leader Leaders can acquire certain traits, which will help them in achieving desired objectives. Several This theory appeals because traits and characteristic are to some degree responsible for goal achievement e.g., for team building one would get a lot of help from integrity and conscientiousness.

Behavioural Leadership

Trait theory ignored the context around leadership success. Behavioural theory built upon internal traits and said that external behaviors(supported by traits) of leaders contribute to the success.

Theory Theme Contributing theorists Notes to self
Behavioural Leadership Traits are not always consistent. Therefore, external behaviour of leaders is determinant of success. Traits can be indicative of what behaviour to expect. But the actual consistency in behaviour determines success Katz, Maccoby, Gurin and Floor(1951), Stogdill and Coons(1957). Feels contributary to success. Consistency is the key and helps in achieving trust of the team.

Contingent Leadership

Contingent theory ( 1967-1990) took into account contextual variable. Leader could have the best traits (trait theory), she could exhibit them in her behavior consistently(behavioral theory), however, the external factors had been missing in accounting for leadership success. Contingency theory takes into account people, situational, organizational and enviornmental factors.

Three of the most notable theories of this era were
1. Fiedler’s Contingency Theory,
2. House’s Path-Goal Theory,
3. Vroom and Yetton’s Normative Theory of Decision Making

Fiedler identified three managerial components:

  1. Leader-member relations
  2. Task structures
  3. Position power.

some contexts favoured leaders who were task-oriented and some favoured those who were relationship oriented. 

Theory Theme Contributing theorists Notes to self
Contingent Leadership No single style of leadership is universal. It all depends on the context.

(Traits + Behaviours) + context

 Fiedler(1967,1971),

Hershey & Blanchard(1969)

 Smaller organisation are often driven by task based leaders, larger organisation have influence power arranged in relationships.

 

Servant Leaderhip

Originally coined in 1970 by Robert Greenleaf, servant leadership is linked to ethics, virtues and morality. Its origin go back through history with people such as confucius, Lao-tzu, Moses and Jesus Christ as examples of servant leader.

Spears identifies 10 characteristics of servant leaders:

  1. Receptive listening.
  2. Empathy.
  3. Healing : Recognizing they have opportunities to make themselves and others whole.
  4. Self Awareness.
  5. Persuation: Convincing over coersion.
  6. Conceptualisation: Nurturing the ability to envision greater possibilities.
  7. Foresight : Intuititvely learning lessons from past.
  8. Stewardship : Comitted first and foremost to serving the needs of the others.
  9. Commitment : to growth of people in personal, professional and spiritual realms.
  10. Building community: identifies means of building communities among individuals working within organizations.

This is my favourite theory because of its  inherent beauty and  contradictions in its implementation. It is beautiful because it is characterised by trust and respect between leaders and followers. It is despicable because of the practical lip service it gets in organization.It is truly difficult to embrace.

Now a days(2022) it is exhbited by inverting organogram, showing CEO at the bottom and employee at the top. I am always amused by CEO’s and top leaders inverting the hirearchy in an organogram and then “asking ( somewhat ordering) ” the followers to lead by “following” the new-found diagram of inverted hirearchy.

My sentiments are echoed in Robert Greenleaf’s questions (1970 ):

“The best test and difficult to administer is: do those served grow as persons; do they, while being served, become healthier, wiser, freer, more autonomous, more likely themselves to become servants? And what is the effect on the least privileged in society; will they benefit, or, at least, will they not be further deprived?”

Theory Theme Contributing theorists Notes to self
Servant Leadership Leader serve the followers.

High quality relationship are characterised by trust and respect between leader and follower. Low quality relationship are characterised by transactional and contractual obligation

 Robert Greenleaf(1970),Graen & Uhl-Bien (1995), Gerstner & Day (1997), Nahrgang & Morgeson(2007) Theoretically wonderful, practically contradictory. Can a follower change  strategy set out by leader?

Transformational Leadership

In a rapidly changing world, transformational leadership model seeks adaptation to change from the leader.In change adaptation,  It focuses on getting commitment from the follower rather than compliance,. A tranformational leadership model proposes a symbiotic relationship in contrast to a transactional relationship .  A mutual simulation converts followers into leaders and leaders into moral agents. James MacGregor Burn’s work Leadership (1987) introduced this concept. James first termed the idea of ‘Transformational leadership’ as a balanced dynamic between leaders and followers predicated upon social influence, rather than power. It looked toward fulfillment of aspirational needs of people.

Bernard Bass ( influenced by James M.)  detailed the structure of transformational leadership to include :

  1. Idealised behaviours : Walking the talk
  2. Inspirational motivation : Offering compelling vision .
  3.  Intellectual simulation: Challenge followers to be innovative and creative .
  4. Individualized consideration : Ability of the leader to demonstrate genuine concern for needs and feelings of follower.

 

Theory Theme Contributing theorists Notes to self
Transformational Leadership Leader-follower symbiosis for achieving common good. James Macgregor Burn (1987), Bernard & Bass (1985,1998) Heavy leaning on the idealistic side of the world. Asks a lot of leader, e.g. leader should be able to inspire motivation and intellectually simulate the follower. These 2 tasks in themselves are too uphill for mere mortals.

 

System Leadership

System leadership recognises that collaboration is essential to solve wicked problems (Heifetz,Kani, and Kramer, 2004). Leaders in system leadership model sacrifice their ego for common good. They move from individual to collaborative responsibility. Reactivity is replaced pro-active collaboration. Leaders collaborate with followers to achieve a shared vision. Power distribution in the organisation is decentralized. Hirearchy has its foundation in mutual trust rather than vertical organogram.

Theory Theme Contributing theorists Notes to self
System Leadership Decentralized power, mutual trust, and letting go of a leader’s ego for achieving shared vision Senge, Hamilton, and Kania (2015), Jim Collins, Fredric Laloux (2014) Especially in a knowledge-based economy, this model serves well. It solves the complexity of interconnected systems. It also nourishes individual autonomy and individual ideation while keeping focus on common goals. 

 

References :

Gene Early, A short history of leadership theories.

EdX, University of Queensland, Becoming an effective leader

Campaign Structure for Marketing platforms

Marketing platforms structure their campaigns in slightly different ways. The difference can be either in hirearchy( nesting levels) or the terminology used for hirearchy. For doing any cross-platform marketing analysis, it’s good to understand at a high level, how these campaigns are structures in different platforms// This article provides a reference point for campaign structures of major digital ad platforms.

 

Facebook’s Campaign Structure

Facebook structures its campaign into Campaign, Ad set and Ads.

AdWords’ Campaign Structure

Ad words have multiple campaign types because it caters for search as well as display. There are 2 primary types of Adwords Campaign structures

Search Campaigns

Display Campaigns

Display campaigns have similar structure as that of search. One key difference is that instead of Keywords and Match types in Adgroups , you select Placement.Keywords in search ads determine which user searches should trigger the relevant ads whereas Placement in display ad identifies where you want your ads to be displayed.

Facebook’s Ad Auction

Facebook’s Auction System

Facebook use auctions to determine which ads to show to the user. This is similar to how google price its ads. In this article we will explore how facebook’s Ad auction system works.

Facebook Ad Auction system has two goals :

  1. Create value for advertisers ( maximise return on investment)
  2. Provide positive user experience ( show relative ads )

Each time an auction takes place, facebook’s algorithm decide which ad to show based on Total Value for each Ad

Total Value  = Advertiser bid  *  Estimated action rate + Ad quality

Lets go into each of these factors individually.

Advertiser bid :

Bid is the price that you are willing to pay for an Ad in dollars.  This Ad bid can be either specificed manually or you can let facebook decide the best price. The important factor to note is that highest bid price does not guarantee that Ad will always be served to the audience your target audience. Final Ad display will always be calculated by Total Value formulae.

Estimated action rate :

Estimated action rate is based on Ad relevance. If an Ad is more aligned with user’s interest, it gets higher score. If same ad is less aligned, it gets lower score. We can understand it with an example.

Ad Audience Interest Relevance Score

(hypothetical score out of 1)
 Estimated action rate
Creative 1 Health & wellbeing 0.8 High
Creative 1 Fast food & desserts 0.1 Low

The scoring method used in the example above is arbitrary and only used for illustration purpose only.

While facebook does not explicitly lists the factors making up estimated action rate, following factors are indicative of what makes up estimated action rate.

  1. Relevant campaign objectives
  2. Relevant audience
  3. High quality creatives
  4. Historical conversion probability

Ad Quality :

Ad quality is determined by several factors. The actual factors are not explicitly given by facebook. However, the objective of Ad quality is that to serve relevant Ads to right audiences. The following factors contribute to Ad Quality score :

  1. CTR ( click-through rate of the ad)
  2. Ads relevancy score
  3. Crowdsourced Feedback of the Ad ( Ads hidden , reported bad)
  4. Text to Image ratio
  5. Landing Page experience
  6. Landing Page speed
  7. Authenticity of landing page
  8. Authenticity of business
  9. Popup, ads or other contents which can be detrimental for user experience
  10. Expected conversion rate
  11. Historical conversion rate

Weightage of factors ?

Facebook does not provide the weightage of each factor in the Total Value formulae. In an ideal world, advertiser’s would have the formulae with weigtahges e.g. if  x,y and z are the weigthages  then formulae would be like:

Total Value  = (x)Advertiser bid  *  (y)Estimated action rate + (z)Ad quality

However, these weightages remain unknown to the advertisers. Only facebook knows the recipe of the secret sauce of x,y & z.

Facebook Charging Model:

Facebook has 3 charging models which it uses for billing its advertisers.

  1. CPM billing
  2. Link click billing
  3. Action based billing

1. CPM billing:

Since facebook is primarily a display based Ad platform. It’s most dominant biling format is CPM. CPM stands for cost per Mile ( 1000). As an example, if an ad’s cost is $1 per CPM. This implies, when the Ad will be served 1000 times, then advertiser will pay $1.

2. Link click billing:

Facebook also gives an option of paying when a user clicks the link on an Ad.

3. Action based billing:

Action based billing charges advertiser on the bases of certain actions e.g. When a user watches a complete video etc.

Lambda and Kappa Architectures

Lambda and Kappa architectures are two popular data processing architectures. They both represent systems designed to handle ingestion, processing and storage of large volumes of data and providing analytics. Understanding these architectures and their respective strengths can help organizations choose the right approach for their specific needs.

What Are Lambda and Kappa Architectures?

Lambda Architecture

Lambda architecture is a data-processing framework designed to handle massive quantities of data by using both batch and real-time stream processing methods. The batch layer processes raw data in batches using tools like Hadoop or Spark, storing the results in a batch view. The speed layer handles incoming data streams with low-latency engines like Storm or Flink, storing the results in a speed view. The serving layer queries both views and combines them to provide a unified data view.

Kappa Architecture

Kappa architecture, on the other hand, is designed to handle real-time data exclusively. A single processing layer handles all data in real-time using tools like Kafka, Flink, or Spark Streaming. There is no batch layer. Instead, all incoming data streams are processed immediately and continuously, storing the results in a real-time view. The serving layer queries this real-time view directly to provide up-to-the-second data insights

Key Principles of Lambda and Kappa Architectures

Lambda Architecture

  1. Dual Data Model:
    • Uses separate models for batch and real-time processing.
    • Batch layer processes historical data ensuring accuracy.
    • Speed layer handles real-time data for low latency insights.
  2. Single Unified View:
    • Combines outputs from both batch and speed layers into a single presentation layer.
    • Provides comprehensive and up-to-date views of the data.
  3. Decoupled Processing Layers:
    • Allows independent scaling and maintenance of batch and speed layers.
    • Enhances flexibility and ease of development.

Kappa Architecture

  1. Real-Time Processing:
    • Focuses entirely on real-time processing.
    • Processes events as they are received, reducing latency.
  2. Single Event Stream:
    • Utilizes a unified event stream for all data.
    • Simplifies scalability and fault tolerance.
  3. Stateless Processing:
    • Each event is processed independently without maintaining state.
    • Facilitates easier scaling across multiple nodes.

Key Features Comparison

FeatureLambda ArchitectureKappa Architecture
Processing ModelDual (Batch + Stream)Single (Stream)
Data ProcessingCombines batch and real-time processingFocuses solely on real-time processing
ComplexityHigher due to dual pipelinesLower with a single processing pipeline
LatencyBalances low latency (stream) and accuracy (batch)Very low latency with real-time processing
ScalabilityScales independently in batch and speed layersScales with a unified stream processing model
Data ConsistencyHigh with batch processing, real-time updates via speed layerConsistent real-time updates
Fault ToleranceHigh, due to separate layers handling different loadsHigh, streamlined with fewer components
Operational OverheadHigher due to maintaining both batch and speed layersLower with a unified stream processing model
Use Case SuitabilityIdeal for mixed batch and real-time needs (e.g., fraud detection)Best for real-time processing needs (e.g., streaming platforms)
Stateful Processing SupportLimited stateful processing capabilitiesSupports stateless processing
Tech StackHadoop, Spark (batch), Storm, Kafka (stream)Kafka, Flink, Spark Streaming

Conclusion

Lambda and Kappa architectures provide essential frameworks for handling big data and real-time analytics. Lambda architecture is well-suited for scenarios requiring both historical accuracy and real-time processing, offering a balanced approach through its dual-layer design. Kappa architecture, with its simplified focus on real-time processing, is ideal for applications that prioritize immediate data insights and require low latency. Choosing the right architecture depends on the specific requirements of the business use case, including the need for batch processing, stateful processing, and the volume of real-time data.

Azure Machine Learning Experiment Using Python

Azure ML studio provides 3 artifacts for conducting machine learning experiments.

  1. Notebooks
  2. Automated ML
  3. Designer

In this article, we will see how we can use notebooks to build a machine learning experiment.

From Azure ML studio, Click on Start now button on the Notebooks (or, alternatively, click on create new -> Notebook)

I have created a new folder TestModel, and a file called main.py from the interface above.

Structure of Azure Experiment:

It’s important to understand the structure of an experiment in azure and the components involved in successfully executing one.

Workspace:

Azure Machine Learning Workspace is the environment which provides all the resources required to run an experiment. For example, if we were to create a word document, then Microsoft Word in this example would be equivalent to a workspace as it provides all the resources.

Experiment:

An Experiment is a group of Runs ( actual instances of experiments). To create a machine learning model, we may have to create experiment runs multiple times. What groups the individual runs together is an experiment.

Run:

A run is an individual instance of an experiment. A run is one single execution of code. We use run to capture output, analyze results and visualize metrics. If we have 5 runs in an experiment. We can compare the same metrics for 5 runs in one experiment to evaluate the best run and work towards anoptimum model.

Environment:

The environment is another important concept in Azure Machine learning. It defines

  1. Python packages
  2. Environment variables
  3. Docker settings

An Environment is exclusive to the workspace it is created in and cannot be used across different workspaces.

Types of environments:

There are 3 types of environments supported in Azure Machine Learning.

  1. Curated: Provided by Azure, intended to be used as-is. Helpful in getting started.
  2. User-Managed: We( set up the environment and install packages that are required on compute target.
  3. System-Managed: Used when we want Conda to manage Python environment and script dependencies.

We can look at curated and system-managed environments from the environment link in Azure ML Studio.

To create an experiment, we use a control script. The control script decides workspace, experiment, run and some other configuration required to run an experiment.

Creating Control Script

A control script is used to control how and where your machine learning code is run.

from azureml.core import Workspace, Experiment, Environment, ScriptRunConfig

ws = Workspace.from_config()

experiment = Experiment(workspace=ws,name="python-test")

config = ScriptRunConfig(source_directory='./', script='main.py', compute_target="Test-Compute-Manan")

run = experiment.submit(config)

am_url= run.get_portal_url()

print(am_url)
Step Code Function
1 ws = Workspace.from_config() Connects to Azure Machine Learning workspace to access all ML resources
2 experiment = Experiment(workspace=ws,name=”python-test”) Experiment class provides a way to organize multiple runs under a single name.
3 config = ScriptRunConfig( This function is used to configure how we want our compute to run our script in Azure Machine Learning Workspace
4 run = experiment.submit(config) This function submits a run. A run is a single execution of your code.
5 am_url= run.get_portal_url()

 

Running the Experiment

Our control script is now capable of instructing Azure Machine Learning workspace to run our experiment from the main.py file.  Azure ML studio automatically takes care of creating experiments and run entries in the workspace we specified. To confirm what our code did, we can head back to our Azure ML workspace. It created an Experiment and a run. Azure automatically creates a fancy display name for a run which in our case is strong malanga. My first few runs failed because of some configuration errors. Running it for 3rd time marks a successful run for the experiment python-test.