Framework 79 Problem-Solving (McKinsey Approach)

Introduction

"...good problem-solving is a process, not a quick mental calculation or a logical deduction..."
Charles Conn et al 2018

This is a disciplined approach to hypothesis-driven problem-solving in change management that involves the application of
- hierarchical decomposition (breaking big problems down into smaller ones)
- step-wise refinement (start by describing what you want to achieve as general functions; break it down into more details which are redefined in successive steps until the whole is fully defined)

"...Problem-solving skills are becoming more critical and account for a higher proportion of workload due to automation and advances in technology such as artificial intelligence..."
David Maloney 2019

"...problem-solving means a process of making better decisions on the complicated challenges of personal life, our workplaces, and the policies sphere..."
Charles Conn et al 2018

To start the process with the following questions

"...1. How do you define a problem in a precise way to meet decision-makers' needs?
     2. How do you disaggregate the issues and develop hypotheses to be explored?
    3. How do you prioritise what to do and what not to do?
    4. How do you develop a work plan and assign analytical tasks?
    5. How do you decide on the fact gathering and analysis to resolve the issues, while avoiding cognitive biases?
    6. How do you go about synthesizing the findings to highlight insights?
    7. How do you communicate them in a compelling way?..."

Charles Conn et al 2018

The learnings from the constant iteration or trial and error are the key to making the process work. Linked with this is the importance of teamwork (including brainstorming) with a diverse group of individuals to reduce the chance of groupthink and cognitive bias while increasing the chance of more creative and appropriate solutions being developed.

Working in teams is an important part of this process, especially in work planning and analysis stages.

"...The best teams have relatively flat structures, good processes and norms, an explicit approach is necessary to avoiding bias..."
Charles Conn et al 2018

Some comments on teams

1. Team Structure and Leadership (the team leader or coordinator has to be like a musical conductor and air traffic controller, ie to make sure the basic elements come together and on time; for brainstorming and idea generation the best teams have little hierarchy which can be a challenge for larger organisations with conventional hierarchical roles - to handle this, set up the teams so that they are outside the normal reporting structures and have a limited life; any authority figures need to play a less direct the role in problem-solving so that the chance of new ideas being formulated increases)

2. Norms (they are hypothesis driven and end product orientated; oscillates frequently between hypothesis and data; look for breakthrough thinking rather than incremental improvements)

3. Behaviours to Avoid Bias and Error (there are around 100 different types of cognitive biases - for more details, see section on cognitive bias)

4. Diversity (choose team members with different backgrounds and viewpoints; this helps create an environment of creativity with an active openness to new ideas and approaches; diverse teams always outperform groups of individuals)

Some additional thoughts on problem-solving

"...Great problem solvers are well read, open to new ideas, reflective, self-critical, persistent......and use teamwork wherever they can..."
Charles Conn et al 2018

1. Multiply logic trees/cleaving (look for different questions and insights; don't settle on the first idea)

2. Challenge your hypotheses (active questioning moves our brain into discovery mode; continual questioning will sharpen and challenge the current thinking)

3. Use brainstorming practices to fight biases and promote creativity, ie

- encourage dissent (different points of view are allowed to be expressed irrespective of expertise, position in the hierarchy, etc)
- use role-play (put yourself in the shoes of other stakeholders)
- used different recording methods (drawings, mapping, verbalising, etc as alternatives to conventional discourse)
- dialectic standard (use sequence of thesis, anti-thesis & synthesis)
- perspective taking (modelling other team members' assertions or beliefs into a plausible proposal; this involves understanding the assumptions implicit in the perspective)
- voting (each team member has the same number of votes on an issue, challenge, solution, etc; senior staff to vote last, so as not to bias the choice of other staff members, especially those lower in the organisational hierarchy)
- solicit outside views (be careful of experts and people who are familiar with the topic)

4. Use good analytical techniques (on the financial side use net present value, margin analysis and use of cash flows rather than accounting book values)

5. Broaden your data sources (explore alternative data sources; realise the limitations of some data sources, ie challenge methodological issues in their collection like representative samples, etc)

Seven Steps

It is a 7 step recipe represented by a circular framework, repeat until you get it right.

sevensteps.jpg

1. Problem definition (get to the heart of the challenge(s), especially causes that are driving the opportunity or challenge; developing a coherent summary of the understanding of the problem; disaggregate the problem in a way into component parts)"...when problems' context and boundaries aren't fully described, there is a lot of room for error. The first step......is to arrive at a problem definition that is agreed upon by those involved in making a decision......test the problem definition against several criteria that are specific, not general......can clearly measure success......the decision is bound both in time frame and by the values of the decision maker......that involves definite action being taken. This step may appear constraining, but it leads to the clarity of purpose essential for good problem solving..."
Charles Conn et al 2018

Participants need to be very clear about the boundaries of the problem, the criteria for success, time-frame and level of action required. A scattergun approach to data gathering and initial analysis generally leads to wasted effort.

Sometimes
"...Existing players literally cannot see the threat from new entrants because their mindset, their problem boundary definition, makes them blind to it..."
Charles Conn et al 2018

This supports the approach of using fresh perspectives to bring breakthrough ideas to old problems, ie use people who aren't constrained by old models, mindsets, etc.

Problem definition or framing is best focused at the "highest level" possible, ie what makes sense for a single unit may not make sense for the company overall.

The more facts are the better in defining problem statements.

The good problem statement has the following characteristics:
i) outcome focused (clarity about the problem to be solved that is expressed in outcomes, not activities or immediate outputs)
ii) ideally specific and measurable
iii) designed to address decision-makers' values and boundaries (including the accuracy needed and scale of aspirations)
iv) encourage creativity and unexpected, different results (don't constrain the thinking)
v) provide holistic solutions

Use SMART (Specific, Measurable, Action orientated, Relevant and Timely) plus outcome focused and holistic.

Some techniques that help are logic/issue tree, weighted factor analysis, problem definition worksheet and design thinking (see techniques in ingredient 2). Usually it is best to start with simple estimates and "rules of thumb".

NB "...a well-defined problem is a problem half solved..."
Charles Conn et al 2018

2. Disaggregation (breaking down big challenges into smaller and more manageable chunks; involves repackaging complex challenges using a simplified contextual framework(s) laid out horizontally on a page, eg landscape orientation)
"...once the problem is defined, it must be disaggregated (or broken down) into component parts or issues......employ logic trees of various types to elegantly dissemble problems into parts for analysis, driving from alternative hypothesis to the answer. There is both an art and science in 'cleaving' problems - revealing their fault lines - that drive better solutions. This is the stage at which theoretical frameworks for economics and science provide useful guides to understand the drivers of your problem solution......try several different cuts at disaggregation to see which yields the most inside..."
Charles Conn et al 2018

Use the different frameworks to help with problem disaggregation and expose insights. Different frames often yield different insights.

At the same time, we need to be careful of our own cognitive biases, eg assuming what you have seen and worked in the past will be successful now:
"...when we encounter generally novel problems, and sometimes persist in using frames that are unhelpful or misleading in a new and different context..."
Charles Conn et al 2018

3. Prioritisation (ruthlessly prioritise the challenges; good prioritisation makes solutions come faster and with less effort)
"...identify which branches of the logic tree have the biggest impact on the problem, including which you can most effect, and focus your initial attention on these......employing a simple matrix of the size of the impact of each lever and ability to move the lever as a way to improving our logic trees. Prioritising analysis helps us find the critical path to the answers efficiently, making the best use of team time and resources..."

Charles Conn et al 2018

Prioritisation table, ie place factors in 1 of the 4 quadrants depending upon impact (high/low) and ability to influence (high/low)

Impact High 1
3

Low 2
4
    Low High
    Ability to
Influence


NB Obviously, you want to focus your attention on factors in quadrant 3 rather than quadrant 2

4. Work Plan and Timetable (build and regularly review plans and processes, ie daily review of what has happened; adjust your thinking and actions appropriately based on the reviews and learnings from frequent iterations)
"...Once the component parts are defined and prioritised......have to link each part to a plan for fact gathering and analysis. This work plan and timetable assigns team members to analytic tasks with specific outputs and completion dates.....best practice in work planning to move quickly and accurately to solutions......includes team norms around generating a diversity of views, use of experts, role-playing, and flattening team hierarchy to achieve better answers...... avoid common pitfalls and biases in decision-making..."
Charles Conn et al 2018

Some key elements of work plans include

- don't do any analysis for which you don't have a hypothesis, ie have an expectation of what the answer(s) are to the questions
- have a very explicit action plan (what has to be done, who has to do it, when will it be finished, etc)

Model Workplan


Issue Hypothesis Analysis Source Responsibility & timing End product
Definition Start with end points from logic tree
The definition of an issue varies from an important question to an unresolved question.
It is phrased so that it can be answered yes or no
Hypothesis is a statement of the likely resolution of the issue.
It includes the reason for answering yes or no
The analysis of a statement of the models that will be explored in order to prove or disprove the hypothesis that hence resolve the issue The source identifies the locations of means of obtaining data to undertake analysis The team member(s) and the timing of delivery of end product or intermediate output The design of the chart or table or other graphic that will show relationship or lack of relationship
Action Make sure each issue is stated in a detailed manner
Define sub issues where necessary
List of all hypotheses
- frontline ideas
- own ideas
- colleagues' ideas.
Discuss with team members
Refine hypotheses
Readjust priorities for analysis
Identify decision-making.
Determine the extent of analysis required
Simple case and/or
Complex justification
Identify readily available data.
Decide on methodology
Decide who will help collect the data and do analysis.
Decide on time frame, with milestones
Draw exhibits.
Develop storyline


NB need to be careful of people getting into premature analysis before understanding what output is needed to solve the problem, ie need to do the thinking on the underlying structure of the problem.
"...We encourage you to think hard about problem structure desired......numbers......If you set up your work plan carefully, you'll do the knockout analysis first, the really important analysis next, and......analysis last..."
Charles Conn et al 2018

Knockout analysis involves making estimates on the importance of a variable and its influence. Can use a prioritisation matrix. This knockout analysis will identify some lines of enquiry that are worth pursuing. The 80/20 rule applies, ie focus your work on the 20% of the problem that yields 80% of the benefit

Reduce the work plans by
"...driving from hypotheses, using rough and ready analysis approaches wherever possible, frequently diving into the data, and constant iteration is to refine the problem statement and the logic tree understanding......too short, but specific, work plans that focus on the most important initial analysis......constantly revise them as new insights come..."
Charles Conn et al 2018

NB Good work plans are a necessary but not a sufficient condition for problem-solving success, ie
"...Make your work plans chunky: short and specific; annual study plans lean: capture key milestones so you deliver on time......Short circuit hierarchy where ever you can - foster an obligation to dissent...... try role-playing for creativity to bring in outside perspectives......one-day answers help sharpen our hypotheses and make our analysis focused and efficient. Knockout analysis provides focus on the critical path in problem solving. Finally, there are a host of team effectiveness disciplines that...... can help...... produce great outcomes and guard against the pitfalls and biases..."
Charles Conn et al 2018

5. Analysis
(getting the critical data required for analysis, ie finding useful data within a set budget; having a toolkit to help with analysis; be objective; start with a thorough understanding of your data through techniques like graphing, visualisation and summary statistics)
"...data gathering and critical analysis is often a long step in the process. The speed and simplicity...... shortcuts or rule of thumb is to get an order of magnitude understanding of each problem component and to assess priorities quickly. This helps us understand where we need to do more work......when and where to use more complex analytic techniques, including game theory, regression, Monte Carlo simulation and machine learning......new online analytical tools make them much more accessible......summarise our best understanding in the form of situation, observation and initial conclusions - and team review sessions to pressure test these hypotheses..."

Some examples of Heuristics and Shortcuts Tools

Tool
What it is
When to use
Watch out
1. Occam's razor The simplest solution that requires the fewest assumptions Always Don't get committed to the first cut answer
2. Order of magnitude cuts: bounding the market What is the maximum potential value? Always. Never start a problem without doing this Don't favour the maximum. Need to look at the minimum
3. 80/20 thinking (Pareto principle) 80% of outcomes come from 20% of causes.
Find the most important factors
Scoping problem under constraint Don't miscalculate risk in large-scale, interconnected settings
4. Rule of 72 (Compound Growth) Divide growth rate into 72 to get doubling period Any growth or compound the problem Step change process. Change in growth rate
5. S-curve/adoption curve Typical model of adoption Adoption of new technologies and products Slow then rapid diffusion
6. Expected value (single point expected value) Expected value is the value of an outcome multiplied by its probability of occurring Any time you have a future uncertain events including value; sets priorities and reaches conclusions Other/better disruptive technologies. If distribution is not normal, ie skewed or longtail
7. Bayesian thinking Conditional probability, ie the probability of an event given that another event took place, ie prior probability When you need to think probabilistically Care to assemble data for calculations; difficult to precisely estimate prior probabilities
8. Reasoning by analogies Creating reference classes; have observed a particular problem and solution before When order of magnitude varies wildly depending on comparator Outlier cases; need right reference class; examples need to be relevant; past success does not guarantee future success
9. Break-even point Break-even volume Quick check on business model viability Knowledge of marginal and fixed costs; fixed costs with scale behaviour
10. Marginal analysis Financial determinants of the next unit; economics of producing more, consuming more, or investing more with limited resources Production, consumption and investment problems Step jumps in cost
11. Distribution of outcomes The possible range of outcomes, ie includes the worst Project cost estimates
New business revenue
M&A
Meaning reversion
Non-normal distribution
Reference point

(Source: Charles Conn et al 2018)

NB all the above tools need to be used carefully as a way to frame the problem and not be applied rigidly or blindly!!!!

Question-based problem-solving, eg root-cause (see elsewhere in this Knowledge Base)

i) Ask questions around "who, what, where, when, how, and why"

Can do this in a fan concept or decision tree format.
ii) Five Whys
iii) Fishbone

In summary
"... - start all analytical work with simple summary statistics and heuristics that help you see the size and shape of your problem levels.
    - don't get a huge datasets and build complicated models before you have done this scoping reconnaissance with rules of thumb
    - be careful to know the limitations of heuristics, project leader potential of reinforcing availability and confirmation biases
    - question-based, rough-problem-solving can help you uncover powerful algorithms for making good decisions and direct your empirical work (when required)
    - root cause and five why's analytics can help you push through approximate drivers to fundamental causes to a variety of problems and not just limited to production and operations environment..."

Charles Conn et al 2018

More complicated analysis (like Bayesian statistics, regression analysis, Monte Carlo simulation, randomised controlled experiments, machine learning, game theory, crowd source solutions, sensitivity analysis, simulation, etc)

Before embarking on more complicated analysis, you need to check that you have
- adequately framed the problem
- developed an appropriate hypothesis
- used some of the simpler techniques to help explore directions of causality and size and impact.

Then you need to ask
- Is the available data accurate and suitable to support using more advanced analytical tools?
- Which tool(s) is/are the right one(s) to use?
- Is there suitable user-friendly software available to help?
- Can this be outsourced?

Selecting the most appropriate analysis approach (using decision-tree)

nature.jpg

 

"...the most important defining question at the outset is to understand the nature of your problem: are you primarily trying to understand the drivers of causation of your problem (how much each element contributes and in what direction), or are you primarily trying to predict a state of world in order to make a decision? The first question leads you mostly down the left-hand branch into various statistical analyses, including creating or discovering experiments. The second question leads you mostly down the right-hand side of the tree into forecasting models, the family of machine or deep learning algorithms, and game theory..."

Charles Conn et al 2018

Some problems have elements of both sides, ie need to combine tools, and some tools like simulation and forecasting can be found on both sides.
There are some areas that you need to be careful of, eg

i) Correlation does not prove causation. For example, US data shows that education, income, walkability and comfort score are all statistically significant and negatively correlated with obesity. On the other hand, they are not necessarily correlated with each other (Charles Conn et al 2018)

ii) Regression models can be misleading if important variables are missing

iii) Too many variables included in the analysis, ie information overload

iv) The choice of analysis can have an impact on the findings

v) A sample must be representative and randomised

vi) Comparisons are best done against a control group

vii) Cognitive biases when analysing the findings

viii) Data needs to be of sufficient quality to be suitable for analysis, ie the analysis is as powerful and weak as the data used

ix) Handling uncertainty as it is virtually impossible to predict the future with any accuracy and assuming that the past will repeat itself is dangerous

6. Synthesising the Findings from the Analysis (finding the answers from the data as a basis for communicating the story and implementing the most appropriate action)
"...problem solving doesn't stop at the point of reaching conclusions from individual analyses. Findings have to be assembled into a logical structure to test the validity and then synthesised in a way that convinces others that you have a good solution..."

Charles Conn et al 2018

This involves another chance to pressure test alternative hypothesis. Sometimes new insights emerge that were not obvious in the early analysis

7. Communicate (telling a compelling story by providing the answers with the justifications based on the earlier steps; building trust with your client; towards evidence-based action that answers the following questions - what should we do and how should we do it?; use visual aids like graphics, diagrams, etc)

NB This needs to be combined with approaches like strategy, execution and transformation
"...the final step is to develop a storyline from the conclusions that link back to the problem statement and issues that were defined. A powerful communication will use a governing thought arrangement that derives from your refined situation - observation - conclusion logic from earlier stages......support with your synthesised findings and assembled into component arguments that may follow inductive or deductive logic. it will either lead with action steps, or pose a series of questions that motivate action, depending an audience receptivity......Done right, your conclusions are an engaging story, supported with facts, analyses, and arguments that convince your audience of the merits of your recommended path..."

Charles Conn et al 2018

It is a good idea to revisit the following
"...What problems are you trying to solve? How has this evolved? What are the key criteria for success that our decision-making......set out in advance? Did you know the boundaries of the problem set by the decision maker?..."
Charles Conn et al 2018

Some tools that can help include

- basic pyramid structure (helps understand how each element of our argument is supported by data, interviews and analysis, ie using situation - observation - resolution statement so that the question "what should I do", eg sequence

i) Governing Thought (your answer to the question in a single statement)
ii) Key Line (the core logic of your case)
iii) Support (on which your logic relies)

- structuring your arguments (there are several types, eg
i) argument structure (using facts and findings to define your problem, identify the causes and how to fix them, eg recommendations (including what needs to change))
ii) grouping structure
(using facts and findings to explain your recommendations and why you should change)

NB use deductive logic (from general principles via specific observations to individual cases) and inductive logic (from individual observations to general conclusions)

An example is in the hardware retail store business where Hechinger was competing with Home Depot

 

hechinger.jpg


".,.see the whole story in one page, with the governing thought and a call to action at the top ( provided by our concise situation-observation-resolution structure), the three major arguments made that underpinned the governing thought, and supporting arguments and data that provide the proof that need correction and the forming of the change. The next day from the storyline summary is to storyboard out your entire presentation..."
Charles Conn et al 2018

Most common pitfalls in the 7 steps

i) Weak Problem Statements (that lack specificity, clarity around decision-making criteria and constraints (like resources, time frames, accuracy, etc)

ii) Asserting the Answer (need to test if the solution is a good fit to the problem(s) at hand; need to be aware of the following cognitive biases in our decision-making, eg availability bias (focusing on facts at hand), anchoring bias (selecting an answer you have already seen), confirmation bias (seeing only data that aligned with your prejudices), sunk-cost bias (linked with loss aversion)

iii) Failure to Dis-aggregate the Problem (need to find the right cleavage points to disaggregate the problem so that able to focus on the crux of the issue)

iv) Neglecting Team Structure and Norms (importance of diversity of experience and diverging views in the group; need to be open-minded; need to keep away from group-think, ie usually happens when people are from similar backgrounds and a traditional hierarchy dominates; dynamics can be competitive and/or collaborative; need processes to reduce the impact of biases)

v) Incomplete Analytical Tools Set (some issues can be resolved with the "back of the envelope" calculations; others require time and sophisticated techniques like a well-designed real-world experiments that allow variables to be controlled, statistical analysis like regression analysis, etc)

vi) Failure to Link Solutions with a Storyline of Action (need more than analysis; need to synthesise and communicate concepts to the audience; need hooks to capture an audience and compel them into action)

vii) Treating the Problem-solving Process as One-off rather than an Iterative One (most problems have a messiness about them that requires oscillating back and forth between hypothesis, analysis and conclusions, each time developing a deepening understanding)

viii) Unable to Predict the Future with any Certainty (analysing the past does not guarantee predicting the future accurately; need to look at a wide range of possible scenarios and their consequences)

ix) Forgetting that humans are storytelling creatures, not logical robots

x) Learning to handle difficult audiences (need to work slowly with the decision-makers towards the appropriate conclusions, ie progressively reveal the answer rather than leading them to it.

In summary
"...great problem-solving consists of good questions that become sharp hypotheses, a logical approach to framing and disaggregating issues, prioritisation to save time, solid team processes to foster creativity and fight bias, smart analytics that start with heuristics and move to the big guns, and finally a commitment to synthesise findings that turn them into a story that galvanises action..."
Charles Conn et al 2018

(sources: Charles Conn et al 2018; David Maloney 2019)

 

Search For Answers

designed by: bluetinweb

We use cookies to provide you with a better service.
By continuing to use our site, you are agreeing to the use of cookies as set in our policy. I understand