Before we start….
There was quite a bit to this challenge, so I am covering it in one blog and one video. This one will look at the strategy and execution of the planned steps I laid out in a LinkedIn post and in the below photo shot; the separate video looks at the technical aspects of cleaning and manipulating the data, as well as assembling the visuals.
For anyone who has stumbled across this blog accidently, and is not familiar with the challenge, details can be found here.
So let’s begin!
As well as Maven providing the raw data from the remote work survey and some background on the contents, the below is what I took as the brief for the challenge, with the crux purposely highlighted in bold by the Maven team.
I used this as the basis for my overall plan for the challenge, which I quickly drafted up using the old school pen and paper method.
Now that was done, it was time to put the plan into motion….
Part A – Considering Outcomes and Goals
When reviewing the brief, I asked myself a few questions that would dictate my outcomes and goals.
- Why am I actually creating this report / dash?
- Who is it for?
- What decision needs to be made as a result of it?
- What can I do by way of design to facilitate that decision process?
Apart from wanting to compete in the Maven challenge, I am creating this report to summarize a large amount of survey data and analyze how remote work impacts productivity and morale among workers.
In terms of “Who?” and “What decision?”, I considered my key stakeholder to be the senior management at a generic company, and anticipated that they would require a high level assessment with a summary of the positive and negative affects of remote working on mainly productivity, but also morale, and how this would shape how employees felt about remote work. This would in turn help the company decide on what plans they would need to enact in a post-COVID world.
After mulling over these questions and answers, it was fairly straightforward in my mind that my overall outcomes would include:
- Outlining the basis for an ideal future remote working policy for a company
- Consider any implications (positive or negative) on worker morale and wellbeing
- Measure how productivity was impacted during remote working
- Summarize the above in a high level explanatory report
These aspects, as well as the fact the requirement was for a single page static report/dash, meant that firstly, I would be analyzing at a high level rather than a granular level. The only reason to go to a granular level would be to analyze a specific industry or demographic, which I did not consider part of my brief (rightly or wrongly).
Secondly, I would also be looking to produce and explanatory set of visuals rather than an exploratory set.
What’s the difference?
Exploratory vs Explanatory
The storytelling with data blog will define this more eloquently than me, but basically an exploratory report would allow the user to manipulate data through filters and slicers and search for their own answers to questions they may have. For explanatory reports, someone has already done the “exploration” on the data, and is presenting what are hopefully the key points with supporting data and visuals.
Part B – Metrics and Comparisons
Now a general approach has been defined, I looked at potential comparisons and metrics that would be used to determine company policy. I needed to keep in mind that if a company is going to base their policy on these, they need to be robust enough to allow a decision to be made.
You may think it’s strange as I haven’t mentioned looking at the data yet, so how would I know what metrics or comparisons I will be able to make?
While this is true, I believe it is good to have an idea of what the most appropriate metrics could be in order to fulfil your brief, even before you look at the data in any kind of detail.
I know I have two data sets, one for 2020 and one for 2021. The obvious metrics would be to try and measure productivity in each separate year, and then do a year-over-year (YoY) comparison. If data is also available, I could also look at a further comparison with “normal working” productivity. This will tell me whether remote working has had a positive or negative effect and whether that was sustained between 2020 and 2021.
Similar to productivity, I would look at changes in any issues that may affect morale or working conditions between “normal work” and the periods of remote work in 2020 and 2021. Again, if possible, I would like to see how the stats change over time.
I am guessing that the survey will ask people’s opinions of working remotely, and whether they are generally in favour of it and whether they would like to retain it as a benefit long term. If so, I could measure any change over time, or the ratio of people in favour versus those not in favour, and this could also feed into the overall future policy to be developed.
Part C – Reviewing the Data
With ultimate goals and an idea of the type of metrics I feel would fulfil the brief, it was time to look at the data itself.
This was a little dauting at first. It wasn’t the fact that each survey had around 1,500 respondents; rather it was that the 2020 survey had 73 questions, and the 2021 survey had 109 questions. Add to that, many of the questions were quite lengthy, and were the headings of each column. This made it quite difficult to know exactly what you were looking at!
My quick solution for reviewing? Copy and transpose the head row from each file into a new table. This makes it much easier to read, and easier to see how many of the questions are actually grouped together.
It also allowed me to add a few columns to each table aligned to the metrics. As I reviewed each years’ questions I could tick whether I thought particular questions may be related to assessing morale, productivity or an overall policy. Having this list also came in very handy as a reference when the data was eventually loaded into Power BI.
As an example of the transposed table, below is a sample of a few questions from the 2020 and 2021 files that would touch on productivity.
Also, to assist anyone that may try out the challenge, I have added a copy of a blank file that could be used to conduct this exercise.
Once I reviewed and listed out all the questions I thought were applicable to my goals, I was happy to move onto the next step and start some provisional review and analysis of the survey results.
Part D – Pull in Data and Develop Plan
There were two distinct csv files for the surveys, and I loaded them into Power Query and Power BI. The respondents in each survey are not linked, therefore there was no immediate need to link the files via a model.
On reviewing the questions and answers for each data set, I narrowed my focus to a limited set that I could build a plan around. This allowed me to create a storyboard and then further detail each aspect – all of which is laid out below.
I usually try to sketch together a outline plan or storyboard when I have reviewed the data and have a general idea of what I want to show and how I want it to structure and flow. There are usually a few changes along the way, as you will see in the final product. However, you will note here that I wanted to show a main headline to capture productivity, morale and the policy, and then expand on each aspect in separate sections.
Productivity – Self Assessment
I wanted to calculate the productivity ratings in 2020 and 2021 and compare. Both surveys had a similar question to ask how the employee rated their own productivity when compared to working in the workplace. As such, I could use it as an effective comparison.
The answers were categorised in a large number of % values displaying degrees of more, less or the same productivity. I decided to combine and reduce the number of categories to 5, as shown below in my sandbox draft version. I also wanted to show the % change for each category to emphasize the increase in productivity between the two years, also shown below. Text could be further added to further summarize and explain the point.
Productivity – Management Assessment
As the above were self-assessments, I wanted to develop the next level of narrative and to see if managers saw the same impacts on productivity. I was able to find a survey question directed to managers, to see how they rated the productivity of their own employees.
Productivity – Occupation
To complete my narrative on productivity, I wanted to go slightly more granular. The results so far were showing generally improved productivity, but for balance I wanted to see if there were particular occupations that were more likely to experience lower productivity. This could allow management to then potentially request further analysis if their companies had employees that had occupations with low rates of productivity.
Morale – Barriers
There were no questions in the surveys which explicitly dealt with morale; however, I noted that in 2021 there were a set of questions related to potential barriers to remote working, and whether people had improved, worsened or largely the same experiences for each barrier when compared to normal working. These barriers included issues such as:
- poor management
I took these as being related to morale. I consolidated these queries using an unpivot technique in Power Query, and again consolidated the ratings to a simple three ratings of, “worsened”, “same/no barrier” or “improved”. This allowed me to quickly see how each issue affected employees in 2021.
To supplement this, I noticed there were additional related standalone questions related to isolation, health and wellbeing. My plan was to bring these all together in an extended visual with a supporting narrative (see below draft result). It was also an opportunity to try out Gestalt’s principle of connection by boxing around some of the bar chart graphics in an attempt to connect them.
Morale – Working Times
From looking at the data and seeing that people were more active and feeling better, I thought then next level would be to examine the typical working day to see what impact remote work was having on people’s general activities.
There were a set of questions in 2021 related to hours people were assigning to daily tasks. I note there were similar questions in 2020, but there were significant errors in the arithmetic, which meant I could not make an equivalent comparison, and therefore discarded them.
I again performed an unpivot of columns in Power Query to consolidate the data, and found an average significant drop in time required for commuting (expected), and an increase in time for personal and family time, which you could potentially relate to increased wellbeing and morale. This was in spite of people actually working slightly longer workday on average.
Policy – Employee Expectations
Finally, I thought the policy section should provide the story arc to bring the productivity and morale factors together. I came across a question that I thought could knit it together – This was how much time employees wished to work remotely in the future.
Again, the answers were categorised into % values in intervals of 10, with some text answers also. I decided to consolidate these into day values to give a more simplified and easier to read view. This would show management that there was significant appetite from employees to work remotely, but that there was also a large number of people who wanted very minimal or no remote work. I highlighted this in red to purposely draw attention, and show there was a potential need for flexibility.
Policy – Retention and Attraction
Finally, and to again supplement the above with closely related data, there were questions to managers as to whether they believed offering remote working would improve employee retention and also attract new talent. Instead of using charts or graphs for these types of supplements, I decided to just use aggregated numbers and supporting text.
Stitch it all together
Once I had completed all my visuals in separate sandbox tabs, I copied them into a combined report/dash, which resembled the sketch I had shown earlier as my initial storyboard.
My initial draft was complete, and I added some summary headings to each section to highlight the key take-aways from each set of charts.
Part E – Review Outcomes and Goals
I felt that individual components of my goals had been achieved by each of the visuals I had included, as well as how they were grouped.
However, as part of my review, I took a tip from storytelling with data and Maven’s thinking like an analyst course, and used the opportunity to get feedback from a member of the Maven community, and also one of my subscribers to this blog (Marjolein Opsteegh).
After working with the data on and off for a few days, the intention was to step back and get a set of fresh eyes to briefly study the collection of visuals and see if it was relatively understood, and whether it flowed naturally and captured the brief.
Marjolein was generous enough to take the time to give some constructive feedback and criticism, which is a real learning process I actually enjoy. There were some general layout and formatting tips, but the crucial comment was to bring the overall outcomes and policy to the fore, and put them top and centre of the dash, rather than at the bottom which I had in my draft.
Part F – Fine Tune and Submit
After the review, I rearranged the running order of the dashboard, which made the summary and overall policy the first thing you see. I then followed it by productivity and morale, with some slight formatting on the headers.
Overall, I was quite pleased with the result, and was happy to apply some of the techniques I had been reading about in Cole Nussbaumer Knaflic’s storytelling with data book, a highly recommended read.
One change I might like to make after having the chance to look at it for a few days however, might be to increase the overall size of the visual, and provide a little more white space to give some of the visuals a little more breathing space from each other.
[UPDATE: As noted at the intro – I opted to support this blog with a walkthrough video of creating each of the visuals, especially those that used the unpivot and consolidation functions in Power Query] – for further work on Maven Challenges, stay tuned on LinkedIn, follow my twitter account which is notified when I write a blog, or alternatively just follow this blog.
Any feedback or queries, please drop them in the comments below – Thanks!