Remote, Hybrid, In-Office, Anywhere: Trends in Workplace Flexibility
The debate over returning to the office vs working remotely (whether that be home or anywhere else) has caused lively discussions over the past several years. Many companies have called for the full return of employees to the office while others have maintained hybrid or fully remote options. Both authors have transitioned from full-time in-office jobs to full or partial remote work and have found a balance that suits their individual styles. What about you? How has your workplace situation changed and where do you stand in the debate?
A growing body of research from the Working from Home Research Project and the Survey of Working Arrangements and Attitudes has received significant media interest. Additionally, the Bureau of Labor Statistics and the Brookings Institute have also published data on trends and preferences for remote work.
In this paper, we highlight the use of Multiple File Import, Workflow Builder, Graph Builder and JMP Live (including new features in JMP 18) to prepare, segment, visualize, and share the survey data. We also conduct our own informal survey in the JMP Community to show how the Community’s remote work trends and preferences compare to the larger public data sets.
Welcome. Thank you for joining us. Myself, Trish Roth and Wendy Tseng, are going to be presenting some demonstration of the use of JMP in analyzing some survey and workplace flexibility data over the next 30 minutes. Thank you for joining us.
As a little background, we selected this topic, one, I have a personal interest in the topic and over the last couple of years, there's been a lot of opinions and articles and discussion online and actually in business journals about workplace flexibility, whether it's good, whether it's bad, who likes it, who doesn't like it, why they like it.
We were looking at some of these articles, Harvard Business Review, Brookings Institute, McKinsey, Forbes magazine, and we thought it lent itself to highlighting some of the ways JMP can be helpful in preparing and analyzing data. We thought it would be interesting discussion. Everyone seems to have an opinion. We thought it would generate some discussion, hopefully, in the JMP community and at the JMP Discovery conference.
What we hope you'll take away from this is, one, a couple of interesting insights or facts about opinions and workplace arrangements and flexible workplace arrangements, and as well, obviously, to take away some tips and tricks and tools in JMP that Wendy and I consider to be somewhat foundational in terms of importing data, preparing data, data wrangling, our data preparation in order to then generate some graphics. We will highlight some new features as well as some customization options that you have in JMP to make your visuals tell the story you want to tell and look visually appealing.
We're going to break the presentation into two parts. Wendy is going to kick us off with some insights using some external data, and then we'll kick it back to me and I will provide some overview of how we use JMP to conduct a survey or analyze survey data that we collected from users in the JMP community and actually in the broader workplace based on our LinkedIn community. I will hand it over to you, Wendy, to kick us off with some of the external data.
All right, sounds good. In our first example, we're going to take a look at findings from the Work From Home Research Group. The Work From Home Research Group conducts a monthly survey of US residents with several questions focused on work arrangements and attitudes. One of the questions they ask is focused around the number of days per week working from home, so WFH. They ask employees how many days per week do they want to work from home, and then they also ask employers how many days per week are they planning for their employees to work from home.
This first graph you're seeing shows the difference between the two. When I graphed this, I was personally surprised at how small the difference is. Employees do desire to have more days working from home than employers, that's why all these numbers are positive. But that gap at the largest in this time series was not even a day and a half, and that was early 2021. Generally speaking, it's been declining to less than a day, but since July 2023, it's been steadily increasing but not surpassing a day yet.
This graph below shows how those differences were calculated. The yellow line is showing what employees desire, about 2, 2.5 days per week working from home. Then the blue line is showing what employers are planning between 1 and 1.5 days per week.
This graph at the bottom is the same graph as the second one, only I've increased the scale. That's to show you in the survey that Trish and I conducted, the respondents wanted to work from home 3.5 days per week. Our respondents, probably as a function of the roles that people have, were much higher than what the Work From Home Research Data was surveying.
Those are some interesting findings. Now let's dive into some JMP tools that I'll focus on. I'm going to highlight a couple of features in Workflow Builder, the ability to organize your workflow steps in these groups, and the ability to package a workflow with the file, and that's what we're posting to our Discovery Talk Community page, so you can run the workflow, and you would already have the Work From Home Research Data bundled in it.
I'm also going to touch upon new in 18 features. In Graph Builder, you can more easily label points by value in 18. Then I'll touch upon something that's called platform presets in JMP 18. We'll see it in Graph Builder, but it exists in many platforms in JMP. It gives you the ability to recreate graphs and recreate analysis very quickly by saving the templates as presets to your software.
Let's first talk about how the Work From Home Research Data is organized, and we'll touch upon the Workflow Builder features. The Work From Home Research Data is compiled monthly, and it's in an Excel workbook. If you go to their website, and you download this, it will be a single workbook with all of the latest data.
The data is organized in separate worksheets, so separate tabs inside of this workbook. Each tab really focuses on different sets of data. Workflow Builder's feature of being able to organize the steps really lend itself well to this data preparation because I could align each of the steps to the specific types of data.
I'll go ahead and open up the workflow here, so you can take a look at it. Here you can see the steps have been grouped and bundled together for ease of organization. Coming to the red triangle, I can export this as a workflow package. Again, this package is what saved out to our discovery paper URL. It will come with the Excel file already attached to it.
Let's go ahead and just run this section which we saw some of those initial graphs built off of. This particular set of steps focuses on the number of days per week that employees want to work from home, and then how many days employers plan for workers to work from home. I'll unselect the other steps so that we can run this workflow and just focus on this set of data and insights. I'll hit Play, and all of my previously recorded steps are playing, which included bringing in the data, joining it together, data preparation, and creating the graphs.
Let's take a look at this particular graph. We're going to talk about how to label the points with the values in this part of the demonstration. If you've used JMP prior to 18, you know that labeling points took a few more steps. Let's create this graph together, so you can see how it can be done in JMP 18.
I'll first start by graphing the date, and then I'll grab the difference between employees and employers. You can see my graph has gotten squished because I have a very long legend name. I'll start by hiding that legend. Then I'm going to change that smoother to a line. Then, when I come to this points control panel, you'll now see the new and 18 feature. I can simply come down here to this drop-down and say label by value. Very simple in terms of being able to label those points now.
The other thing in JMP 18 is what I refer to as platform presets. I'll clear this canvas here and let's start from scratch, so you can see how platform presets work. There are a few steps I took to create that original graph, but now I'll be able to do it by I'm assessing my platform presets in this menu in the upper right.
In this menu, I have a menu that's called My Presets, where I have created several graphs that I want to be able to access quickly. Specifically, we'll take a look at some of these versions I've saved out with line and points labeled.
I'll first start by dragging Date, and then we have that difference between employees, employers. Now, because that's a graph type that I want to access very quickly, I'll come to My Presets, and I've saved this out. Let's go ahead and make a selection. We'll select line plus points with one decimal place. That's how you can customize your graphs and then be able to recreate them very quickly by saving them as templates, as platform presets.
These are also things that you can share with others. You could share your platform presets with others by going to preferences and exporting your platform presets, and someone else can load them into their JMP instance.
Just to summarize, go back to my journal. In this example, we covered organizing a workflow by bundling steps together, creating a workflow package to share the workflow and files with other people, labeling points by value, a new and 18 feature in Graph Builder, and then this concept of platform presets, so saving out graph templates or your analysis templates to your JMP instance.
Let's go to example 2. Example 2 uses Bureau of Labor Statistics data. Since October 2022, the Bureau of Labor Statistics has added new questions to their CPS survey focused on teleworking and working at home for pay. Let's start with some of the findings.
I've specifically focused in on specific occupations, so all management professional-related occupations, in part because the bulk of our survey respondents fell into this category of occupations. Looking at the percent of weekly hours working from home in this time series, you can see at the peak, we're, I guess as a country, averaging 70% of weekly hours working from home in this professional category. At the lowest, it was 63, and right now we're about 66 with the latest data. Our JMP survey respondents came in higher, just about 70% of hours working from home.
We can segment this group of occupations into more granular occupations. Here you can see all of these occupations overlaid. Let's focus on the highest one, the orange represents, so they're hovering in the high '70s, '80s. These are your computer and mathematical occupations. Then, if we look at the lowest percentage, the pink line, that's education, training, and library occupations.
Taking this graph, showing it a slightly different way with each that individual occupation in its own mini graph allows us to see the individual trends better. In this view, we can see that most of the occupations are trending downwards in terms of the percent of hours working from home each week. But some of the occupations, the ones I've highlighted in yellow here have slightly different trends, mostly flat on average across this period, but with some ups and downs.
Let's go to the JMP tools that we'll focus in on with this example. In this example, you'll get to see a couple of more new in 18 features. Multiple file import now in 18 gives you the ability to control the Excel worksheet name that you want to bring in. Previously, you were forced to bring in all of the data in an Excel workbook, but here we can be selective. Then also new in 18 is this new Graph Builder type. This is actually the page layout, but we're able to wrap the graphs into a single view so that you can more easily compare them.
Let's first take a look at how the Bureau of Labor Statistics data is organized. Like the Work From Home Research Data, the workbook contains several worksheets. If we take a look at this picture, you can see there are several tables in this single workbook. That's why multiple file import is particularly useful in this example because each table consists of different information.
In this example, you'll see me bring in specifically table 4, which contains those occupation cuts. What makes this data a little bit different than the Work From Home Research data is the Bureau of Labor Statistics publishes a separate workbook for each month. We want to bring in all of these workbooks, include the file name because it contains the month of the data, but then we specifically just want to concatenate Table 4.
Let's first look at multiple file import. To do that, I'm going to initiate a new workflow to show you how you can use multiple file import in a workflow as well. Before I go to multiple file import, I'll go ahead and click Record here, and now let's go navigate to those files.
I'm browsing to the folder where the BLS tables reside. I'll add a file name column because I want these months to come in my final table. Then I'll come to these settings to show you where you can specify the specific worksheet name in JMP 18. Here I can specify that I want Table 4, and then I can customize where the data starts, where the column headers start, and then where the data ends so that all of my data across all of those months will get combined properly.
Now when I click Import, I'll get a single table with all of these files concatenated, but just with Table 4. Let's stop our workflow, and we'll take a look at this resulting table. Looking at this last column, you can see all of the files came together, and now I have a file name column that I can use to extract the month from.
Let's take a look at the name of this table that JMP gives it as a default. The default name is going to be alphabetically the first file, underscore, the latest file. The latest data that I had in my folder is February 2024, the 02/2024. You can see that here in my File Explorer. That's the default name. I'm pointing this out because we are going to replay this workflow, and I'm going to add more files to that file folder. This will be an opportunity for you to see how flexible Workflow Builder is.
Just as a reminder, I've recorded this step, and it specifically recorded for going up to February 2024. Let's go add some new data to my folder. If you choose to do this, you could go to the Bureau of Labor Statistics and update and get new tables as well and follow these same steps.
Here's my new data, and you can see I've got it up to July 2024. I'm going to copy and then paste it into this folder. Now the folder that my workflow is referencing has five more files in it. The last file is July 2024. Rather than play this workflow that we just built, we're going to play my finished workflow that includes the graphs. This BLS workflow here has that same import multiple files step. It's importing all of the table 4s in those files.
Now I get a pop-up, because it was expecting to see a table that had February and that 02, but it doesn't. Instead of just not working anymore, it's prompting me to select a new table. My new table has the latest of July. Again, it's referencing that table again, so it's prompting me again to tell it what the new table is. It's running through all of the data preparation steps and creating those graphs. We saw these graphs. It's updated to July.
Now we'll just focus on this I'll take this graph because this is new in 18. Let's build this graph together. I'll first put the date here and then the percent of weekly hours. Because this is just focused on those professional occupations, I'll first filter down to just those management and professional occupations. I have occupation here. If I put in the overlay, you can see this is a graph that you had seen in the very beginning when I showed some of the insights.
All of these are forced, obviously, to the same scale. When I move it to the wrap drop zone, I get individual graphs, but they are still forced to the same scale. If I move occupation to the page drop zone, I have individual graphs with individual scales. Then what's new in 18 is the ability to right-click and specify levels per row to wrap all of these graphs into a single view. I'll select 5 because I think visually that looks pretty good. I'll click Done to give us more real estate.
Then here you can see if I unshow title, you can read those names. This is the new in 18. I call it a page wrap, but it really is just that page graph with the option to wrap them all in the same view. Summarizing here. In this second example, we looked at new in 18 multiple file imports, additional feature to import specific worksheets. We also looked at this new Graph type in JMP 18 for Graph Builder. With that, I'm going to go ahead and stop sharing so that Trish can take over from here for the next example.
In the remaining couple of minutes, we'll take a look at the data that we collected in the survey that we conducted in the JMP community. Back to our journal. For reference for folks who haven't seen the survey, I put some clips of the questions that we asked. There's also a link, so it's not too late. If you want your voice counted, you are able to go out to the community and follow the link, and take the survey.
Actually, because we built a workflow, we'll be able to update this information. You can check back and see as more people respond to the survey, what that does to what the graphics look like and the inputs. We asked some demographic questions, we asked some preference questions. We tried to mirror a little bit of what we saw in the BLS research data, as well as the work-from-home data sets in order to try to do a little bit of a comparison. We asked, "Why? Why do people like working Why do people like working in the office or on-premises?"
I'm going to jump right to the survey. I've paced it into the journal, and we'll cover them looking at the workflow quickly. But a number of different techniques within Graph Builder for customizing your graphic design. I have some examples here. In this first example, it's just a standard bar chart, but we've organized it, the most common response at the top down to the least common responses. We've leveraged using the d for the categorical information for ease of readability.
Since some of the response information was a little bit lengthy text, it's much easier for a reader to just view it in reading style versus having a lot of wrap text in the X category. That continues in the journal. If there's a specific technique or something you want to look at in more detail on your own, we've provided some examples right here in the journal.
I am going to switch over to the workflow, which again, you will also get packaged with the information that we published out into our community space. Very similar to what Wendy showed, turned on my recording and did some of my steps. You can turn off the recording and move information into the workflow. You don't have to do it all in one fell sweep. I'm going to go ahead and just run it.
I have some breakpoints set up so that it will stop and allow us to have a look at what it's done. The first thing it did was it just opened the Excel file. We used Microsoft Forms to conduct the survey. Microsoft Forms makes that data available in a CSV file. Obviously, jump, file open, select the CSV or Excel file, and it can very quickly import that data.
I programmed some breaks because I wanted to remind myself like, "Hey, we generated a graph. You might want to capture this graph for later." I programmed in some stop messages in the workflow. It's like, "Okay, good." These are the different roles. There were 93 individuals who took our survey. It was anonymous. We did have some specific questions for students. We only had two students that responded to the survey.
For the purposes of this presentation, I removed that data, so I created a data subset just for students, and I parked that separately in a subset table. If we get more respondents who are students, we can display those responses. But the two students who responded did say having the option for flexible work arrangements was important to them.
Now you can see, because I subsetted the table, I went back and deleted the student responses, this employment role graph updated. Overall, we had 91 people that responded who were mostly currently full-time employed, a couple of consultants, and a couple of retirees or folks that were maybe in transition. Again, standard bar graph. We're not going to spend too much time on that one. I'm going to move this one off to the side.
In the end, basically through the sum of the import steps, what JMP has done is, obviously, created a data table, and we did a lot of cleanup. One of the things I wanted to point out in the workflow script is column recode. I think it's not as commonly used maybe as more traditional recode and the reason that we use that.
The survey, obviously, the questions were asked in a way for humans to understand how we wanted them to respond to the question. You can see an example of what recode looks like for columns. In the survey, the question was "For what employment role are you completing the survey? "That's a lot of text. When you go to make a graphic, that's just going to take up a lot of room, so we shortened them up.
I said, "Okay, call that column, Question 1, Employment Role," and so on, and went through those steps and cleaned up the column names. What that looks like in native JMP was basically to take the column under Columns, Column Names. You just select Recode Column Names, and you're presented with a recode window, and you can just manually say, "Hey, I want to recode this column."
When you say Recode, if you watch on the left-hand side, boom, it's recoded this column name. When you have the recording on in the Workflow Builder, it's going to record that script, which is the instructions for how to recode the column name. Obviously, I did that for a lot of different columns, but I only had to do it once, so big time saver. That's our first tip.
I'm going to go ahead. In Workflow Builder, the arrow shows you where we're at because I had programmed some stops. I'm going to generate another set of graphics. Again, the way that you do that is I can stop the recording, I can go through the Graph Builder steps to make a graph look the way I want, and then I can basically come into the graph once I'm finished with it and say save script to workflow.
I don't have to leave the recording on while I may be trying some different options. Once I'm satisfied with what I have, I can then tell JMP, "Hey, send the script for building that graph into the workflow." As another graph type that we wanted to highlight using a feature called Column Switcher. We had a bunch of demographic data, very similar bar charts. They're just saying, "Hey, how many people responded a particular way to these questions?" Rather than have them being different windows, we maybe just wanted to tab through them.
Again, in Graph Builder, if you wanted to build that, you would come in, and you would just select. Now, again, I'm putting my category on the Y-axis, but I'm still selecting bar chart, and I can order by... How do you say? I can do some cleanup. I can hold down the arrow key or hold down the Alt key while I select the red arrow. I don't want the Y-axis title. I don't want the legend. I'm going to turn them off. It gives me a little bit more real estate.
Now, before I say done, I want to add the other categories. I'm going to come again under the red triangle and go under redo because I essentially want to redo this graph, but I want to switch to some different columns. I'm going to select Column Switcher and then Industry, Occupation, and these are just to recreate what we have here on the screen.
Now I can say done, and it will close the control panel window and I can toggle through these different graphics. I can turn back on the control panel. Again, even in this is… I'm using JMP 17, there are some labeling features that are available as well in JMP 17. I have to tell its count data. I've selected percent since I already have an axis of count. Little bonus that two pieces of information in one graph, I can get the overall count and I can get what the percentage of respondents.
The other feature we wanted to highlight and remind people about was this caption box. If I say right click in the body of the graph, I can add a caption box. Then again, that will bring up a new tool panel in the Graph. Again, it's not a mean, it's really an N. I can right-click on it and I can change the location. I'm going to put it at the bottom since I have more real estate at the bottom. This is giving me a count, again, reminding me I had 91 total respondents to this survey. As I do that, it's going to say done.
Again, because I did the column switcher, these graphs were all linked together. One format, you build it once and then apply it to multiple columns. It can be a real time saver. That is Column Switcher. The other graph, we're going to close the histogram. We're going to show the bar chart.
I have another example of a histogram. I created a calculated column. We'd ask the question, "How often do you prefer to work remotely or from home? How often does your employer require you to work on business premises?" We were presently surprised to see that there was not a big disconnect.
Actually, a lot of people, there was no difference between what they preferred and what their employer required. Quite often, actually, people said their employer had no requirement, but there were definitely respondees who said, "I'm required to be in the office four days a week. I'd really prefer to only be there three days a week."
That's that delta, what's the difference between what my employer is asking and what I would prefer. We did some shading because if your preferences allow you to do more remote than what your employer is requiring, you're happy, so you're in the green when your preference is to do something more than what your employer allows, we shaded it red. This feature of a graph is called range shading. Once you've created your bar chart, let me go to access settings and demonstrate this feature.
One, I wanted it to be balanced. Obviously, there's only five days in the week, so the difference between what you wanted and what your employer required, maximum difference could be five, so I gave a little buffer, so balanced. Set the increment at one so that each category, since we were counting in days, each category is going to show up with a label. We don't need incremental ticks because there's not incremental days, it's full days. But the feature for the shading is shown over here. I think people may be familiar with adding a line.
I set the value at zero so that's no difference. You can customize the line width. But here, if we enable this allow ranges, I said, if you're between 0.5 and 5.5, so your preference is more than what your employer requires, I can select a range color. I selected green. Then I think 25% is a default. It's what saturation of the color do you want, but you can obviously customize this.
I created a label saying preferred, size is required. Then I can set a font type. Then there are some settings for where do I want this to appear. You can see here it's in line in the graph. It did take a little bit. You have some other options. You can play with it, it did take a little bit of trial and error to get it to look the way I wanted. But again, I like jump anything, click the arrows, click the red triangles, find your different options.
Once you're satisfied with that, you can create a second one, so similar process to get the red shading. That's how we ended up with this range shading in this basic bar chart showing percentage of values.
All I'm going to let JMP Workflow Builder run through the rest of the graphics.
One of the other techniques we used was value order and value shading. If I go to this data table and go to number 12, double-click on it real quick, it'll take us over. If we look at the column properties for this, what color is grass? This was a fun question. We got some interesting commentary. I have enforced a value order. It's not being ordered alphabetically. It's not being ordered by what category had the most respondents.
We wanted the response of purple. We learned a little bit about doing surveying. It was basically to make sure people were paying attention to what they're being asked and not randomly just clicking responses. In a formal survey, folks who didn't follow the directions, we would actually exclude your data. We did not do that in this case.
Once I've set an order I want things to show up in, I can also select colors. I just manually went into the color palette and said, purple should be purple, green should be green, and magenta should be magenta. How do I get that to show up in the graph? Again, standard bar chart. I just dropped. I turn on the control panel. You dropped what color is graph into the X category, but you can see I also have it over here. I'll remove it, so you can see what happens. I also have it over here in the color.
It uses my default colors if I don't specify anything else. But if I drop this over, now it's using that column information that I provided to say, enforce these colors, apply these colors. Again, we've used the caption box, we've used the value data table.
I'm going to let the rest of them, the less of them, render the last technique or feature I want to show is multi-response columns. I'm going to let this run. There we go. We're going to come thinking back to the data table, still running. We have had JMP created a whole lot of different graphics, and again, all the different graphics while the session is open, they're all being stored down here because I didn't tell the workflow to close them. I can easily navigate between different options.
One of the things I wanted This multi-response technique will end. Notice here now in the caption box, it's saying we had 307 responses, we only had 91 people complete the survey. How is that possible?
If you remember, if you took the survey, if we look at the survey, you can see if I go to that column, that within this column, there are multiple responses. This person selected visibility and access to company leadership as the reasons why they like to work on business premises. There's better equipment available at the office, there's better networking opportunities. Each one of these responses is separated by a column.
One tip in Microsoft Forms, Microsoft delimits multi-response items by a semicolon to colon. I did have to use the Recode feature to then change those semicolons to commas. If you go on to Recode you can traditionally just make manual entries, edits or Recodes, but I use the replace string function that I replaced all the semicolons with commas so that JMP would understand how to delimit the various options that people had selected Collected. Again, I recorded that so that every time we replay the script, or I get new data, it will automatically do that for me. That's how we get to 307 responses because people could pick multiple.
But then when you go ahead and do the graph, JMP understands, I'm going to categorize each one in their various groups. You can see of all the different categories, what were the most popular ones being on business premises, people like face-to-face collaboration and ability to network.
The very final, was again another use of color scheme. In this, how often are you typically meeting with your coworkers? If I go to this column in the data table, column 19, the numbering turned out to be important just for navigation. Again, if I come here, right-click on Column Info, I came to this column I set a value order, so it's not alphabetical from least frequent to most frequent interaction.
Then I selected value colors, but this time I selected a colors theme. JMP has a number of prebuilt colors, whether your data is sequential, diversion, qualitative. They're very well-designed, so I had selected this blue scheme. You have some ability if I wanted to reverse it, to put the most saturated color on the never, I can do that. There are a lot of ways to customize. But same principle as what color is grass, when you create the graphic, you then tell JMP, "Use that column to drive the colors."
One, it makes it a little bit more interesting to look at. This survey had 20-something questions. Wanted to make some variety in how the graphs look to keep the viewers' attention.
That multi-response piece, just to finish on that point, this was a communication message. Again, if you go to the column, if I look on my graph panel, you can see that some of the columns have a different than the traditional numeric ordinal character icon. It's like this little exclamation point. I'm just going to click on one. Again, if I go to Column Information, you have to tell JMP this column, what modeling type is it? It's a multi-response column. You select that here under the Column Information. Then, when you make a graph, it will divide up the different categories.
Let me go back to our workflow. In our workbook, journal that will provide a couple of different examples of the different techniques. Just lost our textbook here. Let me open it.
We summarize the techniques that we reviewed in this presentation, so you have this for reference. We welcome the discussion. We picked this topic because we thought it would generate some discussion. Find your nearest beach, work remote, celebrate the option that you have to do that. If you have comments or thoughts or want to take the survey, go out to the community and participate. We'd love to hear what you have to say. Hopefully, you learned a few things about JMP in the presentation today. Thank you very much.