This Fine Tuning session is about improving your data analysis processes, including:
- How spreadsheets are keeping you from your actual job
- Freeing your brain for more effective interpretations of your data
- Creating effective spreadsheets when you need them
Before joining Zendesk as a Customer Success Consultant, Sam Chandler was a Zendesk Administrator and has utilized the platform to do everything from building out Customer Service departments to processing purchase orders. She's a former user group leader for the Atlanta area, and has presented on customer service & data topics in various webinars and in-person events.
See all the Fine Tuning series discussions.
Part 1, 8 am: Spreadsheets are not helpful
Does the image above look familiar? Yeah, I’ve been there too. In various points throughout my career, I’ve spent countless Friday afternoons compiling spreadsheet reports that would go all the way up to our C-level execs. And if I missed that Friday deadline, come Monday morning I’d have a trickle-down nagging on my hands that rivaled the likes of Bill Lumbergh.
The most disheartening part of this exercise was not the 2+ hours I spent preparing my report when I had a million other end of week tasks to wrap up. No, the worst part was knowing that no one was reading it. I know no one was reading it - couldn’t have read it - because to my senior leadership team my report was just a sea of black and white boxes. They didn’t understand the context connecting my numerical maze because they weren’t soaked in the data all day like I was.
Why, then - if no one understood my eyeball glazing, thousand yard stare-inducing spreadsheets - was it so imperative that everyone had them in a timely manner? Because too often the act of compiling data is mistaken for the act of analyzing data, and too often managers believe that asking for data is the same as asking their team to interpret that data. So when a customer tells me they want to use Insights as a way to re-create the Excel spreadsheets they’ve been circulating, I hope you’ll understand why I cringe a little on the inside.
I’m going to tell you a secret. And try not to hate me when I say this... Spreadsheets are not work; they are the preparation you must do in order to complete your work. While most customer service managers are spending hours upon hours compiling all of these numbers - pulling from different systems and data dumping into one monolithic spreadsheet - they’re mistaking that for their job. Being there myself, I can totally relate. The effort you’re putting into the project definitely seems like work.
But a customer service manager is not a data entry clerk. You’re here to shed light on root causes so you can make decisions to drive your company forward. In order to accomplish this task, you need to take this mound of data and analyze it for a “civilian” audience. Analysis and reporting - those are your true tasks! And when we study the act of using spreadsheets as a tool to analyze & report, here’s what can be said.
- Inaccurate (88% contain errors!)
- Outdated as soon as they’re sent
- Hard to understand/analyze
- Unwieldy as a collaboration tool
If everyone seems to share a secret disdain of spreadsheets, why the heck are we still using them? As Colin McGrew, Keboola’s Director of Sales & Product Marketing, points out: Spreadsheets are often viewed as convenient and inexpensive, thus making them a go-to option.
Don’t get me wrong - spreadsheets aren’t all bad. There are times when this type of structure is welcome. But how do you know if this is the right tool for your project? McGrew’s article also includes some great questions to assist with your deliberation:
- Are we dedicating a significant amount of time to bring together data from multiple sources?
- Is this a fairly standardized process we are repeating on a regular cadence? (Daily, weekly)
- Does data accuracy come up often in our business conversations?
- Are there multiple contributors and consumers of the information?
- Are we struggling to get team members engaged with the data?
“If you answer yes to even one or two of the above questions,” says McGrew, “it’s probably time to seek a spreadsheet alternative.”
Ok, if spreadsheets are out...what should I do instead?
Tell a story with your data. One of my favorite features of our Insights product is the ability to create dashboards. And if you’re using dashboards to store enormous spreadsheets, then you’re missing the point of this feature - and of your core responsibility as an analyst.
“Most humans think in stories, not in data,” says Daniel Mintz, Chief Data Evangelist at Looker Data Sciences. He adds that, “Handing over a table of numbers to a co-worker without much explanation because ‘I don’t want to prejudice their interpretation of the data’ does a disservice to us both.”
Don’t spend hours every week manually updating the same spreadsheet; create reports and let Zendesk analytics tools automatically update the data for you. Instead, you should be spending your time interpreting your findings for your audience. It's time well spent,” says Mintz, “because when you skip that last step, you often undercut all the good analysis you did. Good analysis presented poorly is just as useful as bad analysis presented well.”
Then how do you visualize your visualizations? My recommendation is to start at the beginning. What story are you trying to tell? Akash Mukherjee of Facebook’s Data Products & People Growth division, shares his list of five questions when trying to communicate any data:
- What question are you trying to answer?
- How big is your data in terms of volumes as well as richness?
- What is the data-savviness of your audience?
- What business domain does your question fall in? Is it HR, Marketing, Finance or something else?
- What are some pre-attentive processing biases that your specific audience has?
Chart types best practices
Now that you understand your audience, you have to figure out what instrument you’re going to use to play your beautiful data song. Below are some best practices on some chart types you know well and others with which you should probably become better acquainted.
Bar Chart - Used to compare quantities of different categories.
Ex: Use different categorical attributes as HOWs to slice built in metrics such as #Tickets or First Reply Time
Line Chart - Track changes or trends over time and show the relationship between two or more variables.
Ex: Ticket solves by week by assignee // # Users by month // # Users by Organization by month
Pie Chart - Compares parts of a whole and should be used carefully. Never compare two pie charts without clearly noting that the size of the pie may have changed as well. You should also limit yourself to no more than 3 variables to compart as a part to a whole to avoid distortion.
Scatter Plot - Shows joint variation of two data items. Scatter plots are useful to evaluate whether or not two variables are related. The more defined the linear arch made from the data points, the stronger the correlation between the your two metrics.
Bubble Chart - Show joint variation of three data items. In order to use it, you’ll need to have 3 metrics in the WHAT field
Ex: #Tickets vs. Full Resolution Time vs. First Reply Time
Bullet Chart - The bullet chart is useful for displaying a single key measure compared to a static target, such as a maximum or a quota.
Here’s a fantastic recipe that utilizes this format:
When you’ve got the proper tools for data compilation & visualization, you free up the brainpower necessary to take your analysis to the next level. Join us later today where we’ll discuss all the cool new data possibilities to fill your brain...
Now it’s your turn! In the comments below share your tips for determining the best way to share your analysis with your audience
Part 2, 11 am: So what?
When outlining expectations for our research papers, one of my college professors employed a simple phrase that would forever change the way I think: “So what?”
His point in saying this was to remind us that anyone could regurgitate a list of facts they found on Google. But to succeed in his class, we would need to take that compilation of facts and create our own interpretation of the results. (He would even go so far as to say “And then what?” if we hadn’t effectively proven our analysis by the end of our abstract. Imagine the terror in a poor little undergrad brain!)
You should approach your data like you’re writing a research paper. Using Zendesk reporting tools solely to create table charts to simply count the tickets that were solved last month is like using Final Cut Pro to edit a Snapchat video: look at all the power you’re wasting.
This type of exercise also breeds tautological arguments, or arguments that begin by assuming the very thing that is meant to be proven by the argument itself. I’ve worked with many a customer who have table charts that include columns with #tickets created and ticket solves by team, only to tally the total number of tickets in the last column. Not only will this exercise circle back on itself, but it’s incredibly limiting to your potential analysis.
If you’re only looking for these surface level metrics, you’re missing out on more complex interpretations of the data. For example, merely looking at your incoming ticket volume doesn’t say much on its own. Is this good? Bad? Same as last month? More than last year?
Here are some examples of how you can take your ticket volume metric to the next analytical level:
- How many of these tickets came from the same user in a given time period? How many channels did your customers use to reach out in that period?
- At what hours do you receive the most volume? What contributes to that?
- What are your highest volumes by issue type?
- When are you breaching your SLA goals? Does it coincide with your peak volume? Should you add more agents, change your hours, or update your processes?
- What is the age of the oldest ticket in your backlog?
- How much revenue is your team responsible for bringing in from their interactions with customers?
For many of these reports, taking it to the next level is as simple as exploring the standard metrics that are funneling into your account already. Something I recommend to all of my customers is taking a look at our Insights Metrics Reference Guide in the Zendesk knowledge base. This is a fantastic guide that details all the finer data points that you can take advantage of in addition to the standard 1st Reply, Full resolution stats to which we’ve grown accustomed.
A question I get a lot is, “What should I be measuring? What are your other customers doing?” I would love to say that there is a secret metric that we should all be monitoring...because I’d be a millionaire by now! But alas, the metrics you should be measuring are as varied as all of you.
Anytime I receive this question, I fire back with three of my own:
- What does success look like to your organization?
- What are the qualities of a highly productive agent?
- What are the elements of a positive customer experience?
Ostensibly these answers might seem like they would be standard across the board, when in actuality they will vary greatly by company and generally trickle down from the first question. For example, some organizations would consider a productive agent to be one that solves the most tickets while others don’t believe ticket solves tell the entire story, instead opting to focus on a more granular metric such as ticket touches.
Once you’ve determined what’s most important, you can begin to narrow down the metrics most closely associated with these values. Once you’ve gotten a baseline for where you stand with these metrics, then comes the fun part! Now you can determine why your metrics are where they are through analysis of the adjacent attributes. (Just like my professor taught us!)
To help jumpstart your creative juices, I’ve created a chart of metrics that many of the organizations I’ve worked with use to measure Agent Productivity, User Experience, and Cost Analysis, alongside some lesser-used metrics that I recommend looking into as well:
Let’s look at a sample workflow to explore the concepts we’ve just discussed.
Scenario: Fashionocracy, an online clothing retailer, is trying to strengthen their self-service offerings and they’re at a loss with what content they should create.
Any time I create a new report, I start with a question: “What am I trying to do?” This helps me to ensure my reports stay on topic. (This is also helpful because in the real world people approach you with a messy word problem to be unraveled as opposed to boxing it up neatly for you into your separate data points.)
What do I want to do: I want to identify which simple ticket topics can be turned into KB articles
How do I do this: I can use the #1Touch metric coupled with a custom “Issue Type” drop down to identify request types that are easily solved with one public ticket comment while still maintaining a high satisfaction rating. (Because if I can solve a ticket with a single comment and still garner a high satisfaction score, this tells me I can probably convert the answer to a self-service channel for an even more effortless customer experience)
Here’s what the report looks like:
WHAT: #1Touch tickets // %Satisfaction (both standard)
HOW: Issue type (custom)
Chart type: Bar chart with primary and secondary metrics
Configuration: Sort chart from largest to smallest based on %Satisfaction
Even when you have a bag of cool new tools like new chart types and metrics, you still have to illustrate your analysis to an audience who might not inherently understand how to interpret these charts. You can still point your audience to the correct analysis of your report through your mighty Insights dashboards. Take full advantage of your dashboard tools by:
1) Using headers and subheaders to show the audience the subject matter and give context of what the report is showing. In this example, the Report title clearly relays the objective of the report while the subheader adds another level of instruction to guide the viewer to the pertinent details displayed.
2) Effectively utilizing your dashboard real estate. In this example, a donut chart replaces a pie chart so that the middle can be used to house a separate headline report communicating further details on the primary chart’s subject matter.
3) Configuring your charts! Enabling features like primary and secondary metrics on your reports can have an enormous impact on your analysis.
I know what you’re thinking: “Yes Sam, these are amazing recommendations. But what about those times when I have to create a spreadsheet? Surely there must be something I can do to make this report type more effective.”
Stay tuned for our 3rd and final section to find out!
Now it’s your turn! In the comments below tell me how you determine the best metrics for your organization.
Part 3, 2 pm: If you must make a spreadsheet, make it a good one
I’m not naive. I know you will create another spreadsheet at some point in your life. But when that time comes, I simply ask that you make it an effective one. If your needs dictate that you absolutely MUST have a table chart report, there’s no reason you should have to spend your Friday afternoons manually updating a crazy matrix of indecipherable digits. Work smarter, not harder. Happy hour is calling after all!
Examples of effective uses for spreadsheets:
- Drill in reports
- Short lists containing a few columns of related attributes
- Exporting large amounts of data to input into other platforms
- Uploading user data points into Zendesk for bulk updates
- When your boss demands that your report be in a spreadsheet :)
When I must create a table chart, here are 5 of my favorite configuration features in Insights that have transformed the way I do it.
1) Conditional Formatting
The #1 absolute best (and easiest!) thing you can do to enhance your table charts is to add conditional formatting. This allows you to add pops of color and other formatting dependent upon a threshold you set.
You can choose from a number of colors and formats based on your content, but one of my favorite ways to use this tool is in conjunction with your SLA goals. For example, if your agents have an SLA goal of 40 solves a day, you can choose to display in red any agents who fall below that. The conditional formatting feature also allows you to add units of measure such as hours or dollar signs to your reports for easier interpretation.
2) Scheduling Sends
Did you know that you can schedule emails of your reports to go out at regularly scheduled intervals? Not only does this automate the process of report distribution so that you can actually take a vacation every once in awhile, but it also ensures that your data is as up to date as the moment the message was sent. (Note: To receive a scheduled send you must be a Light Agent or higher.)
3) Column header rename
By default, the header names on your table chart are the exact names of the metrics or attributes they’re displaying. But sometimes that’s not so effective if your metric isn’t “civilian friendly.” You can left click on your column header and update the name.
This feature is probably what my customers get most excited about. Very frequently, I hear tales of exporting Insights data into excel and creating a pivot table so that the columns can be tallied. But you don’t have to do that! By right-clicking the column header of any metric in your report, you can choose from a selection of mathematical operations that you can use to add totals or subtotals to your reports. (Additional tip: if you have more than 1 attribute in the how field you get the added option of including subtotals in your reports.)
5) Report filters
When dealing with large datasets and spreadsheets, it’s incredibly important to remove any irrelevant data to lessen the clutter. The numeric range filter is an oft forgotten report filter that allows you to automatically remove data points that don’t fall under the threshold you’ve set. Ex: “Only show me users who have created >2 tickets in the past 2 weeks”
Additionally, you might consider adding a ranking filter to your table chart since this format lacks the ability to sort by largest or smallest like some of your other chart types. Here are some great examples of reports that utilize these filters:
Reports Recipes using Ranking Filters
Reports Recipes using Numeric Range Filters
“How do I use Insights to duplicate all 1,000 columns of my Excel pivot table?”
This is not a race to see who can add the most columns to a report. I see customers get frustrated when they can’t create a table chart in Insights that looks 100% identical to their complex Excel spreadsheets. Most frequently I see this occur when an exorbitant number of columns are present in the report as it becomes more difficult to adequately make room for all your columns within the space given:
I like to view this as a built in gut check. There are only so many columns you can add to a report without them starting to stack on one another and get a little wonky. To me, this is an asset. Adding too many columns to a table chart is adding unnecessary clutter to a chart type that already isn’t ideal for visualizing data. Instead, think about whether the data in each of your columns is adding value to your data story. If not, take it out and add it to another report.
If you take nothing else away with you I hope it is this: keep striving to be better. Sure, you could crank out the same ol’ spreadsheet report you always have with very little thought behind the numbers just so you can mark a task as complete. But how is that progressing you forward? You have access to some incredibly powerful data - use it. And with just a pinch of curiosity, you have the ability to not just report on past behaviors but to have a say in the future path of your organization.
How do you jazz up your spreadsheets to highlight your main points? I’d love to hear your thoughts and suggestions on this section and any of the others!
Post ist für Kommentare geschlossen.