Measuring Creative Operations
Preface: This is a post that I wrote for for the globaledit blog. I'm a huge believe in quantifying process and learning from other industries. After re-reading The Goal and looking at the Creative industry, I composed this: One of the books that has really shaped my thinking about creative operations is Eliyahu Goldratt’s The Goal. The basic plot is about a manufacturing plant manager faced with a failing organization and tasked with turning the operation around or the plant will close. The book is commonly used to teach introduction to operations classes and provides intuitive and practical examples on both how to measure and how to improve assembly line style work. Not everything in the book has an analog in a creative operations world but most of the basic concepts apply.
Creative operations isn’t an assembly line in the manufacturing sense but there are clear parallels. You have raw goods (camera files), you have to do something to them (retouch them), you have to QA them (approvals), and you have to ship the product (the final images). The process of creating a creative asset is just that – it is a process with a series of distinct steps.
Through that lens you can start to identify ways to create a high performance Creative Operations team. Your goal is to ship more assets. The first step (and the focus of the rest of this post) is to answer the question: “How is my team performing?”
Quantifying Success
I want to be explicit here: the intent of this measurement is not to create some sort of employee review metric. The intent is to help you understand where your team is and if your changes are having a positive or negative impact. We want to analyze the team as a whole, not the individual participants.
I’ve bucketed a few different metrics into two different categories: those which measure the whole team and those which measure individual steps.
1. TEAM METRICS: Asset turnaround time
Your team is charged with producing something. Therefore, one of the most basic measure is: How long, on average, does it take from the time an order is placed to the time that it is completed?
This measure is an easy-to-understand metric that should drastically change with an improved process. If an average time drops from 7 days to 5, you’re made a huge improvement. Who doesn’t want to get their results faster? It is an intuitive measure for tracking success.
To calculate this:
Look at all of your jobs for the last week. List each job in Excel, one per row. Now, in new column, add the date you received that job. Lastly, add the date the job was completed in a third column. For simplicity, I calculated this to the nearest day. You may want to do hours or weeks. The right level of resolution is highly dependent on the nature of your work.
A | B | C | D | |
1 | Job Name | Start | Stop | Duration |
2 | Job 1 | 1/1/2015 | 1/2/2015 | 1 |
3 | Job 2 | 1/3/2015 | 1/7/2015 | 4 |
4 | Job 3 | 1/3/2015 | 1/4/2015 | 1 |
Average (μ): 2 Days |
.
Now, calculate how many days each job took and take the arithmetic mean (=AVERAGE(D2:D4) ) of the them. Repeat this week after week and plot your weekly values for turnaround time.
Standard Deviation in Turnaround Time
This is a super useful metric that takes a just little bit more effort to compute. It shows you that the average turnaround time is 7 days (+/- 2) days.
This metric is important because it directly measures those times when you’re overworked and the times you have a light workload. The more you fluctuate between those two states, the higher your standard deviation is.
You want to move towards (+/- 1) day instead of (+/- 3) days.
A high deviation from the average means that your team is having very busy days and very light days. This game is about endurance; you want your team to work consistently for a long time, not in short bursts. Constant bursts wear out your team and reduce predictability in deadlines.
To minimize the deviation, you need a well-defined process, clear responsibilities, and mechanisms to identify who is overworked. This means that there should never be a question about “who needs to work on this?” or “what do we do next?” A simple diagram showing how the asset moves from start to finish and who is responsible for what will go a long way.
To calculate this: Good news! If you did the above work, you’re 99% done. In addition to average, in a new cell calculate the Standard Deviation of a Population (=STDEV.P(D2:D4) ). Don’t try to do this one by hand, it will be tedious. This should give you the average deviation from the average for that week. Remember: lower numbers are good.
One note on the above function – I am assuming you included every job in your calculations. If you did a sample of them (say every other one) then you should use the Standard Deviation of a Sample (=STDEV.S() ).
A | B | C | D | |
1 | Job Name | Start | Stop | Duration |
2 | Job 1 | 1/1/2015 | 1/2/2015 | 1 |
3 | Job 2 | 1/3/2015 | 1/7/2015 | 4 |
4 | Job 3 | 1/3/2015 | 1/4/2015 | 1 |
Average (μ): 2 Days | ||||
StDev (σ): 1.4 Days |
.
In the above example, the average job will finish in 2 days. About 80% of all jobs will finish between 0.6 days and 3.4 days. (70% comes from a Normal Distribution table)
Throughput
This is slightly different than the turnaround time. This answers the questions “How many jobs did we complete this month?” This helps you better understand how much work you can take on.
Of course, you want this metric to grow over time. There will be a natural limit to how much work people and your team can do. What is important is that it grows following your new process.
This metric is also useful for capacity planning. If you can constantly handle 50 jobs per month, and you project things to grow to 100 per month a year, then you need to think about hiring or outsourcing some of your steps.
To calculate this: This is the easiest to calculate. Just count how many jobs you did in a week and you’re done. Write this number down somewhere and plot it week after week.
2. INDIVIDUAL STEP METRICS Utilization
If you use the mental model of an assembly line, then all of the above metrics track the system as a whole. One last data point worth tracking is the utilization of each step of the process. If everything happens in order, then the most overworked step of the process is your bottleneck.
As any fan of Eliyahu Goldratt’s The Goal can tell you, the bottleneck is what limits the entire system. By tracking WHERE this is (and there will likely be one next time) you know where to add support. (Hint: The bottleneck will be the team with the large inbox and the light outbox).
One note on this metric: it isn’t a critique of any step. There will ALWAYS be a bottleneck and it may move. Not every person or step takes the same time to do something. The value is knowing where this is so you can add employees to help where it matters most.
To calculate this:
This is actually best solved by walking around the office instead of using Excel. You need to figure out who is slowing down the process. Who has the most work to do and who are other people waiting on to finish?
p.s. As good of a message as The Goal has, the plot is pretty abysmal. The updated version only exacerbates this. If you buy a copy, make sure you don’t buy the updated version.
p.p.s. This isn’t just creative operations. This concept has been applied to Software Engineering for a long time too.
Credit: Image used under Creative Commons / Flickr User Illusive Photography