Mapping process to spreadsheet models to build automation
"Everything Starts Out Looking Like a Toy" (#75)
Join smart, curious folks to get the Data Ops 📊 newsletter each week (it’s free!)
Hi, I’m Greg 👋! I publish this newsletter on finding data products and interesting data observations with the goal of finding patterns and future product insights. (Also, it’s fun.) If you need a background on how we got here, check out What is Data Operations?
This week’s toy: a Tetris reboot. Now you can waste endless hours playing Tetris in a browser, or compete with friends! This is a nice walk down memory lane. Now, if they could make Tetris maps that solved big data problems … Edition 75 of this newsletter is here - it’s January 3, 2022.
The Big Idea
A short long-form essay about data things
⚙️ Elevate work by automating (some of) your job
Automation is a scary thing for a lot of people. Autonomous driving threatens to take away many long-haul trucking jobs. Automated warehouse picking and packing removes a lot of manual labor and forces individual workers to work more like robots than people.
Knowledge workers face a similar challenge. Will machine learning or artificial intelligence replace what I know? It’s too early to tell. For tasks that get done over and over again and depend upon assimilating a lot of facts into patterns, it’s safe to say that machines and process will replace people. People will review information, investigate patterns, and attempt to make sense of what the machines are doing.
Can I learn how to think like a machine? You can try, but machines are much better than people at certain tasks. So why not learn more about how to make machines work for you and build an unfair advantage of tools that not only you can use, but that your team members use to standardize some of the analysis and assessment needed. As a knowledge worker, spend more time using and building your knowledge instead of calculating individual formulas and calculations.
Tucker Marshall, CFO at JM Smucker, says of automation: “It enables our people to be creative, but also unlocks the capacity for them to consider other ideas.” If Fortune 500 CFOs are pointing at automation to level up your thinking, shouldn’t you consider this as well? (Yes, it also saves them money and time … but that’s not the only reason they do it.)
What does automation mean for the typical person?
Automation doesn’t mean trying to model everything you do with a process. It means identifying the things you do most often that are very similar and using templates or other tools to make them:
uniform - if you’re doing a process that takes 10 steps, how do you make sure you’re doing every step every time? If some of the steps are critical, how do you identify them?
effective - is the process really repeatable? One way to check is to give your process or template to someone else and see if there are able to get the same results
transparent - what are the inputs you use to make a decision? What information changes from run to run, and is variable? Is it easy to understand how to read a process and understand what it is doing?
Automation could mean building a repeatable tool you use every day, or it might mean a playbook that only gets used during an important event (e.g. for onboarding a new customer). The point is to spend less time repeating actions that computers do better than you.
What tasks are good candidates to automate?
Ok, you say. I’m ready to start! What kinds of tasks make sense to automate, and how would I even know how to do that?
Decision Models are a great way to map your task to a spreadsheet. Here’s the basic idea. First, understand what’s fixed and what’s variable. Then, identify the constants that are harder to change. Finally, make some assumptions about your ability to achieve the goals in the spreadsheet, and create a way to map your actual results to your assumptions.
Variable items - things that change depending upon success - might look like this. If you are building a forecast of revenue, are you assuming a rate of improvement month over month or quarter over quarter? If you are building a Sales Development spreadsheet predicting leads needed to result in qualified opportunities, what are your assumptions on the percentage of people contacted who end up becoming customers? If you’re not sure, you could try a range of values that represent optimistic, neutral, or pessimistic scenarios.
Fixed or constant items in your model might look like the number of hours that could be worked or the number of calls that could be made in a day. They are the items that might be harder to scale or require significant cost to change (number of people who are in a role).
Speaking of automation: if you’re using data from a Saas system, it’s handy to have a solution that puts the data in the right places for you instead of cutting and pasting the data by hand.
Testing your model
If your Spreadsheet looks like number salad, it won’t be very useful to someone else. Once you have the basic logic in place, take a few minutes to simplify the display of your model and add some descriptive text to clarify the instructions.
A few steps you might try:
Use color to visually highlight important areas. You might use a high-contrast color like yellow to identify cells that take input.
Imagine a typical scenario for your model and add data that people will recognize (the “typical” number of calls or emails for a day in a Sales Development model, for example)
Backtest your model with real data. If at all possible, look at 90 or 180 days of past data and see if your model passes the test of looking reasonable.
Write very simple instructions to use this template on the first tab of your spreadsheet.
Now, find someone else who does the same job or function and ask them to try out the template you made. If they have no idea what to do, it’s time to rethink whether this is an effective tool until you are having valuable conversations about optimizing the process.
Once optimized, future you will thank you for saving you time every new run through the process.
What’s the takeaway? The bigger picture of decision modeling is to learn how to pose questions that fit a model that’s already been answered … here are four examples of typical problems and other places to find sample spreadsheets.
A Thread from This Week
A Twitter thread to dive into a topic
John Cutler shares a gold mine of thoughts on specific ways to notice if someone is good at product thinking. (bookmarking this🧵)
Links for Reading and Sharing
These are links that caught my 👀
1/ What’s behind door #2? - Are you good at estimating probability? Test yourself by reading this article on the Monty Hall problem and then assess how you did. Your expectation of probability may be different than the actual math involved. This is a very important insight for the next time you need to consider the likelihood something will happen compared with your assumption of how often that option will occur.
2/ When will Web3 take off? - Tim O’Reilly is a well-respected veteran of all things Internet, so reading his take on Web3 is instructive. O’Reilly makes a good case in this piece for a reasoned view of Web3. Crypto Bulls will say “ok Boomer” and dismiss his request that the systems that are built with Web3 increase the utility and speed of those we have today. Bears should pay attention that there is a “there” there with distributed contracts, new forms of money, and expanded opportunities to build new systems. O’Reilly is worth a read.
3/ Get better at automating text - If you want to get an excellent introduction to text manipulation using a library, check out Miller. It solves some of the core issues around changing and reformatting text files when they don’t quite match. Miller and tools like it are key to identifying the initial pass at automation when you’re not quite sure you need to build a production-quality solution but don’t want to build Apps Script in Google Sheets or manipulate a data pipeline. There’s some interesting to work to do here to make this easier.
On the Reading/Watching List
Some things to watch or read, in no particular order
Reading: A blueprint on Taking a Big Swing. Morgan Brown, Scott Tousley, and Natalie Rothfels explain in detail how to set up a “home run” in growth for your startup. As they explain:
💡 A home run is a step function change in X driven by a big swing innovation in Y.
X = revenue, WAUs, traffic, etc.
Y = product line, acquisition channel, customer segment, regional market, etc.
Not all growth tactics lead to a meaningful change. To find one, you need to focus on the levers of the business and select areas to test based on your current understanding of how things go. Once you find signal, you’re ready to think about how to take a big swing at the problem. This is a great blueprint to build that loop.
Listening: If you need an excellent rabbit hole of the past year in music, check out the playlist below from Fluxblog. Break out of the typical go beyond what’s typical of commercial radio (if you like this on the regular, be sure to listen to KEXP). I liked this list because it contained some artists and songs I’ve never heard before. It’s astonishing the amount of musical diversity that can happen in a year. (Request for product: bespoke playlists based on single songs, not built by algorithms).
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
I’m grateful you read this far. Thank you. If you found this useful, consider sharing with a friend.
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon