What is Forecasting?
Mar 26, 2020 • Eva Humphrey
At its most basic, forecasting means using historical data and educated assumptions to predict a future state. At East Fork, we use historic sales to determine our typical % of Revenue is for each of our forms and glazes. If that's not enough to get you to click in and keep reading then we can't help you...
Hello, hello! Please give it up for our Forecaster & Analyst, Eva Humphrey, as she gives you a good look into the wild world of using data to figure out how much of what to make and when. Take it away, Eva...
What is Forecasting?
At its most basic, forecasting means using historical data and educated assumptions to predict a future state. At East Fork, we use historic sales to determine our typical % of Revenue is for each of our forms and glazes. Then we combine this average breakdown with the revenue projections determined by the finance team to estimate how many of each pot we can expect to sell in the coming weeks. We add a buffer to that quantity because we don’t want to sell down to zero, and another to account for sales while we work to get the pots to our channels. Then we subtract the pots we currently have on our shelves (and pots waiting to be QC’d - less our average second's rate and pots currently in transit) to get the list of pots we should glaze. Then a bunch of lovely, hardworking, skilled folks glaze those pots and send them to our warehouse & stores.
Pie chart * $$ signs on a magic 8 ball + pie chart * calendar w/ 2 weeks highlighted - pots on a shelf = order ticket
Pretty simple right? So why does it seem like we’re always out of stock? Good question! There are a couple of complications that make it hard for us to keep everything on the shelves.
- We make our own pots! This seems sort of obvious, but we can’t just call up our suppliers and say, “We’ll take twice as many next month, thanks”. Alex wrote a nice long article about this, but our total production capacity is limited and builds fairly slowly. New machines take time to arrive & be set up, it takes time to find and train new people. This reality drives all other trade-offs and decisions. Fun fact, in January and February of this year, we sold a dollar value of pots equal to 94% of the value of firsts quality pots we produced. Not a ton of pots leftover for stock levels to recover from holiday sales.
- Seasonal & limited-release glazes are central to who we are as a pottery business. They’re how we attract new customers and keep our favorite people happy and coming back for more. They also mean that we spend months making and stockpiling fun, special pots before they get their shining moment—and then disappear in a few hours. Sometimes that means we won’t have some forms, like mugs & dinner plates, available in our core colors for a couple of weeks. These seasonal builds are always a balancing act; we want everyone who wants a special pot to have the opportunity to buy one, but we also want to keep core pots on our shelves as much as possible.
^ We use this spreadsheet to map out our seasonal builds and estimate how much of our production capacity will be allocated to each build and each sales channel for every week of the year. Of course, we always adjust in real-time.
- Seasonal builds also introduce extra unknowns. We make educated guesses that our new glazes will perform similarly to our cores, but sometimes that isn’t quite right. For example, we pre-sold twice as many beach party ice cream bowls and a lot less beach party dinner plates than we would have guessed from our usual revenue breakdown by form.
- Our historical data isn’t perfect. When we’re stocked out of a particular pot for a few days (or a few weeks) it lowers our calculated % of revenue for that pot below its natural potential. If we don’t catch this, we under-order, stock out again, and start the cycle over. To account for this, we order based on a “target” % of revenue by form & glaze that’s based on a long time horizon (about 6 months) and gut-checked by our CFO. We also add customer special requests on top of our calculated store orders and are working on adding to our eCommerce order based on a count of active back in stock notification requests.
- Production takes a nonzero amount of time. RAM formed pots are made in big batches, Jiggered pots in smaller batches, and we can only glaze what we have available in bisque. Over the past couple of months, we’ve started paying more attention to this secondary forecast, predicting what pots we need to form to ensure we will have the correct bisque on hand to glaze the weekly order. Since different forms have different batch sizes and different loss rates from forming to bisque, this is a bit of a puzzle for the forming team leads, but it has gotten much easier (recently) as we’ve improved tracking of in-process inventory.
^ For example, this graph of our in-process and on-hand inventory from March 16 indicates that we needed to focus a little more on Juice Cup & Soup Bowl production on the Jigger team and it was time to press more Everyday Bowls, Cake Plates & Dinner Plates on the RAM press.
^ At the same time, we could see that we had enough bisque available to fill the glazing order for all the potters bowls, breakfast bowls, tot cups and mugs, but we were short on bisque to fill orders for everyday bowls, dinner plates, side plates and juice cups.
- Sometimes our production capacity distribution doesn’t quite match our sales breakdown by form. We can shift production capacity among forms with the same production method, but not between methods, and because we grow capacity one area at a time, our capacity allocation is typically just a little mismatched from our target form spread/revenue breakdown. For example, we recently bought another RAM press, because while we could keep our Jiggered/Jollied forms in stock, we just couldn’t keep up with plates!
- Production challenges/hiccups (i.e. most of one team gets sick, a whole run of plates slump because they were trimmed too wet, a mysterious uptick in bloating, we have to close all of production down to help contain a pandemic, etc.). It’s like driving a busted grocery cart, or one of our big QC tables covered in pots across an uneven floor. There are rules, you can plot a course and steer, but it takes some muscle, and there’s enough of a delay between action and effect that you’re always holding your breath a little.
- Sometimes we know we have our inventory in the wrong place, but it takes an agonizingly long time to course-correct. For example, our average inventory breakdown in February really didn’t match what we wanted it to be, but we were having a hard time getting PL-3 firsts, and that drove down production of all other RAM pressed forms (SB-2, PB-1, PL-2, PL-1).
Yes, all of our spreadsheets and dashboards are color-coded in accordance with the East Fork brand book. We like things to look good!
And if you're curious about what software we use...
We use Shopify (& TradeGecko) to track completed inventory and make sales, but pretty much everything outside of that is managed with Google tools. We use an ELT tool called Blendo to transfer sales & inventory data from Shopify to Google BigQuery and then we rely on Google Sheets, Google BigQuery, Google Data Studio & Google Apps Script to wrangle that information and extract useful insights.
Just hopping on here to second what Jenna had to say – I work in data & analytics, absolutely love your pottery, and this post made me smile :) Thanks for doing what you do, and sharing with us what it looks like to make it all happen.
As someone who works in data & analytics and also loves your pottery, this post makes me very happy :) Love your transparency and what this company stand for!