A Different Approach to Categories

One of the design decisions we made about Tiller from the beginning was to not auto-categorize transactions in our standard template. In the age of machine learning, this may seem like a step backwards. However, our experience with auto-categorization (and many tools with this feature) is that it falls short in several aspects, and there are real benefits to manual categorization.

It's quick

First, manual categorization can be fast. The way our standard template is designed, it’s possible to categorize with a keystroke. Hit the first letter of the category, hit return, and voila, the category is entered and you’re onto the next row.

Build awareness

Second, manual categorization encourages us to pay attention. Take the water bill, for example. It would seem to be a great candidate for an auto-categorization algorithm, right? Perhaps, but by taking a fraction of a second to categorize it manually, I might notice that our water bill spiked this month. Maybe a leaky toilet or an irrigation leak? If I didn’t have to touch that transaction again in Tiller, I might not have noticed the change.

Align with your spending intent

Third, manually categorizing allows us to reflect on our purchase intent rather than simply describing the merchant. Take a simple visit to the local coffee shop. Tiller doesn’t know why I was there. On one occasion it might be a treat to grab a coffee and scone. The next time it might be a work-related meeting that’s deductible. That third trip could be for a gift, when I buy my favorite local coffee to give to a friend. Each reflect a different intent that I may want to capture with a different category.

More accurate

Finally, manual categories are accurate. We know that when we categorize, we get it right. At Tiller we’ve tried many tools that get auto-categories mostly right, maybe 80% of the time. Fixing the 20% that are wrong is a hassle. When the categories aren’t always right, we’re left wondering if a report or analysis is accurate. Bad data in, bad data out.

So we’re big believers in the benefits of manual categorization. We encourage Tiller customers to create their own categories that reflect their unique purchase intents. Start with fewer categories, as simple is usually better. Use unique first letters so keyboard shortcuts work faster. Don’t worry about your historic data, focus on today, tomorrow, and the future. Finally, take the time to categorize by hand. If you have a dozen new transactions, it will take you five or 10 seconds. That small investment of time pays big dividends with engagement. You’ll pay attention and feel in control, and that’s a win all around.