How We Built Contact Categorization (and Why It Changes Filtering)

When you receive an email from noreply@amazon.com, you instantly know what it is: a shopping notification. Your brain categorizes it without thinking. We wanted Cleanbox to do the same thing — automatically, for every sender, from the first email.

The database behind it

Cleanbox maintains a curated mapping of domains to categories. When an email arrives from linkedin.com, the system looks up the registered domain and assigns the category "Social Networks." When it arrives from uber.com, it gets "Travel." From stripe.com, it gets "Finance."

We currently have 20 categories: Social Networks, Shopping, Entertainment, Finance, Travel, Education, Food Delivery, Logistic Services, Business Services, News & Magazines, Fun & Activities, Gaming & Gambling, Home & Supplies, Internet & Communication, Discounts & Promotions, Restaurants & Dining, Events & Tickets, Productivity & Tools, Hobbies & Interests, and Adult.

The database is continuously expanded. When new services gain popularity, we add their domains. Users can also manually assign categories for domains not yet in our database.

Why 20 categories? Why not 5 or 50?

We tried both extremes during development. Five categories (Personal, Business, Marketing, Transactional, Spam) were too broad — "Marketing" from your favorite shop and "Marketing" from a casino feel very different. Fifty categories were too granular — users could not remember which category was which.

Twenty hits the sweet spot: specific enough to be useful in filters, broad enough to be memorable.

The real power: category-based filtering

Categorization is nice for the dashboard pie chart. But the real value is in filter rules. You can create filters that act on entire categories:

• Move all Shopping emails to a "Receipts" folder — Catches Amazon, eBay, Etsy, Zalando, and hundreds more, including shops you have not bought from yet
• Deny all Gaming & Gambling emails — One filter, zero tolerance, every gambling domain covered
• Mark Social Networks as read — LinkedIn, Facebook, Instagram notifications delivered but silent
• Flag Finance emails — PayPal, Stripe, bank notifications immediately visible

The key insight: category filters work on emails you have never received before. A traditional "from contains amazon" filter only catches Amazon. A category filter catches every shopping domain in our database — including new ones added after you create the filter.

On the dashboard

Your dashboard shows a "Top categories" widget with percentage breakdowns. This gives you a bird's eye view of what types of senders dominate your inbox. Most people are surprised by the results — Social Networks and Shopping usually account for 40-60% of all contacts.

For the full list of categories with example senders, see Contact categories explained.