His raw columns measured the wrong things. Computed variables fixed it without touching the data.
The setup
Saddleback Outfitters sells tents, packs, and the kind of jackets people save up for. Tom runs performance marketing and lives in the customer export. His raw data had the basics: total revenue, order count, signup date, total margin. Useful columns. But, as he’d learn, the wrong shape of useful.
The problem
Every time Tom clustered on his raw columns, the segments came out sorted by one thing in disguise: total spend. A customer who’d placed forty small orders over five years and a customer who’d bought two expensive tents last month landed in the same “high revenue” bucket, even though they’re completely different people to market to.
“The raw numbers measured how much, not how,” Tom said. “I didn’t care that someone spent $2,000 total. I cared what their average order looked like, how profitable they were, and whether they were new or loyal.” The columns that captured that didn’t exist in his export. And he wasn’t about to rebuild the source data in a spreadsheet every time.
The turning point
Tom used Divisio’s computed variables to derive the metrics he actually wanted, right in the app, without altering a single cell of his source file. In a few clicks he built:
- Average order value = total revenue ÷ order count
- Margin % = total margin ÷ total revenue
- Days since signup = today’s date minus signup date (tenure)
These were new columns calculated on the fly: ratios and differences layered on top of his data, not edits to it. Suddenly he could cluster on behavior and quality instead of raw totals. He segmented on AOV, margin %, and tenure, and the groups finally made marketing sense: new high-AOV premium buyers, loyal-but-thrifty frequent shoppers, low-margin discount hunters, and a healthy core of profitable regulars.
And it cost nothing: computed variables are on the free tier.
How he did it
- Dropped his customer export into Divisio and left the source file untouched.
- Built three computed variables: AOV (a ratio), margin % (a ratio), days-since-signup (a difference).
- Ran pattern detection on the derived metrics instead of the raw totals.
- Read the heatmap summary and saw segments split by behavior and profitability, not just size.
- Reworked his paid audiences around AOV and margin tiers.
The payoff
The discount-hunter / low-margin segment was the eye-opener: a group Tom had been bidding up in his ad campaigns because their total spend looked high, when their margin % showed they were barely worth acquiring. He cut spend against them and shifted it toward high-AOV premium buyers.
Same ad budget, better-shaped audiences, and blended ROAS climbed by a solid margin within two ad cycles, mostly by not paying to chase unprofitable customers.
“My data had the answer, just in the wrong units. Computing AOV and margin in the app, without rebuilding my export, completely changed who my segments said to chase.”
Tom, performance marketer at Saddleback Outfitters
Feature spotlight: Computed variables (Free)
Derive new columns on the fly (ratios, differences, logs) without touching your source data. When your raw export measures totals but you need to segment on behavior (AOV, margin %, tenure, rates), computed variables let you build exactly the metric that matters in a few clicks, then cluster on it. It’s on the free tier, because the right variable is half the battle.