Most owners set prices once, copy a competitor, or round to a number that feels right, then never touch it again. Data driven pricing means something much simpler than an algorithm: pull the sales history you already have in your POS or invoicing system and let it tell you which prices are wrong. I've watched a retail chain in Tangerang find close to 80 million IDR a year in margin they were leaving on the table, using nothing but a year of transaction data they'd never opened.
You don't need a data science team for this. You need three columns: item, quantity sold, and margin per unit, over the last six to twelve months. Sort by revenue, sort by margin, and look at where those two rankings disagree.
Start with the mismatch, not the average
The mistake most businesses make is pricing everything the same way, a flat markup percentage across the whole catalog. But your sales data almost always shows two groups worth treating differently:
- High-volume, low-margin items that customers buy constantly and barely notice the price of. These can usually absorb a small increase.
- Low-volume, high-margin items that only sell because of a discount you set once and forgot about. These are often underpriced relative to what customers would actually pay.
A simple worked example: say a bakery chain sells 40 different pastry SKUs. Data pulled from POS shows their top-selling item, a basic croissant, sells 3,000 units a month at a 12% margin. Their least-discussed item, a specialty tart, sells 80 units a month at a 45% margin but hasn't had a price review in two years. Raising the croissant by just 500 IDR (a price customers won't consciously register) adds 1.5 million IDR a month. Testing a 10% price increase on the tart, since demand is clearly inelastic at low volume, adds another few hundred thousand. Neither move required new customers or new marketing. That's data driven pricing in practice: small, evidence-backed moves that compound.
Look at repeat rate before you touch price
Before raising any price, check whether that item drives repeat purchases. An item with a high repeat rate is more price-sensitive to loyal customers who will notice a change, even a small one, because they buy it often. An item bought once or occasionally has much more room to move without anyone noticing or complaining.
This is where most "just raise prices 5% across the board" advice fails. It treats a commodity staple the same as an impulse buy. Segment first:
- Pull 12 months of transaction data by SKU.
- Calculate margin per unit and total volume per SKU.
- Flag items with margin below your category average but volume above it, these are your underpriced bestsellers.
- Flag items with high margin but declining or flat volume, these need volume, not more margin.
- Test price changes on the smallest, most isolated group first, and measure for 30 days before rolling wider.
Discounts are where margin quietly dies
If you're already collecting the data, look at your discount line separately from your list price. In our experience with a multifinance company's retail partner network, the printed price list was fine, discounts approved informally at the counter were the actual leak. Nobody tracked how often a 10% "loyalty discount" got handed out to first-time customers who never asked for one.
Pull a report of discount frequency and discount value by staff member or by branch, not just by product. You'll often find that one location or one shift is discounting 3x more than others, with no corresponding increase in sales volume. That's pure margin loss with zero return, and it's invisible until you build the report.
Build the habit into a dashboard, not a project
Data driven pricing fails when it's treated as a one-time initiative. The businesses that get compounding value from this build a lightweight monthly view: top 20 SKUs by revenue, their margin trend, and their discount frequency, refreshed automatically from the same POS or ERP data you already have. That's a smaller build than most owners assume, and it turns pricing into an ongoing five-minute monthly review instead of a guessing game repeated once a year. For more on why dashboards need to drive a decision rather than just look good, see Business Dashboards: For Decisions, Not Decoration.
You also don't need more customer data to do this well. The pricing signal is almost entirely inside transactions you've already recorded; see Customer Data: Collect Less, Use More for the same principle applied more broadly.
The takeaway
Data driven pricing isn't about sophistication, it's about actually looking at data you already own. Pull last year's sales by SKU, separate volume from margin, check your discount patterns separately from your list prices, and test small changes on isolated items before rolling them out wider. The businesses leaving money on the table usually aren't pricing badly, they're pricing blind. Fix the visibility first, and the pricing decisions get obvious.