Data-driven pricing sounds like something that requires a data scientist and an algorithm. For a small business, it requires a spreadsheet and three afternoons. The data you need is already sitting in your sales records; almost nobody looks at it, because pricing in most SMEs is folklore, "we have always priced it this way", "the market price is X", "customers will not pay more", none of it ever tested.

Let me be clear about what this article is not. It is not about algorithmic dynamic pricing, the airline-style systems that reprice by the hour. For an SME that is overkill, expensive, and mostly a distraction. What actually moves margin at SME scale is much more basic: knowing which of your prices are wrong, and in which direction.

That knowledge is cheap. Here are the three analyses I run with almost every trading or service client, in the order of how often they surprise the owner. Each one is spreadsheet-level work using data you already generate.

Analysis 1: discount leakage, which discounts actually moved volume

Most businesses give discounts constantly, promo prices, loyal-customer prices, "bulk" prices negotiated years ago, the sales team's standing 5 percent for anyone who asks. The untested assumption is that discounts buy volume. Sometimes they do. Often they just give away margin to people who would have bought anyway.

Pull 6 to 12 months of sales and, for each discounted transaction, record the SKU, the discount depth, and the quantity. Then compare volume in discount periods against normal periods for the same SKU.

You are sorting every discount into three buckets:

  • Worked: volume rose enough that total margin went up despite the lower price. Keep these, maybe even deepen them.
  • Neutral: volume barely moved. The discount was a gift. A distributor client found that their standing 7 percent for a group of "loyal" customers produced purchasing behavior identical to full-price customers. Removing it added roughly Rp 18 million per month of margin with zero customer loss, because the customers were loyal to service and availability, not to the 7 percent.
  • Toxic: the discount trained customers to wait. If your monthly promo teaches buyers to stock up cheap and skip the following weeks, your "promotion" is just a price cut with a delay.

The math worth taping to the wall: at a 30 percent margin, a 10 percent discount needs roughly 50 percent more volume just to break even on margin. Most discounts never come close. Nobody checks.

Analysis 2: margin per SKU, which products subsidize which

Owners usually know revenue per product. Far fewer know margin per product after the real costs, and the ranking of the two lists is never the same.

Build a table: for each SKU or service, monthly revenue, direct cost, gross margin in rupiah, and honest allocations for the costs that hide, delivery, payment fees, returns and damage, storage for slow movers, and labor time for services. Sort by total margin contribution.

Two findings appear almost every time:

  1. The hero product earns less than believed. The top revenue item often carries the deepest discounts, the highest delivery burden, or the most returns. A food producer I worked with discovered their flagship product, 40 percent of revenue, contributed 12 percent of margin once delivery and returns were allocated. The quiet mid-list products were funding the business.
  2. Some SKUs are genuinely negative. After honest cost allocation, a handful of products lose money on every sale. Sometimes that is deliberate, a traffic driver that pulls buyers in. Fine, but it should be a decision, not a discovery. If a money-losing SKU is not pulling any strategic weight, reprice it or cut it.

Once you can see who subsidizes whom, pricing decisions stop being ideological. Raising the price of a subsidized SKU by 5 percent is no longer scary, because you can see exactly what it currently costs you to keep it cheap.

Analysis 3: safe price tests, treating price as a hypothesis

The first two analyses read history. This one generates new data. The core mindset shift of data-driven pricing is treating every price as a hypothesis you are allowed to test, instead of folklore you inherited.

You do not need software for testing, you need a controlled setup and the discipline to measure. Safe testing patterns for an SME:

  • Test on new customers first. Existing customers have price memory; new ones do not. Quote new inquiries at 5 to 10 percent higher for a month and track the close rate. If it does not move, your old price was a donation.
  • Test one branch or one channel. Multi-location businesses can raise a price at one branch and compare. Online versus offline channels can carry different prices with a plausible reason, delivery included, for instance.
  • Test the tail, not the hero. Start with low-volume SKUs where a wrong price costs little, build confidence in the process, then move up the list.
  • Test bundles and sizes before naked price increases. Repackaging changes the reference point, a larger pack at better unit economics for you, and often meets less resistance than a visible increase on the identical item.

Rules that keep tests safe: change one thing at a time, decide the success metric before you start, close rate, volume, total margin, and set a duration, two to four weeks is usually enough for a fast-moving product. Then actually decide: keep, revert, or test the next step. A test without a decision at the end is just noise.

One honest caveat: demand for many SME products is less price-sensitive than owners fear, but not infinitely so, and competitor reactions are real. Which is why you test small and measure, instead of announcing a chain-wide increase and praying.

Make it a habit, not a project

The failure mode of pricing work is doing it once, capturing the gains, and letting folklore grow back. The maintenance version is light: a quarterly two-hour review of the same three views, discount performance, margin ranking, and the results of whatever test ran that quarter. Put it in the calendar with your accountant or your co-founder so it actually happens.

If your sales data is too messy to run these analyses at all, scattered across notebooks, cashier memory, and three versions of Excel, that is your real first project, and it is the same foundation everything else needs too. I made that argument in why your business needs a technology strategy, not just a website. And pricing pairs naturally with the collection side of revenue: a correct price invoiced late is still a cash flow problem, which is why automating invoicing and payment reminders is the sibling of this article.

The takeaway

Data-driven pricing at SME scale is three spreadsheet analyses and a testing habit:

  1. Audit your discounts: keep the ones that provably move margin, kill the gifts and the toxic ones.
  2. Rank SKUs by true margin contribution and find out who subsidizes whom.
  3. Run one small, controlled price test per quarter, hypothesis in, decision out.

None of this needs an algorithm. It needs the willingness to check whether the prices you inherited are actually true. In most businesses I have looked at, several of them are not, and finding them is the highest-margin work you will do this quarter.