Rule-based repricing applies logic you wrote and can read back afterwards: floors, ceilings, and instructions such as match the cheapest stocked competitor but never go below cost plus 12 per cent. Dynamic pricing lets a model set the price from demand signals, and you accept that you cannot always explain one specific number.
Most companies need rules. Most pricing losses come from stale prices and prices set too low, not from a shortage of mathematical sophistication.
The definitions, without the marketing
Rule-based repricing is deterministic. Same inputs, same output, every time. You can write the rule in a document and a colleague can predict what it will do on Friday.
Dynamic pricing is a model adjusting price against demand, inventory and time remaining. Airlines and hotels have run it for decades because their inventory expires. A room unsold on Tuesday night is not sold later, it is gone.
The two are not opposites. Almost every real dynamic system is a model wrapped in rules, because the rules are what stop it selling at nine euros on a bad Thursday.
When rules are enough
- Your catalogue is large and each individual product sells infrequently
- Your costs are known per unit and change on a schedule, not continuously
- Your inventory does not expire, so no deadline forces a discount
- Buyers are professional and expect price stability rather than movement
- You need to explain a price to a customer, a regulator or a key account, which covers most of B2B distribution, pharmacy and construction supply
Transaction frequency decides more than anything else. A catalogue of 8,000 products taking 200 orders a day means the average product sells once every 40 days: 8,000 divided by 200. You cannot estimate a demand curve from one sale every 40 days. A model asked to try will find patterns in noise, and it will report them confidently.
When dynamic pricing earns its keep
- Inventory with an expiry date: hotel nights, flight seats, event tickets, short shelf life stock
- High traffic on individual products, so demand is measurable within days rather than quarters
- A market where customers already expect prices to move, which is a cultural fact about your sector, not a technical one
- An organisation that can run and read experiments, because a model without measurement is a rule set with worse documentation
If three of those four are false, buying a model buys you complexity and nothing else.
The precondition both approaches share
Neither approach is safe on top of wrong matches. A rule that says match the cheapest competitor will happily match your 1.5 litre bottle to their 0.75 litre bottle and halve your price. It will do it at 03:00, and it will do it to every product in the family.
Before automating anything, sample your matched pairs and count how many are wrong. If it is above 2 per cent, fix matching before touching prices. This is where price intelligence tools, PriceRoom included, spend most of their engineering, and it is the part worth auditing yourself rather than taking on trust.
A rule set worth copying
A starting shape, in evaluation order.
- Hard floor: never below landed cost plus your minimum margin. Not a variable, not negotiable.
- Reference: the cheapest competitor who is in stock, has a verified match, and was observed in the last 24 hours. Ignore everybody else.
- Position: hold inside a band, for example between 0 and 3 per cent above that reference, rather than aiming to undercut it.
- Ceiling on movement: never more than a set percentage away from your own previous price without human approval. This catches parsing errors before customers do.
- Damping: at most one change per product per day, and no change smaller than an amount that would plausibly affect a purchase decision.
- Exceptions: a written list of products the rules do not touch, each with an owner and a review date.
The band in rule three is where the money sits. Being 2 per cent above a competitor loses far fewer orders than most teams assume, and it is cheap to test before accepting that parity is required.
Costs nobody budgets for
- Somebody owns the rule set. If nobody owns it, it will still be running unchanged in two years, against costs that have moved.
- Every price change is a change to feeds, marketplaces, printed material and open quotes. Fast repricing with slow downstream systems produces visible inconsistency.
- Frequent movement teaches customers to wait. In some segments that is acceptable. In B2B contract business it damages the relationship.
- Model-driven pricing needs a written answer to why this price when a large account asks. Prepare it before you need it.
How to decide this afternoon
Three questions settle it for most companies.
- Does my inventory expire on a date? If no, rules are almost certainly enough.
- Does an average product see enough transactions per month for demand to be measurable? If no, a model has nothing to learn from.
- Can I explain my current prices today? If no, fix that before adding a system that makes prices harder to explain.
A maintained rule set with correct matches and fresh data beats a sophisticated model fed by stale, mismatched inputs. That comparison is not close, and it is the one most companies are actually facing.