What can AI do for a lumber yard's pricing and stock?

In pricing, AI can suggest a price for each quote line, learned from the rep's own past quotes and set against today's replacement cost, and let the rep override it where they know better. In stock, it can set reorder points and order quantities from what suppliers actually deliver, when they actually deliver it, and from recent demand. Both are more modest than dynamic pricing or a forecasting engine. Both have better evidence behind them.

What the usual answer says

The usual answer describes AI tracking stock across thousands of SKUs, forecasting demand, generating purchase orders automatically, and turning emailed requests into quotes. Each is possible. Little of it cites evidence from a real sales floor.

Pricing: the field evidence

The best field evidence comes from a distributor much like a lumber yard. Yael Karlinsky-Shichor and Oded Netzer, in Marketing Science in 2024, worked with a business-to-business aluminium retailer. They built a model of each salesperson's pricing from their own past quotes and showed its price in real time during an eight-day experiment. They found that "reducing intertemporal behavioral biases by providing the model's price to the salesperson increases profits for treated quotes by 11% relative to a control condition". That came to over $26,000 in the eight days, or "over $1.4 million when extrapolated yearly", an extrapolation from a short test.

Two limits matter. Compliance was low: in the earlier 2018 dissertation on the same experiment, reps adopted the model's price on about 10% of treated lines, and "complied more when pricing for frequently contacted clients or for frequently purchased product categories". And people did better on some quotes. Using counterfactual analysis, the published paper finds "salespeople generate higher profits when pricing out-of-the-ordinary or complex quotes", and proposes a hybrid that sends each quote to the model or the rep.

For a lumber yard, that maps well. Commodity items for regular contractors suit the model, and that is where reps in the study were most willing to use it. Special orders, unusual items and new accounts suit the rep.

Pricing: tie it to replacement cost

A model of a rep's own pricing also learns the rep's habits, including any under-pricing to favoured accounts. And on moving lumber markets, history anchors on old prices. So the suggestion should rest on what the next load will cost, not the average of what is in the yard, and should be compared with what the yard's best-performing reps charge for the same item. A price suggestion built on stale average cost will be consistently wrong.

Stock: the lead time is the weak input

Replenishment rests on two numbers: expected demand over the lead time, and the lead time itself. From Quarri's own work with a lumber and millwork manufacturer: under 4 in 10 purchase orders arrived on time, across 13,000+ closed orders, and late ones averaged about nine days.

Check how your system holds lead times. If it holds the supplier's promise, a reorder point set on it runs short whenever the promise slips. The fix is in the purchasing records: order date and receipt date for every line, by supplier and item. From those, set lead times per supplier as they actually are, including how much they vary, and put safety stock where a supplier is unreliable rather than everywhere.

How to roll it out

Start with pricing on the busiest commodity items, where quote volume is high and the model learns fastest. Show suggestions to reps for a few months without asking them to follow them, and compare margins on quotes where they did and didn't. Measure overrides too: where reps override often and win, the model is missing something they know. For stock, pull a year of purchase order and receipt dates, compute actual lead times by supplier, and reset reorder points first for the items with the largest gap between promise and reality.

What AI adds on top

Demand forecasting helps most for items with steady sales and seasonal patterns; for slow and lumpy items, in our experience, simple rules on recent sales and cover are hard to beat. AI is most useful in the reading and matching: turning supplier confirmations and receipts into clean lead-time records, and flagging items whose cover is rising.

When it doesn't apply

Yards selling entirely on fixed contract price lists have little quote-by-quote pricing to improve. Yards buying from one reliable supplier on short, steady lead times have less to gain from lead-time data. And the pricing study is from an aluminium distributor, not a lumber yard, so it shows the mechanism, not the size of the gain in lumber.

Quarri for wood products is built around how a wood products plant runs, where the order book meets real capacity.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.