How do hardwood and softwood mills use data differently?

They chase different money, so different data pays. A softwood dimension mill earns on volume: logs through the headrig per shift, tons of logs per thousand board feet, uptime, and a product mix matched to price. Small changes in conversion, across large volumes, are worth a lot. Its data work is about conversion by log source, stops, and whether the optimiser's settings still match the market. A hardwood mill earns on grade: each board's value depends on its clear cuttings, and species and log quality vary widely. Its most valuable data is grade outturn by log source and species, how accurately it saws to thickness, and which grade each board is worth most in. Hardwood mills are usually less automated, and they often have more to gain from each piece of data.

What the usual answer says

Descriptions of the two sectors say softwood mills are highly automated, with scanning and optimisation driving volume and speed. Hardwood mills, they say, focus on grading, species sorting and quality, and have adopted technology more slowly. That describes the equipment. It stops short of saying what each kind of mill should do with the data it has.

What sawing accuracy is worth in hardwood

Edward Thomas and Urs Buehlmann, in two papers published in 2023, simulated hardwood logs sawn with different amounts of sawing variation, the deviation from target thickness. Reducing it in the simulation gave "a minimal average recovery improvement of 3 percent". They value that at $336,000 a year for an 8 million board feet mill, assuming lumber at $1,400 per thousand board feet.

In the second paper they point to a data gap: "sawmill personnel generally do not know at which point efforts to reduce (SV) become more costly than oversizing the boards". Answering that needs thickness measurements by machine and by shift. A mill that doesn't keep them can't find its own break-even point.

What grade decisions are worth

Sailesh Adhikari, Brian Bond and Henry Quesada, in a study published in November 2023, sawed 126 yellow poplar logs. They graded the lumber three ways: standard hardwood appearance grades, a structural grade for cross-laminated timber, and a mix. "Producing mix-grade lumber added approximately 27% more revenue than producing NHLA-grade lumber on average" with a cant sawing method. Structural hardwood lumber has no established market price, so the authors priced it at production cost plus a 15% margin. The 27% shows what a grade choice could be worth if that market develops, not revenue a mill can bank today. The yields, by contrast, were measured.

For data, the point is grade outturn recorded by log, species and sawing pattern. Set against prices for each grade, it shows a hardwood mill which logs and boards are worth more under another rule.

Checking the log buy

A hardwood mill that records grade outturn by log source can check its log buying. Do logs bought as a given grade from a given supplier yield the lumber grades that price assumed? The tally answers that. Over a season, that record can inform what to pay each supplier for each log class. It needs grade tallies joined to log tickets, which is often the missing step.

What softwood mills get from their data

On the softwood side, Amanda Lang of Forisk reported a November 2024 survey of 38 pine sawmills. Weighted average conversion ran at 4.37, 4.27 and 4.07 tons of logs per thousand board feet for mills of increasing size. Conversion had improved since 2019 at every size, by 13% for the smaller mills. A larger mill, the survey found, makes a thousand board feet with 7% less raw material. And "Southern yellow pine lumber recovery rates are strongly correlated to diameter, taper and quality".

That points softwood data at conversion by log source and diameter class, and at stops and the constraint station. It also points at whether the optimiser's price file and grade settings still match what sells. At softwood volumes, a small gain in conversion is worth more in total than most per-board gains in hardwood. Our reading is that the softwood mill's data problem is keeping a well-instrumented system honest: scanner calibration, stale price files, and supplier recovery that drifts.

Where both meet

Both need logs tied to their source, boards tied to their logs or at least their shifts, and production tied to what it sold for. Where grade tallies are on paper, AI can read them into board-level records by log source and shift. Outturn can then be joined to log tickets without keying.

When it doesn't apply

Mills cutting both hardwood and softwood need both agendas. Hardwood mills producing mainly industrial products, such as pallet cants and ties, sit closer to the softwood pattern, because volume matters more than grade. And the hardwood studies are a simulation and a one-species trial. They show what the levers are worth in principle, not at a given mill.

Quarri for sawmills is built around how a sawmill runs, from log intake to shipped order.

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.