Are digital twins in sawmills delivering, according to the research?

In the one sawmill where researchers have documented a full deployment, yes. A 2025 case study in the International Journal of Production Research describes "a pioneering sawmill". It calls the mill's twin "Unique for its technical complexity, proven economic impact, and complete life cycle analysis". The authors say the study "provides scarce real-world evidence of a fully automated DT with tangible commercial success". Their own word, "scarce", marks the limit. The published record is one deployed mill, one laboratory prototype and a review that judged sawmill twins "realistic" in 2022. It shows twins can deliver. It cannot yet show how often they do.

What the usual answer says

The usual answer is that sawmill digital twins model production lines with live data to simulate and optimise throughput, recovery and maintenance. That describes the promise. The research is thinner than the confident tone suggests, and more encouraging than a sceptic might expect.

The 2025 case study

Kober, Polzer, Kulasingham, Kulasingham and Xu studied the development of a twin at one sawmill across its whole life cycle. We read the abstract only, so the mill's quantified results are not reported here. The abstract opens by noting that "the maturity of DTs in manufacturing remains limited due to various challenges". It then presents the sawmill as the exception, and draws out success criteria from it, "such as fostering organisational trust to support DT adoption". It describes a successful outcome as one "rarely documented in current DT literature".

The 2022 review

Chabanet, Bril El-Haouzi, Morin, Gaudreault and Thomas wrote the review that tops search results for this question. It defines a twin as "a collection of various models and data updated in real time", with "predictive and analytical capabilities". Its case for sawmills is that twins "would, indeed, allow the testing and optimisation of production plans, including optimising the parameters of embedded real-time optimisers". Without a twin, it says, "the impact of these parameters on the production as a whole is difficult to assess without historical data, which are not always accessible or require costly trials and errors". The twin is offered as the way around that.

The review is optimistic about feasibility. "The state of existing technologies and simulation tools makes the development of DT realistic for the sawmill industry." Equipment makers "supply simulators based on real pieces of software running their machines", which "are therefore able to precisely model the behaviour of the production line, given sufficient data". X-ray optimisation systems are on sale, "but they are too expensive for many sawmills".

The challenges it names are specific. Variation across mills "makes the development of a solution suitable for all cases improbable", so "it is important to propose finished implementations". And "the adaptive capacities of sawmill DT models over time and their response to changes remain challenging and require further investigation".

The prototype and the review with no sawmills

A 2023 chapter from the same group combined a computationally intensive sawing simulator with a machine learning model to predict the lumber sawn from each log. We read its abstract. In numeric experiments it reported "improvements from 11% to 18%" against its baseline. That is a laboratory result, not a mill's.

The figure most often quoted comes from elsewhere. A 2024 review by Tagarakis and colleagues covered 34 digital twin papers, 9 of them in forestry, and found that "only one of the reviewed technologies was classified as deployed". Its forestry papers concern stands, drones and fire, and the one deployed system simulates a rice crop. It says nothing about sawmills, and shouldn't be read as if it did.

The strongest case against "rarely"

Scarce publication is not the same as scarce deployment. Commercial scanner and optimiser simulators, which the 2022 review says can model a line precisely, are twins in most respects. Few mills or vendors write their results up for journals. So the true share of mills getting value from a twin may be higher than the literature shows. It may also be lower, since failures are published even less. The research can't settle it either way.

Reading it through a data lens

The open challenge in the 2022 review is keeping a twin accurate as logs, equipment and markets change. In our view that is mainly a records question. A twin's prediction about a new setting can only be trusted if earlier predictions were checked against what the mill actually produced from the same logs. That needs the log scans, the settings in force and the resulting tally tied to each other, and kept.

The 2025 study's point about trust follows from the same logic. People act on a twin's recommendation when they have seen its past recommendations borne out. A mill that can show that record has a better chance of the twin being used, and so of it paying.

Where it touches an operation

For a mill weighing a twin, these papers point to three questions. Which decision will it improve? How will its predictions be compared with actual output, and how often? Who will act on what it recommends, and on what evidence? A vendor that can answer the second question with the mill's own data is offering something the literature says is rare.

What we cannot take from it

We read the 2025 study and the 2023 chapter as abstracts only. One documented deployment says nothing about the average mill. The 2022 review predates newer simulation tools.

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.