// Insights
Seeing the wood for the trees.
What we learn building data platforms for forestry and lumber operators, reconciliation, margin, mill economics and the messy reality of industrial data.
Trade · 14 Sep 2026
Canada put 50% on dissolving pulp in September. RYAM stopped making it in the first quarter.
Canada's 50% counter-tariff on dissolving pulp came into force on 8 September. RYAM's own filing shows it permanently stopped making dissolving pulp at Témiscaming two quarters earlier, and the rest of that site stops on 15 September.
Read the article →
Research · 2 Sep 2026
The barrier to AI in the forest supply chain isn't the models. A 69-study review says it's the data.
Sixty-nine studies, seven barriers to adoption, and only one of them is about the models. The remedies the authors propose are both data-integration institutions.
Read more →
Trade · 7 Aug 2026
Demand for Canadian lumber is recovering. The tariff effect isn't in the numbers yet.
Export value has climbed 13.7% from its March low, on price and volume together. The 21.4% half-year fall still being quoted describes the quarter the market has spent since climbing out of.
Read more →
Trade · 31 Jul 2026
The tariff everyone is quoting is 50%. The one that moves the market is 45%.
The new 50% tariff exempts softwood lumber. The duty softwood already carries is 45.16% all-in, and one producer's filing shows $202m of it in six months.
Read more →
Analysis · 16 Jul 2026
Weyerhaeuser's $1bn isn't an AI number. It might also be too small.
The number that travelled around the industry this year is that Weyerhaeuser will add $1bn of annual profit by 2030 using AI. That isn't what the company said.
Read more →
// Articles by section
The case for a timber data platform
How is a timber data platform different from an ERP?An ERP records transactions as they happen. A data platform keeps them, joins them with other systems and answers questions across time.Does a forest products business need a data platform, a data warehouse or a new ERP?It depends where your wrong numbers come from: bad entry points to the ERP, short history to a warehouse, disagreeing systems to a platform.What is a semantic layer, and why does AI need one?A written model of what your data means. It turns AI's plausible wrong answers into refusals, at the cost of questions not yet modelled.Is a digital twin worth building for a forest or a mill?For a mill, rarely as a whole: model the decision that costs most when wrong. For a forest, the shared inventory a twin needs may come first.What should a forestry database hold, and where?Dated measurements and events with their source documents, keyed to the same stand and contract IDs as the map, plus a record of how those IDs change.What does it cost to leave forestry data on paper?Three costs: keying and handling, which benchmarks measure; exceptions found late; and questions never asked of the records, which nothing measures.Should you replace your ERP or put a data layer over it?Replace it when it can't be supported or can't record the business. If replacement is coming, a data layer built first can carry the history across.Why do reports from different systems never agree?Different definitions, different moments, or an error. A re-run test separates timing from the rest, including figures for closed months that keep moving.
See all 8 →AI in forestry and lumber
How are forestry and lumber companies using AI in 2026?On machines and imagery, and in decisions such as matching production to demand. What 2026's public record lacks is a measured result.What is the role of AI in forestry?An assistant that reads more data than people can, in monitoring and operations. Its output is only as good as what it is checked against.What works in AI for forest management, and what is hype?AI works where it measures what the sensor sees, or is calibrated and checked against your own records. Untested claims of volume or value are the hype.Where is AI working in the lumber industry, and where isn't it?It works on the cut: grading, defects, sawing decisions. It lags on setup-aware scheduling, deep in research but rare in mills.What can AI do in a sawmill beyond the scanner?Machine-level work like maintenance and drying, more vision, and work on the mill's records. The records get least attention and can drift furthest.How is AI being used in the wood products industry?Mostly at the scanner and the saw. Its gains are quoted as less waste, more yield or more value, and the baseline decides how large they look.How does AI lumber grading work?Sensors image each board, a model finds defects and the grade rules are applied. Ask whether accuracy is measured on grade or on value.What does AI do in a forest products company's back office?The generic uses are payables and knowledge search. The timber-specific one is matching loads, rates and settlements to what was booked.Can AI be trusted with timber numbers?Not as a system, only answer by answer. Trust a figure when it comes with the records behind it and a check that every record was counted.What data do you need before AI is useful?It depends on the question: the records the answer is built from, enough history to cover its cycle, and keys that link them. Not all your data.What is an AI agent, and what could one do in a lumber business?A model that takes steps with tools towards a goal. In lumber, the first good jobs read many records and draft actions for a person to approve.Can you ask your ERP questions in plain English?Yes. AI writes the query much better than it did. Whether the answer is right depends on written definitions of what your ERP's fields mean, and a test.Can AI estimate lumber quantities from building plans?Partly. On a vendor benchmark, AI beat estimators on framing and found more items, but a quantity counted as right within 25%.Why hasn't AI reached day-to-day operations in forestry?Two reasons fit better than reluctance: records that arrive after the decision, and records split across parties. Measure both for your own operation.What is the difference between AI on the saw and AI in the office?Less the technology than the test. Machine grading is validated against a known answer; office AI rarely is, and its errors are harder to see.Does AI make mistakes with numbers, and how do you catch them?Yes, and some are confident: asking again won't catch them. Control totals, recalculation outside the model and record counts will.What does "deterministic" mean, and why does it matter for AI on financial data?Same inputs, same answer, every time. Language models aren't, even at temperature zero, so reportable figures should come from defined calculations.
See all 17 →Maps and GIS
What is agentic GIS?A GIS run by an AI agent that plans spatial analysis from a plain request. In 2026 tests the best agent fully solved a third of real tasks.Can you ask a map questions in plain English?Yes, if the layers are clean and the words are defined. Near and adjacent have no fixed meaning, and an AI picks one silently. Write them down.Why do forest operations keep their maps and their numbers in different systems?A stand has many loads, contracts and payments. Putting them on a map needs a shared key and a summary rule, or figures drop or double.What can AI add to GIS in forest management?New layers the GIS never had, such as canopy height from imagery, each with a known error that decides what it is fit for. Plus help running analyses.What happens when spatial data and financial data share one screen?Unplaced transactions and disagreeing totals show up first, and a table would show them too. The map's own payoff is questions like haul cost by block.What does forestry mapping software do, and what does it leave out?It records where stands, roads and boundaries are. How accurate each line is, and when it was drawn, is stored only if set up, and rarely reaches reports.What can LiDAR data tell a forest manager, and what can't it?LiDAR measures height and canopy well, models diameter and volume less well, and can miss trees. Check it against the scale record, not the map.How do you connect harvest plans, contracts and remaining volume on a map?Key contracts to sale areas and tickets to contracts. Remaining volume is an estimate less removals, so show it as an estimate.
See all 8 →Software and systems
What should you look for in lumber ERP software in 2026?Beyond features: how it records your awkward transactions, and whether your data comes out cleanly. Test both in the demo with your own examples.What kinds of sawmill software are there, and which do you need?Four kinds, each holding a different record: optimisers, planning, ERP and maintenance. What you need depends on which record your questions need.What does the forestry software market look like in 2026?Published size estimates differ by a factor of two and don't always add up. What matters is whether the tools' links reach tickets, contracts and accounts.What is the difference between forestry management software and forestry inventory software?Inventory software estimates what is on the ground at a date, with error. Management software records what is done. Check the handoff keeps date and error.What can lumber inventory management software tell you, and what can't it?It tells you what the records say, by item and location. It can't tell you what is in the yard, or which way the records are wrong. A count shows both.Why does sawmill accounting software rarely match the mill?Logs enter by scale and lumber leaves by tally, and the ratio moves with the logs. A standard factor, or one average, hides where margin comes from.What can sawmill optimisation software do, and what can't it?It finds the highest-value way to cut each log against a price file. It can't tell whether the prices are current or whether anyone ordered the product.What is timber inventory software?Software that turns plot and tree measurements into estimates of standing volume, products and value. The good kind reports how uncertain each estimate is.What does lumber yard ERP software need to handle?Units of measure, stored once on each item so every module agrees. Every lumber ERP claims it; the test is where conversions live, and whether they drift.Can you get data out of an old lumber ERP with no API?Usually, yes. The reliable routes are the database and the reports the system already prints. Driving its screens with bots is the least reliable route.Why do software vendors charge for data access?Partly real costs, partly because access is where switching costs live. The 2024 cloud exit-fee waivers show how such fees behave when rules change.Should a forest products company build its own data platform or buy one?Buy the plumbing, as almost everyone does. The decision is who builds, owns and maintains the timber meaning: units, tallies, keys and rules.What is the difference between an ERP, a WMS and a data platform?They record the same stock at different grains. The WMS knows where it is, the ERP what it is worth, and a data platform keeps both over time.How do scale house systems and ERPs share data?Tickets pass from scale to ERP as weight, and the ERP turns weight into volume and money with a factor. The keys and the factor are where it breaks.What doesn't a converter's ERP tell you?For a packaging converter costing at standard: the grade really run, trim and set-up waste by order, and the cost of serving each customer.How do you evaluate a data platform for a timber business?On your own data, with questions whose answers you already know. Score it exactly, include questions it should refuse, and test how it handles change.What should an AI vendor tell you about your data before you sign?What it reads and with whose permissions, where it is kept, what it retains and for how long, who else touches it, and how you get it back.Lumber ERP alternatives for mid-size distributorsBefore choosing a new ERP, check the problem is the ERP. If it is reporting and scattered data, a data layer over the current one avoids a migration.
See all 18 →Documents and paperwork
Can AI read handwritten trip tickets and scale slips?Mostly, yes. The risk is a misread that looks like a valid number, so check each ticket against its own arithmetic and a second record.What is the difference between OCR and AI document extraction?OCR turns an image into characters; AI extraction turns it into fields. Quality depends on what the model sees and whether its confidence is calibrated.Why isn't a scanned document the same as having the data?A scan is a picture of the data. You can look at it, but you can't sum it, join it to other records or check it against them until it becomes fields.How accurate is AI at reading invoices and settlement statements?Per field, often very accurate. Per document, less so, because errors multiply across fields, and settlements have many. Measure it on your own documents.How do you digitise forestry paperwork?Capture digitally where you control the form, read what partners keep sending on paper, and start with documents that trigger payment.What does forest inventory data analysis involve?Turning tree measurements into estimates through a chain of equations. Change the equations and the answer changes, with nothing changed on the ground.How do you check that AI read a document correctly?With checks that don't depend on a person spotting the error: the document's own arithmetic, a match to other records, and review against the source image.What should happen when AI can't read a document?It should say so, give a reason and send the document to a queue, never guess. Then the reasons become data that shows which senders and forms to fix.
See all 8 →AI and data by vertical and segment
What can AI do for forest managers and TIMOs?Two jobs: better inventory for the appraisal that drives most of the return, and property-level variance against plan, explained from the documents.Can AI forecast timber harvest volumes from weather data?Yes, but weather matters through ground access far more than through machine output. Forecast the days a site can be worked, from your own history.How does AI reconcile scale tickets, stumpage and invoices?Load by load: it matches the load record, scale ticket and settlement, applies the contract rate by product, and explains what's left.Can you find a mill's bottleneck from the data alone?Usually, if machines log their states with times: running, down, blocked, starved. The machine with most downtime is not always the one limiting output.Can AI schedule a sawmill against open orders?Yes, but the simple due-date rule many mills use already does well. A 2024 study shows why; AI earns its place when logs, machines and orders change.Which questions can AI answer for a millwork manufacturer?Two kinds: questions about drawings and documents, and questions about your own sales, purchasing and stock. The second kind needs joined records.Can AI flag dead stock before it ages?Mostly, yes, and the best place is the purchase order. The signs are in data a yard holds: cover rising, one customer, special-order origin.Why does averaged cost data mislead buyers?An average moves with the mix as well as with prices, so blended cost per MBF can fall while the same basket gets dearer. Split price from mix.Can AI optimise purchase orders across suppliers?Yes, where there is a choice and the data compares. At one lumber manufacturer about 7 in 10 stocked items had one supplier; start with the data.What can a paper or packaging converter ask of its own data?Which work is long-run and stable, which is a long tail of short runs, and what each really earns after changeovers, substitutions and service.How do hardwood and softwood mills use data differently?Softwood mills chase volume and uptime, hardwood mills grade. So different data pays: stops and mix for one, grade outturn by log for the other.What does AI change for a logging contractor's paperwork?Two things: less time on forms, and a cheap way to check settlements against your own loads. A one-month test shows which one pays for you.Where does AI fit in a box plant's order book?At intake, combining and date-promising. Trim optimisers are old, proven software; the newer gain is reading orders in cleanly and promising dates.Can AI plan remanufacturing runs to cut changeovers?Yes, for lumber and millwork reman. The planning methods date from the 1960s; the gap is a record of setups by product pair. Measure, then group.What could AI tell a timberland investor that a quarterly report doesn't?How the manager is executing between appraisals: harvest against plan, stumpage against the price deck, cruise against scale, and costs, month by month.How are engineered wood producers using production data?Inside the plant, the most useful job for process data is predicting panel properties before the lab test, with a range. The evidence is on particleboard.What can AI do for a lumber yard's pricing and stock?Suggest quote prices from each rep's own history and today's cost, leave unusual quotes to people, and set reorder points from real lead times.
See all 17 →Data practice
How do unit-of-measure errors corrupt spend data?In two ways: some change the spend total, most only the unit cost. Errors large on one line look small in the total. The ratios they leave can find them.How do you prove where a number came from?For a business figure: keep the rule and its version, the records it used, the documents behind them, and when it ran. Judgement calls need an approver.Is it safe to connect AI to your ERP?It can be. Beyond least privilege and logging: keep document reading apart from anything that acts, and treat a confident wrong answer as a safety risk.Does AI train on your company's data?It depends on the plan and the account. Business terms we checked forbid it; consumer plans may allow it. Staff bringing their own AI are the gap.Where should a timber business host its data, US or EU?Where its personal data and contracts point; most timber data isn't personal. Decide by data type, know the transfer mechanism, keep a way out.What is data lineage?The record of where data came from and what was done to it, kept for tables, fields or single records. For timber data, the record level matters.How clean does data need to be before AI is useful?Clean enough for the question. For totals, systematic errors across many rows matter more than typos, and automated cleaning handles them worst.How do you keep reports trustworthy when the underlying data changes?State what each report knew and when, keep history so any past report can be reproduced, and show what changed since the last issue and why.
See all 8 →Public timber and lumber data
Where can you find public timber and lumber price data?At four levels: stumpage, delivered logs, wholesale lumber and trade. Each is a different price on a different schedule. Match level to decision.What does the US Forest Service cut and sold report show?Quarterly volumes and values of timber sold and cut from national forests, by region, species and sale value. A supply signal, not a price series.How do you read the producer price index for wood products?As an index of domestic producers' net revenue, not a price. Tariffs paid are excluded, recent months are provisional, and series differ in base and type.What do SEC filings reveal about forest products companies' costs?It depends on the company. Since 2024 segment notes show the costs management reviews: fiber and energy for one pulp producer, one cost line for others.What public data exists on lumber and log trade?Customs data by product code, trade price indices, and FAO's annual series. Exporter and importer figures rarely agree, and some gaps are filled.What do European timber price indices track?Different things under one name: standing or roadside sales, public or private sellers, real or nominal, under or over bark. Check before comparing.How reliable is public stumpage price data?It depends on who reported and how many. Some published averages rest on three sales; weighting and unit conversions move them too. Check first.What do rail carloadings say about lumber demand?Less than the headline suggests. Lumber hides in a forest products group, carloads track shipments and stocks, and figures are revised for a year.What does Eurostat production data show for wood products?Annual sawnwood, panel, pulp, paper and roundwood volumes by country, released 10 to 12 months after year end, with many recent cells blank.Can public data benchmark a private timber business?On prices, yes, with care. On returns and operations, only partly: public benchmarks cover other owners, or publish averages that hide the best.Why is public forestry data so hard to use?Inventories are built to estimate populations, not your land. Plots are moved and swapped for privacy, and county removals estimates carry wide errors.How do you combine public market data with your own numbers?Join each record to the public figure as it stood on that date, on the same basis. Store vintages where revisions are large, and test AI after its cutoff.
See all 12 →Regulation and reporting as data problems
What data does EUDR actually require from wood exporters?Plot geolocation for every wood component, full species names, harvest dates and legality evidence, kept segregated to shipment, from 30 December 2026.How do you prepare Lacey Act declarations from your own records?Hold genus and species, harvest country, unit and any special use designation on each item record, then generate each electronic declaration from it.What records does chain-of-custody certification need?Purchases, sales, material accounting, conversion factors and controlled wood checks, kept five years. An optional 2026 module adds more, by role.How do you track tariff and duty costs by product?Record each duty layer per entry line by code, origin and date, accrue it at receipt and true it up. Refunds and rate changes need the detail.What data does carbon reporting need from a forest products business?Fuel, power and haul data by site, biogenic CO2 kept apart, and co-products allocated. The 2026 land sector standard leaves forest product chains out.How are softwood lumber duties calculated, and what data do you need?A deposit at entry, trued up years later when a review sets final rates per producer, with interest. Keep each entry's value, date, producer and deposit.What records do auditors ask a sawmill for?Tax examiners want contracts, scale tickets and reconciliations; financial auditors want records plus proof each system report is complete and accurate.How do you keep compliance data current without a spreadsheet?Hold each certificate, permit and declaration as a record tied to the registry that owns it and to the purchases it covers, and keep every version.
See all 8 →Listed companies and AI
How is West Fraser using AI and data?Its filings name no AI project, but outside them it is piloting autonomous log hauling and hiring AI developers, while unifying legacy IT platforms.How is UFP Industries using AI?It says it is investing in AI and machine learning but names no project or figure. The capital it describes and sizes goes into automation.How is Louisiana-Pacific using technology?Its 10-K shows product technology, a published OEE metric by business, and AI in standard risk language as "in the early phases". Siding OEE rose in 2026.How is Stora Enso using AI?In its forests, for planning and monitoring: a digital forest twin in its 2024 and 2025 reports, with biodiversity the newest aim. Partners show more.How is UPM using AI?A company-wide AI programme, pilots in orders and invoices, a field assistant used by hundreds of foresters, and machine vision in pulp mills.How is Holmen using data?Above all to value its forest: a regression on property sales, three methods each within 1.2% of the weighted value, and a 2019 inventory rolled forward.How is SCA using technology in forestry?Laser scanning and harvester data combined with AI into a digital twin for harvest planning, an EU single-tree project, and AI log scanning at a sawmill.How is Mondi using AI?In its reports, mainly as a board topic focused on production. Outside them, a university AI tool it uses to choose paper grades, and AI in HR since 2019.How is International Paper using AI?Its 10-K treats AI only as a risk; a vendor story says it runs ML on mill sensor data. The data work the 10-K describes is 80/20 segmentation.How is Södra using data?Through members' forest management plans, now made with remote sensing and AI models, and automated sustainability data collection.How is Domtar using AI in its sawmills?To grade boards at planer mills against each customer specification, at seven named mills. It gives no accuracy figures; we add the records to keep.
See all 11 →Research through a data lens
How well does AI grade logs, according to the research?Off-the-shelf image models in a 2026 study: 92.6% on species with look-alike oaks merged, 40 to 55% on value, where guessing the middle band scores 50%.Can harvester data predict productivity? What the studies foundYes, to about R² 0.7 in two studies. One built its model on 1.12% of 144,973 records without saying why. Tree size drives most of it.Does machine learning forecast lumber prices better than the market?The four recent studies we read don't test it in their abstracts. They compare models with models, not with a no-change forecast or futures.What has computer vision achieved in log yards?About 1% diameter and volume error where capture is controlled, on log ends or a sorting line. From ordinary photos, one model missed four logs in ten.Are digital twins in sawmills delivering, according to the research?In one documented case, yes: a 2025 study reports a sawmill twin with commercial success. The published evidence is still one mill deep.How accurate is LiDAR forest inventory in published trials?At stand level, about as accurate as a field cruise when both are checked against mill scale. At tree and diameter-class level, much less so.What does the research say about optimising log truck routes?One large 2026 preprint reports a 35% distance cut against executed plans. It points to pooling contractors' trucks, but does not test that.Why don't forest supply chains share data? What the studies showOne study of harvester data found no clear owner and no accepted way to price it. A consumer-goods study found sharing pays in plan alignment.How do small manufacturers adopt AI, according to the research?Slowly, and mostly at the edges. Where manufacturers take AI into production, a 2025 study finds a short-run dip, deepest in older firms.How accurately does document AI read handwritten forms?Close to a careful person on free text, weaker on handwritten tables, and wrong in plausible ways. What 2025 and 2026 benchmarks found.What works in demand forecasting for wood products?Time series for the short run, structural models and scenarios for the long run, and drivers only if they are easier to project than demand.What does the research say about decision support in forest planning?Strong on calculation, thin on reported proof. A 2025 review of 23 harvest planning tools found none reporting independent validation or open code.What is data integration worth to a manufacturer? What studies measuredNo independent study has measured integration alone. The closest, of 30,000+ plants, found data use paid 1 to 3% only with the right complements.Does AI improve yield in secondary wood processing?Against fixed-length cutting, yes: a 2024 trial lifted yield from 69.2% to 81.5%. Against a good optimiser, untested, and false detections cost wood.
See all 14 →