What 68 sawmills told us about their data

This article is based on a public document: Fernández, M.; Hernández, G.; Troncoso, L.; Elgueta, P. (2019). Brechas tecnológicas-productivas en la Pyme de aserrío de las regiones de Maule, Ñuble, Bio Bio, La Araucanía, Los Ríos y Los Lagos (Technology and productivity gaps in SME sawmills of six Chilean regions). Instituto Forestal, Technical Report No. 223, ISBN 978-956-318-160-9, DOI 10.52904/20.500.12220/29194. The report is in Spanish; what follows is my reading of it, seven years later and from the software side.

A large industry made of small companies

In 2018 Chile had 984 sawmills in operation. Together they produced 8.3 million cubic metres of sawn lumber and employed 16,373 people directly. Almost nine out of ten were in the macro-region stretching from Maule to Los Lagos. This is the industry that feeds timber construction, and in 2019 two public policies were betting on it: CORFO's Madera de Alto Valor programme aimed for 30 % more timber-frame houses by 2025, and the national Forest Policy proposed doubling them by 2035.

For that to happen, lumber has to leave the sawmill with dimensions and moisture content within tolerance. And to know whether it does, someone has to measure it and write the number down. That is where the problem we went to measure began.

What we asked, and whom

Between 11 March and 12 July 2019 we surveyed 68 sawmills producing between 10,000 and 120,000 m³/year across six regions, out of a population of 95 plants in that range — a 6.2 % sampling error at 95 % confidence. We talked to owners, managers and production chiefs, on site. I coordinated the study — from questionnaire to analysis — and wrote the report, which I sign together with three colleagues from INFOR's Wood Technology and Products unit.

The survey had two halves. The first was about machinery: debarking, log sorting, headrig, edging, packaging. The second, the one I cared about most, was about management: which indicators each plant used in each area — raw materials, sawmill, drying, planing, maintenance, quality control, planning — and, for each one, how the data was captured: by hand, semi-automatically or automatically, and with what software it was processed.

What we found

The machinery half gave the expected portrait of a mature industry: 83 % of output came from lines with a band saw and carriage or twin band saws as the headrig; stacking and stickering were mostly manual; maintenance was, to a large extent, corrective.

The data half was what justified the study:

• Only 38 % of the sawmills had a quality-control function, and 41 % performed dimensional control in their process.
• Of those that did, 79 % measured with a digital caliper, wrote the reading on a paper form, and later someone typed the numbers into a computer. 21 % had adopted Bluetooth calipers that sent the reading to an application.
• 36 % kept documented moisture-content control at dry grading; again, recorded by hand and digitised afterwards.
• 66 % had some production-control system, but its functionality and reliability fell short: poor inventory management, hard traceability, information that was not real-time, data-entry errors. Planning was done, above all, in spreadsheets.
• Most had no information on line stoppages, or had it in a format that could not be analysed.

And the figure I found most important: 43 % said they were interested in adopting technology for more efficient quality control; 28 % wanted to automate data acquisition and move to web or desktop software; 25 % explicitly asked for tools to monitor operational losses. The demand existed. What did not exist was an offer within reach of a small company.

The reading that gets lost between the caliper and the computer

If you look at the 79 % workflow, the problem is not measurement: a digital caliper measures well. The problem is the next two steps. Writing by hand introduces transcription errors and slows measurement down, so less gets measured. Typing it in later adds a second layer of errors and a delay: by the time the reading reaches the spreadsheet, the batch it came from is already packaged or in the kiln. The control exists on paper, but it controls nothing, because it arrives late.

Seen from the software side, this is a data-acquisition problem, not an analytics one. Before thinking about dashboards or KPIs, the gap between the instrument and the record had to be closed: the reading should reach the record without passing through paper, with its context — lot, section, operator, time — and the analysis should work on data that was clean from the source.

From diagnosis to specification

That reasoning predated the report and became a specification. First, in 2017 and within a joint INFOR–CETMA project, as a spreadsheet with macros used with the plants to test the calculations — dimensional variation, tolerances per section, per-shift summaries — and the logic of what is recorded versus what is derived. Then, in 2018, as the design of a mobile application that receives the caliper reading over Bluetooth, attaches the context and sends it to a web platform where the reports are generated. That specification went into the formulation of a CORFO Technology Diffusion Project (PDT) that Universidad de Concepción was awarded in 2019 and ran from 2020 with CORMA; by then I had returned to INFOR and took no part in its execution. At INFOR, the same specification was the basis of the module implemented with an external developer: I coordinated the hiring and the initial stage, and the workbook's calculation logic was ported into the system. I tell that story in detail in The spreadsheet that became a system.

One engineering decision is worth telling, because it is the one that decides whether such a tool ever reaches a small company. The market offered integrated solutions with proprietary hardware, technically sound and priced well beyond the 20,000 m³/year sawmill we had surveyed. We chose standard, open-connection, low-cost Bluetooth calipers, and put the intelligence in the software rather than in the instrument. It is the same choice I would make today for any shop-floor capture system: the hardware can be bought anywhere; what gets designed is the flow of the data.

INFOR owns that family of applications for SME sawmills — dimensional control, moisture, downtime, production — and maintains it today at controlmaderainfor.cl. I do not speak for its later evolution or its code; I can say the starting point was the diagnosis in this report.

What I took with me

Three things that still hold in what I do now, which is a different industry and a different stack:

• Measure the gap before proposing the solution. Without the survey, the obvious answer would have been "a management system". With it, the answer was much smaller and much more useful: fix the capture.
• Total cost decides adoption. A technically better solution that nobody buys improves no indicator. Open, cheap hardware; intelligence in the software.
• Data that arrives late is not control. It is the same principle I apply today to an ETL ingestion or an inference queue: the latency between the event and the record is part of the design, not a detail.

A note on the report's scope. In 2024, Gysling, Mejías and Herrera used it in Forest Products Journal as a source on the technological deficiencies of SME sawmills. That is a fair use, with one nuance the report itself makes clear: the machinery base we found was mature — ring debarkers, band saws with carriage and twins behind 83 % of output; the gap was not in the iron, but in how the data was captured and used.

Suggested citation: Fernández, M., Hernández, G., Troncoso, L. & Elgueta, P. (2019). Brechas tecnológicas-productivas en la Pyme de aserrío de las regiones de Maule, Ñuble, Bio Bio, La Araucanía, Los Ríos y Los Lagos. Technical Report No. 223. Instituto Forestal, Chile. 52 pp.