Codetrace

𝗛𝗮𝗽𝗽𝘆 𝗠𝗲𝗿𝗱𝗲𝗸𝗮 𝗗𝗮𝘆.

Sixty-nine years on, Malaysia is still being built — by engineers, operators, technicians and the people on every factory floor who keep things running.

That's the work we're part of, and we're glad to do it here.

Selamat Hari Merdeka from all of us at CodeTrace. 🇲🇾

𝗦𝗮𝗹𝗮𝗺 𝗠𝗮𝘂𝗹𝗶𝗱𝘂𝗿 𝗥𝗮𝘀𝘂𝗹 ✨

Today we pause to reflect on the character of Prophet Muhammad: his honesty in dealing with others, his patience, and his compassion.

Warm wishes from all of us at CODETRACE to our Muslim colleagues, customers and partners. 💕

𝗠𝗼𝘀𝘁 𝗳𝗮𝗰𝘁𝗼𝗿𝘆 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝗱𝗶𝗲 𝗮𝘁 𝘁𝗵𝗲 𝘄𝗵𝗶𝘁𝗲𝗯𝗼𝗮𝗿𝗱, 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗯𝘂𝗱𝗴𝗲𝘁 𝗺𝗲𝗲𝘁𝗶𝗻𝗴.

The picture is too big. A full smart-factory transformation, every line, all at once. 𝗢𝗻𝗹𝘆 𝟭𝟴% 𝗼𝗳 𝗠𝗮𝗹𝗮𝘆𝘀𝗶𝗮𝗻 𝗺𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗲𝗿𝘀 𝗵𝗮𝘃𝗲 𝗳𝘂𝗹𝗹𝘆 𝗶𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗲𝗱 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝟰.𝟬. 46% are still in the early stages.

So don't start there. Start with one process.

The one that's hardest to staff. The one that scraps the most parts. 𝗠𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝗵𝗮𝗻𝗱𝗹𝗶𝗻𝗴, 𝗽𝗮𝗹𝗹𝗲𝘁𝗶𝘀𝗶𝗻𝗴, 𝘃𝗶𝘀𝘂𝗮𝗹 𝗶𝗻𝘀𝗽𝗲𝗰𝘁𝗶𝗼𝗻, 𝗿𝗲𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 𝘄𝗲𝗹𝗱𝗶𝗻𝗴. Automate that single cell, then measure four numbers against your old baseline: labour cost per unit, defect rate, throughput, downtime.

If the cell pays, point the saving at your next-worst bottleneck. Repeat. That's how a plant becomes a smart factory, one proven step at a time.

You're not buying a factory of the future. You're buying back one shift you can't staff.

𝗣𝗶𝗰𝗸 𝗼𝗻𝗲 𝗽𝗿𝗼𝗰𝗲𝘀𝘀. 𝗪𝗲'𝗹𝗹 𝘀𝗰𝗼𝗽𝗲 𝘁𝗵𝗲 𝗿𝗲𝘀𝘁.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛 𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑓𝑎𝑐𝑡𝑜𝑟𝑦-𝑎𝑢𝑡𝑜𝑚𝑎𝑡𝑖𝑜𝑛-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗬𝗼𝘂 𝗰𝗮𝗻'𝘁 𝗵𝗶𝗿𝗲 𝘆𝗼𝘂𝗿 𝘄𝗮𝘆 𝗼𝘂𝘁 𝗼𝗳 𝗠𝗮𝗹𝗮𝘆𝘀𝗶𝗮'𝘀 𝗹𝗮𝗯𝗼𝘂𝗿 𝗰𝗿𝘂𝗻𝗰𝗵 𝗮𝗻𝘆𝗺𝗼𝗿𝗲.

Here's the reframe most factory owners miss: automation isn't displacing your people. In a shortage, there's no surplus to displace — only shifts you can't staff and work that isn't getting done.

𝗦𝗼 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗲 𝘁𝗵𝗲 𝘁𝗮𝘀𝗸, 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗽𝗲𝗿𝘀𝗼𝗻:

→ Robotics takes the physical motions — handling, welding, packing — with no fatigue, overtime or turnover

→ Machine vision (AOI) inspects every unit at line speed, catching what a tired eye misses

→ Your people move up into robotics, automation and process roles

And you don't have to boil the ocean.

𝗢𝗻𝗹𝘆 𝟭𝟴% 𝗼𝗳 𝗠𝗮𝗹𝗮𝘆𝘀𝗶𝗮𝗻 𝗺𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗲𝗿𝘀 𝗵𝗮𝘃𝗲 𝗳𝘂𝗹𝗹𝘆 𝗮𝗱𝗼𝗽𝘁𝗲𝗱 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝟰.𝟬 — the national roadmap targets a jump in robot density from 55 to 195 per 10,000 workers by 2030. The field is wide open. Automate one high-labour or high-defect line, measure it against the wage bill and scrap rate, then reinvest into the next bottleneck.

The real risk isn't the price of the first machine. It's waiting, while the plants that started pull ahead.

𝗬𝗼𝘂 𝗰𝗮𝗻'𝘁 𝗵𝗶𝗿𝗲 𝘁𝗵𝗲 𝘀𝗵𝗼𝗿𝘁𝗮𝗴𝗲 𝗮𝘄𝗮𝘆. 𝗬𝗼𝘂 𝗰𝗮𝗻 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗲 𝗶𝘁 𝗮𝘄𝗮𝘆.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛 ℎ𝑡𝑡𝑝𝑠://𝑤𝑤𝑤.𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑓𝑎𝑐𝑡𝑜𝑟𝑦-𝑎𝑢𝑡𝑜𝑚𝑎𝑡𝑖𝑜𝑛-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗠𝗼𝘀𝘁 𝗼𝗳 𝘄𝗵𝗮𝘁 𝘆𝗼𝘂𝗿 𝗔𝗢𝗜 𝗳𝗹𝗮𝗴𝘀 𝗶𝘀 𝗻𝗼𝘁 𝗮 𝗱𝗲𝗳𝗲𝗰𝘁.

On typical SMT lines, 30 to 80 percent of AOI calls are false. Every one still gets checked by a person. That is not just wasted labour. Fatigued review of endless good boards desensitizes the operator, and desensitized operators start passing the real defects. High false call rates manufacture escapes.

𝗧𝗵𝗲 𝘄𝗿𝗼𝗻𝗴 𝗳𝗶𝘅 𝗶𝘀 𝗹𝗼𝗼𝘀𝗲𝗻𝗶𝗻𝗴 𝘁𝗵𝗿𝗲𝘀𝗵𝗼𝗹𝗱𝘀. Every widened tolerance is a door a real defect walks through, and the parts generating your false calls are often your top sources of real ones.

Below ~10 percent, rule tuning hits diminishing returns and ML takes the review, cutting up to 60 percent of manual verification. The catch is training data: real defect images are scarce because defects are rare. That is the gap we attack with 𝗡𝗩𝗜𝗗𝗜𝗔-𝗽𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗱𝗲𝗳𝗲𝗰𝘁 𝗶𝗺𝗮𝗴𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻.

𝗡𝗲𝘃𝗲𝗿 𝘁𝗿𝗮𝗱𝗲 𝗲𝘀𝗰𝗮𝗽𝗲𝘀 𝗳𝗼𝗿 𝗾𝘂𝗶𝗲𝘁. 𝗗𝗿𝗼𝘄𝗻𝗶𝗻𝗴 𝗶𝗻 𝗳𝗹𝗮𝗴𝘀? 𝗦𝗲𝗻𝗱 𝘂𝘀 𝘆𝗼𝘂𝗿 𝗙𝗖𝗥.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛: ℎ𝑡𝑡𝑝𝑠://𝑤𝑤𝑤.𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑚𝑎𝑐ℎ𝑖𝑛𝑒-𝑣𝑖𝑠𝑖𝑜𝑛-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗔 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗰𝗮𝘁𝗰𝗵𝗲𝘀 𝗺𝗼𝗿𝗲 𝗮𝗻𝗱 𝗻𝗲𝘃𝗲𝗿 𝘁𝗶𝗿𝗲𝘀.

Most AOI-vs-manual debates argue the wrong number. 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗿𝗮𝘁𝗲 𝗶𝘀 𝗿𝗲𝗮𝗹 (a human runs 70 to 80% under production pacing, a tuned AOI 95%+), 𝗯𝘂𝘁 𝗶𝘁 𝗶𝘀𝗻'𝘁 𝘄𝗵𝗲𝗿𝗲 𝘁𝗵𝗲 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝘀𝗶𝘁𝘀. 𝗧𝗵𝗲 𝗰𝗼𝘀𝘁 𝗶𝘀.

𝗛𝗲𝗿𝗲'𝘀 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲 𝗥𝗢𝗜 𝗱𝗲𝗰𝗸 𝘂𝘀𝘂𝗮𝗹𝗹𝘆 𝗹𝗲𝗮𝘃𝗲𝘀 𝗼𝘂𝘁:

The human column looks cheap. RM 1.9 to 2.5k a month base, until you multiply by coverage. Three shifts need three times the headcount, plus relief and turnover. Manual cost scales linearly with volume and shift count.

The machine column looks expensive. RM 50 to 200k up front, until you amortize it over five years. But two costs never make the quote: programming every product and revision, and the verification operator re-judging a 30 to 80% false-call ratio. That operator is part of the AOI's cost, not a leftover from the manual era.

The honest answer for most factories is a hybrid line. Machines on the repeatable. People on the judgment.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛: ℎ𝑡𝑡𝑝𝑠://𝑤𝑤𝑤.𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑚𝑎𝑐ℎ𝑖𝑛𝑒-𝑣𝑖𝑠𝑖𝑜𝑛-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗬𝗼𝘂𝗿 𝘄𝗮𝗿𝗲𝗵𝗼𝘂𝘀𝗲 𝗹𝗼𝗼𝗸𝘀 𝗳𝘂𝗹𝗹. Most of what you're paying for is empty air above the racking.

A recent study put average peak occupancy at 85.6%, and 47% of facilities said they needed more space. But occupancy only tells you whether your locations are filled, not how much of the building you're actually using.

𝗜𝗻 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗽𝗮𝗹𝗹𝗲𝘁 𝗿𝗮𝗰𝗸𝗶𝗻𝗴, 𝘀𝘁𝗼𝗿𝗲𝗱 𝗴𝗼𝗼𝗱𝘀 𝘁𝗮𝗸𝗲 𝘂𝗽 𝗷𝘂𝘀𝘁 𝟮𝟮–𝟮𝟳% 𝗼𝗳 𝘁𝗵𝗲 𝗰𝘂𝗯𝗶𝗰 𝘃𝗼𝗹𝘂𝗺𝗲. The rest is aisles, clearances, and the gap between your top pallet and the roof.

So before you sign a new lease:
→ 𝗔𝗱𝗱 𝗯𝗲𝗮𝗺 𝗹𝗲𝘃𝗲𝗹𝘀. Taller racking is the cheapest volume you'll ever buy.
→ 𝗡𝗮𝗿𝗿𝗼𝘄 𝘁𝗵𝗲 𝗮𝗶𝘀𝗹𝗲𝘀. Turn drive lanes into pallet positions.
→ 𝗦𝗹𝗼𝘁 𝗯𝘆 𝘃𝗲𝗹𝗼𝗰𝗶𝘁𝘆, 𝗮𝗻𝗱 𝗰𝗹𝗲𝗮𝗿 𝗼𝘂𝘁 𝗱𝗲𝗮𝗱 𝘀𝘁𝗼𝗰𝗸.
→ 𝗢𝗻𝗹𝘆 𝘄𝗵𝗲𝗻 𝘁𝗵𝗲 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗶𝘀 𝗳𝘂𝗹𝗹 𝗼𝗳 𝘀𝘁𝗼𝗰𝗸, not air, does automated vertical storage earn its capital, reclaiming up to 85% of the floor shelving wastes.

𝗠𝗲𝗮𝘀𝘂𝗿𝗲 𝘆𝗼𝘂𝗿 𝗰𝘂𝗯𝗶𝗰 𝘂𝘀𝗲 𝗯𝗲𝗳𝗼𝗿𝗲 𝘆𝗼𝘂 𝗹𝗲𝗮𝘀𝗲.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛: ℎ𝑡𝑡𝑝𝑠://𝑤𝑤𝑤.𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑠𝑚𝑎𝑟𝑡-𝑤𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗬𝗼𝘂𝗿 𝘀𝘁𝗼𝗰𝗸 𝗰𝗼𝘂𝗻𝘁 𝗶𝘀 𝘄𝗿𝗼𝗻𝗴. Not because your team is careless, but because the average warehouse runs at just 65–75% inventory accuracy, while the benchmark most aim for is 97% or higher.

𝗔 𝗿𝗲𝗰𝗼𝗿𝗱 𝗱𝗼𝗲𝘀𝗻'𝘁 𝗴𝗼 𝘄𝗿𝗼𝗻𝗴 𝗮𝗹𝗹 𝗮𝘁 𝗼𝗻𝗰𝗲. It drifts one handling event at a time: a mislabeled carton at receiving, a misread label at picking, stock moved without a system transaction, shrinkage hiding between annual counts. 𝗕𝗮𝗿𝗰𝗼𝗱𝗲𝘀 𝗼𝗻𝗹𝘆 𝗰𝗮𝗽𝘁𝘂𝗿𝗲 𝘄𝗵𝗮𝘁 𝗮 𝗽𝗲𝗿𝘀𝗼𝗻 𝗿𝗲𝗺𝗲𝗺𝗯𝗲𝗿𝘀 𝘁𝗼 𝘀𝗰𝗮𝗻.

Two fixes actually move the number:
→ 𝗖𝗼𝘂𝗻𝘁 𝗰𝗼𝗻𝘁𝗶𝗻𝘂𝗼𝘂𝘀𝗹𝘆, 𝗻𝗼𝘁 𝗼𝗻𝗰𝗲 𝗮 𝘆𝗲𝗮𝗿. Cycle counting with blind counts and root-cause reviews costs nothing but discipline and surfaces errors within days, while they're still traceable.

→ 𝗧𝗵𝗲𝗻 𝘁𝗮𝗸𝗲 𝗱𝗮𝘁𝗮 𝗲𝗻𝘁𝗿𝘆 𝗼𝘂𝘁 𝗼𝗳 𝗵𝘂𝗺𝗮𝗻 𝗵𝗮𝗻𝗱𝘀. RFID pushes accuracy past 99.9%; AI vision counts cases automatically and inches toward 100%.
Run them in that order — fix the process, then automate the input. Never point a camera at a broken process. One analysis found raising accuracy from 65% to 93% lifted gross sales by ~9%.

𝗧𝗵𝗲 𝘀𝘆𝘀𝘁𝗲𝗺 𝗻𝘂𝗺𝗯𝗲𝗿 𝘀𝗵𝗼𝘂𝗹𝗱 𝗯𝗲 𝘁𝗵𝗲 𝘀𝗵𝗲𝗹𝗳 𝗻𝘂𝗺𝗯𝗲𝗿. 𝗡𝗼𝘁𝗵𝗶𝗻𝗴 𝗹𝗲𝘀𝘀 𝗰𝗼𝘂𝗻𝘁𝘀.

𝑅𝑒𝑎𝑑 𝑚𝑜𝑟𝑒 𝑜𝑛: ℎ𝑡𝑡𝑝𝑠://𝑤𝑤𝑤.𝑐𝑜𝑑𝑒𝑡𝑟𝑎𝑐𝑒.𝑐𝑜𝑚.𝑚𝑦/𝑠𝑚𝑎𝑟𝑡-𝑤𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒-𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠/

𝗚𝗿𝗮𝘁𝗲𝗳𝘂𝗹 𝘁𝗼 𝗵𝗼𝘀𝘁 𝘁𝗵𝗲 𝘁𝗲𝗮𝗺 𝗳𝗿𝗼𝗺 𝗠𝗜𝗧𝗜 𝗮𝘁 𝗼𝘂𝗿 𝗦𝗵𝗮𝗵 𝗔𝗹𝗮𝗺 𝗛𝗤 𝗳𝗮𝗰𝗶𝗹𝗶𝘁𝘆.

Thank you for taking the time to see our work up close. It means a lot to walk visitors through what we build and why it matters for 𝗠𝗮𝗹𝗮𝘆𝘀𝗶𝗮𝗻 𝗺𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴.

Conversations like these keep us focused on solving real problems on the factory floor. We're proud to be doing this from Malaysia, and we're just getting started.

Terima kasih, Ministry of Investment, Trade and Industry, Malaysia 🥰