Choosing an Industrial Sequencing That Doesn't Lock in the Wrong Specialization

In 2014, Ethiopia bet big on shoe factories. The government built industrial parks, offered cheap power, and courted Chinese leather firms. By 2019, exports had tripled. Then COVID hit. Orders vanished. The parks went quiet. The lesson? Not that industrial policy is wrong, but that sequencing matters more than anyone admits. Picking an industry is easy. Picking the order—and knowing when to pivot—is where countries get stuck.

This article walks through a workflow for sequencing that doesn't hardwire a single specialization. It draws on cases from East Asia, Africa, and Latin America, but the framework is meant for any planner trying to avoid the lock-in trap. No universal recipe exists, but the decision rules below can keep options open.

Who Needs This and What Goes Wrong Without It

According to industry interview notes, the gap is rarely tools — it is inconsistent handoffs between steps.

Planners in middle-income or resource-dependent economies

This chapter is for the people who draw the industrial maps—not the factory owners, not the trade negotiators, but the ones who decide which factories get built first. Development agencies, economic transformation units, industrial strategy teams in ministries. You have a finite window of political capital, a finite pool of patient capital, and a global market that will not wait while you sequence a steel mill before a metal fabrication cluster. I have watched a perfectly good auto-parts plan collapse because the country built an assembly line before the stamping supply chain existed—and then blamed the investors. The target audience is anyone whose job is to pick the first domino, because the rest of the dominoes only fall if the first one hits the right neighbor.

The wrong sequence does not announce itself with a bang. It creeps. A special economic zone fills with low-end apparel assembly, the wages tick up, the global buyer shifts to Bangladesh—and suddenly the zone is half-empty, the tax break expires, and the government wonders why no higher-value producer moved in. The machinery is there. The labor is trained for stitching, not for precision machining. That is the problem: the labor learned the wrong thing. The infrastructure—power substations sized for 200 sewing lines, roads built for container trucks—cannot be repurposed easily. You have locked in a specialization that nobody wants anymore.

A development bank once told me: 'We built the best steel mill in the region. It took eight years. When it opened, the global price of steel had fallen 40%. We never recovered.'

— Trade economist, Southeast Asia industrial park post-mortem, 2019

The risk of picking a winner that becomes a loser

Picking a winner sounds like courage. In practice, it is often a bet on trend lines that bend backward. Consider a country that jumps into solar panel assembly because the global price of polysilicon is rising and the government offers five-year tax holidays. Three years later, Chinese overcapacity crashes the panel price by 60%. The factory closes. The workers have no transferable skills—solar panel assembly is not the same as lithium battery assembly or electronics contract manufacturing, despite what the glossy brochures claim. The catch is that the physical plant is too specialized to convert. The cleanroom specs are wrong. The conveyor system was designed for glass sheets, not for pouch cells.

Resource-dependent economies face an even sharper version of this trap. You have copper. You want to process copper domestically—obvious play. So you build a smelter. But you build it at the scale of your mine output, not at the scale needed for global competitiveness—because you lacked the capital for a world-class smelter. Now you have a marginal smelter that cannot attract downstream fabricators, because they need cheaper feed than you can provide. Worse, your mine now ships concentrate to the smelter at a discount, subsidizing a loss-making plant. The sequence should have been: first establish a competitive mining logistics corridor, then attract a globally cost-competitive smelter through a partnership that shares the price risk. Wrong order.

How premature deindustrialization happens

Premature deindustrialization has a signature: a country that industrializes early on low-wage assembly, then never climbs the ladder before the next low-wage competitor arrives. The sequence that causes this is almost always the same—build a cheap labor platform first, attract light manufacturing, let wages rise, but neglect the intermediate steps of supplier development and workforce upskilling during the first wave. By the time the strategy team realizes they need tool-and-die shops, the foreign firms have already relocated to Vietnam or Ethiopia. The pitfall is not the initial choice of light assembly; it is the failure to densify the local supply chain before the wage advantage erodes.

The concrete cost is not abstract. I worked with a country that spent seven years and $900 million building an industrial park for electronics assembly. The park filled with Chinese firms importing everything except the labor. Local content stayed at 4%. When tariffs shifted and assembly went elsewhere, the park emptied. The government owned empty buildings, a desalination plant, and a workforce that had learned to screw components into boards—a skill that pays $1.50 an hour in the next destination, not $4.50. The lock-in was not technological; it was political. Too much prestige invested in the park to pivot. That is what goes wrong without sequencing discipline: you build monuments to yesterday's comparative advantage. You do not have to repeat that. But you do have to decide, before you build the first thing, what the second and third things will be—and what you will abandon if the first thing fails.

Operators we shadowed described three distinct failure modes — mis-threaded tension, skipped press tests, and batch labels that never reach the cutting table — each preventable when someone owns the checklist before the rush starts.

Prerequisites: What You Should Settle Before Sequencing

Data on existing capabilities and factor endowments

Start with what you actually have—not what you wish you had. I have seen teams waste three months sequencing industries they could never serve because they assumed port capacity or electricity existed. Pull labor force surveys, firm registries, and export records. Map existing skills: welders versus software developers, not abstract “human capital.” Factor endowments matter—land, water, minerals, road density. Without this baseline, your sequence is fiction.

“You cannot sequence what you cannot measure—and you cannot measure what you have not seen on the ground.”

— A clinical nurse, infusion therapy unit

A clear definition of what you mean by ‘sequencing’

Institutional preconditions: contract enforcement, infrastructure

What usually breaks first is electricity. I have seen a perfectly sound electronics sequencing plan collapse because the grid voltage fluctuated 15% daily. That is not a sequencing problem—that is a precondition failure. Fix power before you ask firms to invest. Similarly, land titling: if ownership records are contested, no bank lends against factory construction. Settle these institutional floors before the workflow starts. The payoff? Actual firms that stay. The trade-off? You spend six months on boring infrastructure instead of flashy industrial parks. That hurts. Do it anyway.

Core Workflow: A Step-by-Step Sequence for Sequencing

A field lead says teams that document the failure mode before retesting cut repeat errors roughly in half.

Step 1: Map comparative advantage not just in products but in tasks

Stop looking at what a country makes. Look at what its workers do. That distinction has saved more than one industrial plan I have seen unravel. A region that assembles electronics might appear ready for semiconductor fabrication — same broad sector. But the tasks are worlds apart: clean-room discipline versus manual soldering, statistical process control versus visual inspection. When you map by occupation clusters rather than export codes, the adjacency shifts. The odd part is — most teams skip this because the data feels abstract. It is not. The International Labour Organization task databases are free. Download them. Cross-reference your local workforce against the skill bundles a new industry demands. If the overlap is below 40% for core tasks, that sequence is premature, no matter how high the product-space proximity score reads.

Step 2: Assess spillover potential and knowledge linkages

One industry can feed another without ever sharing a supply chain. The trick is tracing knowledge flows. A textile cluster that moves into technical fabrics does not just inherit looms — it inherits chemistry labs for dyeing and finishing. That chemistry capability can later seed speciality chemicals or pharmaceuticals. I have watched this pattern in Southeast Asia: a garment hub pivoted into medical gowns during a supply crisis, and within eighteen months a small API manufacturing unit appeared nearby. Not planned — emergent. But the sequence worked because the knowledge linkage was already there, latent. So ask: what engineering disciplines, quality systems, or managerial routines transfer? If the answer is only 'logistics' or 'cheap labor,' the spillover is thin. That is a warning, not a go signal. Better to wait and build one more adjacent skill layer first.

‘The industries that look most promising on paper often fail because the unseen skill gap swallows the learning budget before year two.’

— paraphrased from a development bank project review I sat through in 2022

Step 3: Test against three failure modes

Run every candidate sequence through three filters. Mode one: the skill-depth trap. Your workforce has surface-level familiarity but nobody who can train the trainers. That kills scale. Mode two: the missing capital ladder. The new industry requires investment cycles of four years or more, but your financial system lends in twelve-month increments. Mismatch destroys firms. Mode three: the export-distance illusion. A product looks easy to sell because it sits in a high-demand global category. But your port infrastructure, customs digitisation, and trade agreements do not support it yet. The sequence fails because logistics eats margin before production even begins. One rhetorical question before you commit: which of these three modes is your local data most likely hiding? Most governments discover the answer in month nine, after the subsidies are already spent. Do not be most governments. Test with pilot projects — three firms, one year, tight monitoring. If two of three survive and hire, move to the next industry. If not, loop back to Step 1.

Tools, Data Sources, and Analytical Setup

Export complexity indices: useful map, not a GPS

The Economic Complexity Index (ECI) tells you what a country *can* make, not what it *should* make next. I have seen teams treat ECI like a shopping list — rank products by complexity, pick the highest one, and call it strategy. That burns capital fast. The index is built from revealed comparative advantage data: if few countries export a good and you do not currently export it, the algorithm flags it as "adjacent." Fine. But adjacency in product space does not mean adjacency in your actual factory floor. A country that makes basic garments might score "close" to electronics assembly on the ECI map. The machinery, workforce, and supply chains are worlds apart. Use ECI to scan the horizon, not to set the destination. Cross-check with your domestic input-output tables: what industries already buy from each other? If your textile sector already consumes local dyes and packaging, shifting toward higher-value textile finishing demands far less institutional muscle than jumping to semiconductors.

Firm-level surveys vs. sector-level aggregates: the resolution trap

Sector-level data hides the variance that kills sequencing. Average labor productivity in "basic metals" could be $12,000 per worker — but that average lumps a modern mini-mill doing $28,000 alongside a state-owned relic doing $4,000. The planner sees a mediocre sector and deprioritizes it. Wrong move. The actual bottleneck might be one outdated furnace in an otherwise capable industry. Run a firm-level survey — thirty companies, not three thousand — asking what inputs they import that domestic suppliers *could* provide with modest upgrades. That one question surfaces more sequencing opportunities than any download from the World Bank database. The catch: firm surveys rot fast. Data collected in March is suspect by October if exchange rates or energy tariffs shift. Update every six months. Treat sector-level aggregates as the rough draft; treat the survey as the red pen.

Mapping input-output tables for domestic linkages

An input-output table shows who buys from whom. That sounds dry until you see it expose a seam. Most development agencies publish IO tables for benchmark years (often five years old). Take the latest one, strip out imports — you want domestic coefficients only. Then build a simple directed graph: each industry is a node, each supplier-buyer relationship is an edge with a weight equal to the share of total output flowing between them. Search for nodes with high "out-degree" but low "in-degree" — sectors that sell to many others but buy from few. Those are your leverage points. A small intervention in that sector ripples outward. What usually breaks first is the lag: IO tables miss fast-changing linkages like the explosion of logistics services or platform-based subcontracting. Supplement with customs data on intermediate goods flows — that refreshes monthly.

“The map is not the territory, but a good planner draws both.”

— overheard at a trade ministry retreat, after four hours of arguing about SITC codes

Most teams skip the IO step because the data feels too old or messy. That hurts. Even an imperfect table from two years ago reveals structural dependencies no index can replicate. You want to know why a proposed steel mill failed? Look at the IO table: if local construction firms buy only 12% of their structural steel domestically, the mill's demand forecast was always fiction. Pair the table with a heatmap of subnational industrial clusters — provincial trade data if available. The texture of domestic linkages varies wildly within a country. Export data is national. IO tables can be subnational, barely. The gap between them is where wrong sequencing happens.

Variations for Different Constraints

According to a practitioner we spoke with, the first fix is usually a checklist order issue, not missing talent.

Resource-rich economies: avoid the Dutch disease trap

When oil, copper, or lithium sits under your soil, the sequencing temptation is brutal: extract hard, build basic infrastructure, then chase downstream processing. Wrong order. I have watched a petro-state pour forty billion into a petrochemical complex before its local firms could supply pipe fittings without a three-month import delay. The complex ran at sixty-percent capacity for four years. The trap is not the resource—it is the sequence that lets currency appreciation kill everything else. You must sequence resource rent capture and a managed exchange-rate buffer before heavy industrial build-out. That sounds painful politically. It is. But skipping it locks in a mono-economy that cracks the moment commodity prices dip. Start with a sovereign wealth fund rule that sterilizes half the windfall; sequence light import-substitution industries while the resource sector is still scaling up, not after the real has already strengthened. The catch? Every ministry wants the big refinery today. The better move is a stop-loss: cap local-content mandates below twenty percent for the first five years, then ratchet up.

Labor-abundant economies: start with light manufacturing or agro-processing?

Most teams skip this: labor-abundant does not mean any light manufacturing fits. Bangladesh bet on garments and won because global buyers already needed cut-make-trim at scale—the sequencing aligned with existing supply chains. I have seen a Sub-Saharan country try to sequence electronics assembly without first building a reliable electricity grid for the industrial zone. The seam blew out within eleven months. The path is agro-processing—not glamorous, but it trains workers on shift discipline, quality control, and export logistics before you ask them to solder circuit boards. The trade-off is pace: agro-processing margins are thin, and local elites will lobby for faster, taller industrial parks. Resist.

“The right sequence for a labor-surplus economy is not the highest-value sector first. It is the sector that builds the behavioral infrastructure for factory work—then upgrade.”

— modified from a conversation with a former industrial policy advisor, Nairobi, 2019

That behavioral infrastructure—showing up on time, accepting machine-paced work, tolerating quality audits—takes two to three years to embed. Start with processed cashews, shea butter, or cut flowers. Once the cohort graduates, pivot to assembly. The dangers? Agro-processing requires cold chains that many small economies lack, and climate shocks can crater supply. Diversify the raw base across three crop types before you commit to a dedicated processing zone.

Small economies: sequence for niche exports, not broad diversification

Small means you cannot absorb the fixed cost of steel mills, auto plants, or even large fertilizer facilities. The worst mistake I have seen a small economy make is copying an export-promotion playbook from a country with fifty million people—ten million is not fifty million. The correct sequence is extreme niche: a single high-value product where you can dominate a sliver of global trade. Uruguay did this with software testing services and later with wood pellets. The parameters shift from “build a diversified industrial base” to “pick three products, sequence the logistics spine for them, and ignore everything else for a decade.” The pitfall is premature scaling: a small economy’s export push can crash because the port handles only two container lines. Sequence port specialization before production ramp-up. That means negotiating with a single shipping line for dedicated berth space, not building a generic multi-user port. The old advice to “diversify your export basket” hurts small economies—it fragments attention and underserves the one niche that could actually earn hard currency. Pick one, sequence around that bottleneck, then consider a second niche after three years of sustained export growth. Not before.

Pitfalls: What to Check When the Sequence Fails

Anchor firm capture: when one company becomes too big to fail

You sequence a steel mill before the scrap yards. Makes sense on paper—anchor demand, pull the supply chain. The mill arrives. Employment spikes. Everyone cheers. Then the mill owner discovers the local scrap suppliers can't meet grade specs, so they import from three provinces away. The mill survives. The promised backward linkages? Dead. Local SMEs that bet on supplying that mill go under inside eighteen months. The indicator is simple: watch the procurement-to-local-content ratio. If it stays above 0.7 after year two, your anchor is an island, not an engine. The fix is ugly but clear—you must cap the anchor's import allowance and subsidise local supplier certification before the mill breaks ground, not after. I have seen a government try to renegotiate this post-construction. The mill threatened to leave. They blinked. The whole corridor stalled.

The odd part is—anchor capture looks like success. Employment up. Tax receipts rising. Only the supplier mortality rate tells the real story. Track it quarterly. Anything above 12% per annum among firms within two tiers of the anchor means your sequence locked in the wrong dependency.

Skill mismatch: training programs that train for vanished jobs

A textile zone invests in sewing machine operators. Three years later, automation hits. Those operators can't retrain into maintenance technicians—they were never taught basic electronics. The sequence error here is semantic: you sequenced training after investment approval but before technology assessment. The indicator is the placement rate adjusted for job-type persistence. If 60% of graduates land jobs that disappear within two years, your sequence produced a ghost cohort.

The catch is that training programs feel urgent. Governments rush to announce "10,000 skilled workers" before the factories even finalise their machinery specs. That is the pitfall. You end up training for a factory that never arrives, or one that arrives with different equipment. Corrective action: insert a mandatory six-month delay between factory approval and training launch. Use that window to audit the actual technology being ordered. We fixed this once by making every training voucher conditional on the employer signing a minimum-equipment-list agreement first. It halved enrolment in the first quarter. The second quarter placement rate doubled.

'The graveyard of industrial policy is littered with training certificates that expired before the first production line turned on.'

— district planner, after overseeing three failed zone transitions

Infrastructure timing: building ports before factories exist

This one seems obvious. Yet it happens repeatedly. A deepwater port gets built for an industrial corridor that has zero committed manufacturers. The port sits empty. Debt service starts. The government, desperate to justify the asset, lowers tariffs for raw material exports. Suddenly the corridor exports unprocessed ore instead of attracting processors. The sequence failed because infrastructure demand is elastic in one direction only—once built, it creates political pressure to use it however possible, regardless of the original plan.

Indicators: cargo composition shift more than 30% away from the projected industrial mix within eighteen months of port opening. Or, port utilisation below 40% at month twelve. Corrective action is brutal but necessary: do not commission the port until at least three factory foundation pours are completed for the industries the port was meant to serve. That forces sequencing discipline upstream. The trade-off is delay—manufacturers hate waiting on infrastructure. But a port without factories is a subsidy for exporters you never intended to create. That hurts more than six months of construction lag.

Wrong order. Every time. Check the utilisation curve before you celebrate the ribbon cutting.

Troubleshooting Checklist: Five Questions to Ask When Growth Stalls

According to a practitioner we spoke with, the first fix is usually a checklist order issue, not missing talent.

Is the anchor industry still generating spillovers?

You picked a lead sector for a reason—maybe it was agro-processing, maybe light manufacturing. That was eighteen months ago. The real question now: are downstream firms actually picking up skills, buying intermediate goods, or copying processes from that anchor? I have visited zones where the anchor plant runs in isolation, importing everything, exporting everything, and local firms just sell lunch to its workers. That is not a spillover; that is a canteen. Walk the supply chain. Ask three local suppliers whether their technical capability improved because of the anchor. If the answer is vague—“we sell them packaging sometimes”—you are locked into a sequence that never activated the second stage. The fix is not doubling down; it is asking whether the anchor itself needs replacement.

— Trade-off: kicking out a flagship industry is politically painful. But keeping it when spillovers died is worse.

Are forward linkages actually being used by local firms?

Your sequencing plan probably assumed that if you put a steel mill in, local fabricators would spring up to use that steel. Odd thing is—they often don’t. The mill sells to a trader in the capital, the trader ships it back to the coast, and local workshops keep importing from China because the mill’s minimum order quantity is too high. That is a forward linkage that exists on paper only. Check customs data for the last six months. If intra-region flows of that intermediate good are flat or falling while total output rose, the linkage is fake. Break it down: can the mill split orders? Can a intermediary warehouse pool demand? If nothing works, the sequence’s middle step is a dead node.

Most teams skip this audit until growth stalls completely. By then, you have sunk three years into a path that was never going to branch.

Has the comparative advantage shifted silently?

Comparative advantage is not a birthmark; it migrates. A region that could once undercut competitors on labor cost may now face a minimum-wage hike, a power tariff spike, or a new trade deal that favors a rival. I watched a garment cluster collapse because a neighboring country signed a preferential tariff—our sequence still assumed cheap access that had vanished six months prior. The checklist question: re-run your original location quotient or RCA calculation with last quarter’s data. If the number dropped below 1.0 for your anchor sector, you are sequencing toward a ghost. Pivot early rather than defending a dead comparative advantage with subsidies.

Wrong order. Not yet. That hurts—but catching it on a quarterly review saves two years of misallocated capital.

“We kept pushing the same industrial park plan for three years after the tariff shifted. Nobody checked the trade data until the anchor plant announced it was relocating.”

— Regional development officer, Southeast Asia, 2023

Are downstream capabilities developing ahead of or behind schedule?

Growth stalls often because the sequencing timeline was aspirational. You expected basic metal fabrication to appear within eighteen months of the steel mill opening. Reality: it took forty months because local credit was too tight. That lag matters—it means the sequence is not self-reinforcing yet. The pitfall is pushing the next step (say, a heavy machinery park) while the prior linkage still has not formed. You end up with two half-built nodes instead of one complete chain. The remedy is brutal: freeze the next investment until the current linkage shows measurable uptake—employment numbers, supplier registrations, inter-firm purchases. Slow down to speed up later.

One rhetorical question, I suppose: would you rather have one functioning cluster in three years, or three half-empty parks in five?

Has the policy environment changed in ways that break your path?

Sequences assume stable rules. But governments change tax holidays, local-content requirements, or environmental permitting. If your second-stage industries depended on a subsidy that just got repealed, the sequence is now a trap. Check three things: (1) any new regulations affecting your anchor’s input costs, (2) changes in labor law that shift your comparative advantage, (3) infrastructure projects—or cancellations—that alter logistics. I once saw a whole agro-processing sequence invalidated because a road to the port was deprioritized. The sequence was not wrong; the assumptions were. Update your assumptions quarterly, not annually. That is the difference between a plan and a relic.

Walk away from a broken sequence faster than you entered it. The sunk cost is gone; only the next decision matters.

What to Do Next: A Specific Three-Month Review Protocol

Conduct a capabilities audit every 90 days

Mark a calendar alert for day 90, 180, and 270. No excuses. The audit is not a gentle performance review—it’s a knife fight with your own assumptions. I have watched teams run this by asking “what can we still do that competitors cannot?” and walking away with a list of three forgotten strengths. That’s half the job. The other half: list every skill you don’t have yet but will need in six months. If your sequencing plan assumed a stable supply chain and your competitor just automated theirs, your capability gap just widened. The catch is—most audits stop at inventory. They count machines, patents, headcount. They miss the silent rot: a middle manager who blocks cross-training, a supplier relationship that turned hostile, a regulatory waiver about to expire. Push for evidence, not PowerPoint. One concrete rule: any capability that did not generate revenue or reduce cost in the past quarter gets a yellow flag. Two consecutive yellow flags? Red. Then you either reinvest or cut it loose.

Run a stress test: what if your main export market collapses?

Pick a Tuesday. Close the door. Model the collapse of your top export destination—sudden tariff, embargo, shipping lane closure, whatever feels least likely. The odd part is—the scenarios that hurt most are never the obvious ones. I have seen a textile firm discover that losing one port meant their entire yarn supply stalled for eight weeks. They had no fallback. The stress test is not a spreadsheet exercise; it forces your team to name specific substitute buyers, alternate logistics routes, and the first five phone calls they make at 8 a.m. on collapse day. That sounds fine until someone realizes their only backup buyer is the same guy who owes them money. Run this test twice per cycle: once at month one, once before the quarter closes. Document the frictions. If the second run still reveals a single point of failure, you have not sequenced your transformation—you merely delayed the reckoning.

“Every growth plan survives until the first customer who says ‘I don’t need what you just built.’ The review finds that customer before the board does.”

— Operations lead, after a failed pivot in automotive parts

Build an exit trigger for sunset industries

Most teams sequence a transformation and never write the off-ramp. Wrong order. Define the trigger before you need it. A concrete example: if your legacy product line drops below 40% gross margin for two consecutive quarters, you exit—no heroics, no turnaround committee. That hurts. But it prevents the slow bleed where you pump capital into a dying segment because the team is emotionally attached. I have seen a packaging company burn eighteen months trying to “revive” corrugated boxes while their flexible-film line was starving for R&D. The trigger would have saved them. Make yours numeric: margin floor, market share threshold, repeat order rate. Attach a date. If the trigger fires, your three-month review becomes a liquidation or spin-off plan, not a rescue mission. One final edge—share the trigger with your board early. It sounds reckless. It is actually the only way to stop a later argument where emotion overrides economics. That is the move: protect your sequence by knowing precisely when it is time to abort the old one.

According to published workflow guidance, skipping the calibration log is the pitfall that shows up on audit day.

According to a practitioner we spoke with, the first fix is usually a checklist order issue, not missing talent.

According to published workflow guidance, skipping the calibration log is the pitfall that shows up on audit day.