When Good Integrations Go Bad: Lessons from Failed Manufacturing System Rollouts

Key takeaways

• Most failed manufacturing system rollouts in food plants are not software failures but leadership, testing, and change‑management failures.
• Patterns repeat: inadequate end‑to‑end testing, poor change management, underestimated integration complexity, and thin training on the plant floor.
• Leaders can spot trouble early by watching a short list of signals (around scope creep, testing, data readiness, and workforce fatigue) and steering with a simple risk and readiness framework.

Food manufacturers are feeling intense pressure to digitize as retailers demand better traceability, regulators tighten reporting, and margins are squeezed by raw‑material volatility and labor shortages. It is no surprise then that enterprise resource planning (ERP), manufacturing execution systems (MES), and plant‑to‑cloud integrations are everywhere on board agendas.

But even well‑run food companies can see “good” integrations go bad, and fast.

In early 2023, J&J Snack Foods rolled out a new ERP platform across its network. Within weeks, the company was dealing with order delays, frozen goods distribution issues, and enough supply chain disruption that executives had to explain the impact on earnings calls.

More than two decades earlier, Hershey’s famous ERP implementation became a cautionary tale. A compressed schedule, limited testing, and a go‑live just before Halloween contributed to missed orders of over $100 million and a very public profit hit.

If anything, the risks like these have grown as integrations span cloud systems, plant‑floor equipment, and increasingly, AI tools.

Deloitte’s 2025 survey of 600 manufacturing executives found that 92% believe smart manufacturing will be the main driver of competitiveness over the next three years, yet a majority also report significant challenges managing complex transformations and operational risk.

In other words, everyone needs integrations, but not everyone is ready to absorb the disruption.

Let’s take a deep dive into what goes wrong in high‑profile food manufacturing rollouts, as well as how leaders can spot problems early and steer projects back on track.

Why food manufacturing rollouts are uniquely fragile

System rollouts are hard in any industry, but food and beverage manufacturing adds several complications:

• Perishability and tight lead times. Inventory cannot sit for weeks while systems stabilize. Shelf life, cold‑chain requirements, and promotional windows leave little room for extended disruption.
• Heavy regulation and traceability. Systems must handle lot tracking, allergen controls, and recall scenarios with precision. Mis‑configured integrations can instantly create compliance gaps.
• High product and process variability. Recipe changes, seasonal SKUs, private‑label products, and frequent line changeovers all stress poorly tested integrations.
• Decentralized operations. Many food companies run multiple plants with different maturity levels, local systems, and work practices. So a “one size fits all” design often cracks on contact with reality.

Recent guidance on MES rollouts in food manufacturing emphasizes these same themes: integration with existing systems, change management, and food‑specific compliance are repeatedly cited as major challenges.

At the same time, the workforce is already stretched. Auburn University’s 2024 Smart Manufacturing Adoption Study found that workforce operations is the top business challenge, with 61% of respondents ranking it in their top three issues.

When you layer a complex integration project on top of that reality, failure patterns begin to look alarmingly predictable.

Pattern 1: Inadequate end‑to‑end testing

If there is a single thread that links Hershey’s 1999 failure to more recent incidents, it is testing treated as a formality rather than a business survival tool.

In Hershey’s case, the company attempted to implement three major systems — SAP R/3 for ERP, Manugistics for supply chain, and Siebel for customer relationship management — on a compressed 30‑month schedule instead of the recommended 48 months. To meet the deadline, key testing phases were shortened or skipped. The systems then went live just before the Halloween shipping peak. When issues emerged, there was no safety net: orders could not be processed even though inventory was available.

In more modern MES projects in food plants, consultants report similar patterns: technically “successful” factory acceptance tests that never simulate real changeovers, allergen clean‑downs, or recall scenarios — and certainly not the stress of running three shifts at holiday volume.

Testing framework for food manufacturers

Leaders do not need to be technologists to demand better testing. A simple three‑layer framework can dramatically reduce risk:

1. Process‑fit testing (“conference room pilots”): Map processes end-to-end: order to pallet loading, including quality holds and rework. Use real master data (customers, materials, recipes) and realistic order mixes. Confirm regulatory support: labeling, allergen controls, and traceability.
2. Integration and performance testing: Simulate actual transaction volumes and seasonality; include interfaces for weigh scales, printers, lab systems, WMS, and external partners; and validate fail-over/recovery procedures (e.g., label printer or line PC failure).
3. User acceptance testing on the plant floor: Simulate realistic and “ugly” scenarios, including partial shipments, last-minute recipe changes, failed quality checks, and urgent recalls, with line leads, quality technicians, maintenance, and schedulers.

If your most stressed plant manager does not believe the system will survive a holiday promotion week, you are not ready to go live.

Early warning signs testing is insufficient

• Generic test scripts are used instead of food-specific ones.
• Defect backlogs remain flat or increase as cutover nears.
• Business users are disengaged during testing, focusing on email or skipping sessions.
• Critical scenarios are not covered by any test.

These signs often appear months before a manufacturing system go-live.

Pattern 2: Poor change management and workforce readiness

Deloitte’s research highlights the human side of digital transformation. Nearly half of respondents report moderate to significant challenges filling production and operations roles, and 35% cite “adapting workers to the Factory of the Future” as a top workforce concern.

Resistance from workers, lack of stakeholder alignment, and inadequate training are common causes of implementation setbacks or outright failure. Yet in many rollouts, change management is still a thin communications plan and a couple of “town hall” meetings. 

In the J&J Snack Foods case, external analysts point out how multi‑site ERP rollouts can force standard processes onto plants with very different maturity levels, system literacy, and local expectations. When these differences are not surfaced and managed, the result is confusion and workarounds, followed quickly by performance issues and customer impact.

Simple change management strategies for the plant floor

Food manufacturing leaders can dramatically raise their odds of success by insisting on a few basics:

• Change impact map by role. Create a one-page before/after view detailing changes for critical roles (operator, scheduler, QA tech, etc.).
• Visible sponsorship from operations leaders. Adoption increases when plant managers, not IT, lead the rollout as their initiative.
• Floor-level champions. Identify respected operators/supervisors as “super users” for extra training and post-go-live support.
• Two-way communication loops. Use daily cutover huddles to ask: What broke? What was the workaround? What help is needed?

Early warning signs of change management failure

• Plant managers referring to the project as “the IT system” rather than “our new way of working.”
• Operators hearing about the rollout mainly through rumors rather than direct briefings.
• Super users being picked solely based on availability, not influence or respect.
• Training and hypercare support being cut to meet budget or timeline targets.

Change resistance is rarely about laziness. It is usually a rational response to poorly communicated risk.

Pattern 3: Underestimated integration complexity

Food manufacturers often run a patchwork of systems: legacy ERP in headquarters, custom scheduling tools, spreadsheets in the plants, standalone quality databases, and equipment from multiple automation vendors. Integrations weave these into something that looks like an end‑to‑end system, until a change pulls on the wrong thread.

The J&J Snack Foods analysis highlights three recurring pitfalls in multi‑site ERP rollouts: unstandardized local processes masquerading as “global templates,” hidden data inconsistencies, and resource imbalances between early pilot sites and later waves.

The challenge isn’t unique to food. Clean Energy Associates’ 2024 quality assessment of battery energy storage systems found that system‑level integration problems now account for 72% of manufacturing defects, up from 48%. When manufacturing systems become more connected, integration quickly becomes the main point of failure.

An “integration complexity canvas” for food plants

Before approving a rollout, assess integration risk using these five dimensions:

1. Product/process complexity: Consider the number of recipes, formats, variants, cleaning regimes, and the frequency of new SKUs or promotions.
2. Regulatory/customer requirements: Evaluate required traceability (unit vs. batch) and the strictness of customer compliance and chargebacks.
3. Network/multi-site structure: Determine if plants are similar or different, and if decision-making is centralized or local.
4. IT/OT landscape: Tally how many systems (ERP, MES, lab, etc.) and plant-floor devices (scales, PLCs, printers, etc.) must exchange data or be touched at go-live.
5. Data quality/governance: Assess the cleanliness of item masters, recipes, BOMs, and customer data, and clarify who owns data decisions across functions.

Use a simple red–amber–green score for each dimension. If you see more red than green, insist on either a smaller initial scope, a pilot site with lower risk, or more time and resources for integration design and testing.

Early warning signs that complexity is underestimated

• Phrases like “we’ll harmonize processes later” appear frequently in project updates.
• Integration diagrams change every few weeks, but the cutover date never moves.
• No single executive can explain, end‑to‑end, how an order flows through the new landscape.
• Data cleanup is stuck in “planning” mode while build and configuration race ahead.

Complexity is not the enemy; unacknowledged complexity is.

Pattern 4: Insufficient training and support

Training is often the first line item cut when budgets get tight, with the logic being, “Our operators are smart; they’ll figure it out.”

In reality, rushed or generic training almost guarantees post‑go‑live turbulence. Deloitte’s smart manufacturing research shows that human capital is one of the least mature capability areas in many manufacturers, and only 48% of companies report having a formal smart‑manufacturing training and adoption standard.

New systems are landing on overloaded teams. So successful implementations require more than one-off classroom sessions, but role‑specific training, ongoing support, and continuous monitoring and improvement.

What effective training looks like in a food plant

Effective food manufacturing training programs are typically characterized by five key traits:

1. Role-based: Training focuses on actual tasks (e.g., starting a batch) rather than just system menus.
2. Scenario-driven: Operators practice challenging situations, such as rejected quality checks or urgent rework.
3. Hands-on in a safe environment: A “sandbox” system with realistic data allows risk-free practice.
4. Multi-shift and multi-language coverage: Schedules accommodate three-shift operations and language needs, with shift champions reinforcing learning.
5. Sustained hypercare: Extra floor support is provided for the first 60-90 days to quickly log and triage issues.

Early warning signs that training is under‑resourced

• Training plans show a single 2‑hour session to cover everything from receiving to shipping.
• Night and weekend shifts have no dedicated training sessions.
• There is no plan to backfill operators who attend training, so managers quietly tell them not to go.
• After training, operators still refer to “the old way” as the only reliable method.

If you see these signs, treat training as a critical path item, not an optional extra.

Use a leadership framework that ensures success

Food manufacturing leaders can use a four-stage framework to govern system rollouts:

1. Align: Define clear business outcomes (e.g., fewer stock-outs, faster traceability), acceptable disruption levels, and establish an empowered cross-functional steering committee.
2. Design for plant-floor reality: Start with current-state process maps, use an integration complexity canvas to identify risks, and choose a rollout strategy based on risk.
3. Prove: Test “nightmare scenarios” like a recall or system outage. Use strict quality gates, requiring all critical defects resolved and KPIs met before go-live. Run a full cutover rehearsal.
4. Land: Manage the first 90 days with a dedicated “war room” monitoring operational metrics (e.g., scrap rates, manual workarounds) and rapid decision-making. After 90 days, conduct a structured “lessons learned” review to improve the playbook.

Are we off track? A quick diagnostic for leaders

To spot a failing rollout early, ask your team simple questions across key areas:

• Scope and timeline: Have we added significant scope (sites, functions, AI, etc.) without adjusting the timeline/resources? When did we last move the go-live date for quality, not external pressure?
• Testing and quality: Can we immediately see evidence of testing a recall, major promotion, or equipment failure? Are serious defects decreasing before go-live, or are we “accepting” more?
• Data and process: Who decides when a process change conflicts with plant practice? Are we cleaning/standardizing data before configuration finishes?
• People and training: Can all plant managers describe their teams’ training and support? Do we know training gaps by shift and language?
• Operational readiness: If we launch tomorrow, which customer, plant, or product line is our biggest worry, and why? What is our rollback plan for a Day One failure?

Manufacturing system rollouts are critical operational transformations. Without careful planning, they’ll end in failure. 

Food manufacturing leaders must treat integrations as strategic, plant-centric projects, focusing on robust testing, honest complexity assessment, and investing in people to avoid system failure.

FAQ for food manufacturing leaders

Q: What is systems integration in a food plant?

A: Integration is how systems (ERP, MES, WMS, etc.) communicate with each other and the plant floor. Weak integration forces manual data entry, leads to spreadsheets, and causes errors under pressure.

Q: How much downtime should we plan for during go-live?

A: Plan for several days of reduced performance (e.g., lower speeds, limited products, overtime), not zero downtime. Define this explicitly, inform commercial teams, and schedule go-live away from peak seasons.

Q: When should plant managers/operators join the project?

A: Much earlier than usual. Successful companies involve plant leaders during selection and early design to evaluate real-world fit (sanitation, allergens, changeovers, etc.). They become advocates; waiting until training guarantees resistance.

Q: What metrics should we watch in the first 90 days?

Focus on operational and customer metrics:

• Order fill rate and on-time performance
• Line uptime/unplanned downtime
• Scrap, rework, and giveaway
• Number/severity of workarounds (shadow spreadsheets)
• Customer complaints/chargeback

Sustained deterioration signals a system issue, not just a people issue.

Q: How can smaller manufacturers de-risk integrations?

A: Narrow scope (fewer plants/products); use proven cloud solutions and templates; partner with vendors for hands-on support; rely on super users. Focused scope and strong training often outweigh cutting-edge tech, addressing primary barriers like workforce/capital constraints.

Q: Where does AI fit? Does it increase risk?

A: AI layers on ERP/MES for forecasting, quality, and maintenance, but requires stable, high-quality core system data. AI isn’t a fix for a misaligned ERP system. Treat AI initiatives as second-wave projects built on a solid integration foundation, not part of a “big bang” rollout.