Why Million-Dollar Digital Projects Become Shelfware: Four Traps Enterprises Miss

Enterprises spend millions on ERP, data platforms, and BI dashboards. At acceptance, systems appear complete and dashboards look impressive. Yet six months later, employees still use Excel, departments communicate via WeChat, and management needs ad-hoc data pulls. The digital systems become mere showpieces for visits and reports.

Wang Ziru, Fanruan CEO Chen Yan, and FDE expert Shen Tao discussed the deep reasons in a recent interview. The core issue: enterprises never resolved the relationship among business, data, and organization from the start. Four specific problems dominate.

01 Buying Systems Equals Digitalization — Projects Go Live but Business Doesn't Change

Many firms equate digitalization with software procurement: ERP for finance, MES for production, OA for office, BI for analysis. More systems, bigger budgets, yet the same problems persist. After ERP rollout, departments still export Excel for rework; after BI launch, finance still manually prepares data; after MES, planners still adjust schedules in spreadsheets.

Why? Companies focus on whether the system went live, not on what business the system must change . Wang Ziru emphasized: "Enterprise digital transformation cannot only focus on technology construction; it must return to real business scenarios." For example, production scheduling isn't just moving an Excel plan into software — it involves equipment status, order priority, material supply, and production rules. If those business logics aren't clarified, the new software merely replicates old manual steps online.

Before project initiation, enterprises should answer three questions:

What is the most pressing problem in current business?

Which business links need to change through digitalization?

After launch, how to measure actual effect ?

Digitalization is not buying software; it's using technology to change how business operates.

02 Processes Moved Online but Business Remains Disconnected

Sales uses CRM, procurement uses ERP, production uses MES, warehouse uses WMS. Each department has its own system, yet inter-department information still relies on manual handoffs. A customer order should flow through inventory check, production scheduling, material procurement, execution, and shipment — a complete chain. In reality, sales re-sends order details to production, planners copy data to Excel, procurement re-verifies material needs. Systems abound, but duplicate entry, manual transfer, and information gaps remain.

Wang Ziru warned: "As long as my approval flow is online, I've completed business digitalization — actually not completely." The problem: approvals are online, real business execution stays offline. Moreover, enterprises must record not just final results but the business process . Wang Ziru stressed: "Every change, the previous version is valuable process data. It should be retained, recorded, traceable." Example: production plans adjust due to equipment failure or rush orders. If only the final plan is kept, the reasons and impacts of each adjustment are lost.

Enterprises must enable continuous data chains across procurement, production, sales, warehousing .

03 Many Data Systems Built but No Unified Data System

This is the deep reason huge investments yield no value. Firms have data warehouses, data mid-platforms, BI platforms, and piles of reports. Yet cross-department analysis fails. When profit drops, finance blames cost, sales blames product mix, production blames efficiency, procurement blames material prices. Each department has its own data; each explanation seems plausible, but the enterprise cannot form a unified conclusion quickly.

Cause: data is collected but not governed . Same metric, different definitions across departments; same business, different IDs in different systems; systems lack effective linkage.

Wang Ziru shared his experience driving indicator management: organizing departments to create an indicator management white paper, unifying internal metric definitions . On the indicator dictionary's role: "The dictionary's value isn't that everyone checks it daily; it's that when people argue, the dictionary settles it." Data governance appears to standardize fields, metrics, and calculations, but actually unifies different departments' understanding of business .

Enterprises must ground data governance in business:

Data integration: Connect ERP, MES, CRM, WMS to solve data silos.

Indicator governance: Unify definitions of revenue, cost, profit, inventory.

Data modeling: Build relationships among customers, orders, products, materials.

Data quality: Ensure accuracy, completeness, continuity, traceability.

After governance, a new question arises: how to let business people actually use the data? Traditionally, fixed reports and BI dashboards. Management views operational reports, departments view KPI dashboards, ad-hoc needs require data staff to re-extract and build reports.

In the AI era, data consumption is changing . Shen Tao proposed future consumption beyond reports and BI: HTML reports, personalized "thousand faces" presentations, even natural language querying. This lets data permeate every business link. This drives Fanruan's AI product layout.

Fanruan's Dual-Track AI Strategy

On-premise (large enterprises): Centered on Dora as a localized Data Agent platform. Users configure agents manually or generate via dialogue. FineDataLink connects disparate systems, integrates data via warehousing for agents. FineBI serves as data foundation with indicator center, data management, data source management, semantic layer — solidifying existing indicator systems and business rules. FineBI's indicator center is upgrading to 2.0 to adapt traditional BI governance to AI-era needs.

Cloud (SMEs): "Use to govern" approach. Example: FineBI Cloud e-commerce clients connect standard platform APIs; after account linking, they get operational analysis and attribution results in 1-2 days for standardized scenarios.

Both tracks share a philosophy: let new data consumption forms amplify past data reproduction and governance capabilities. Shen Tao: "Our core idea is to surround new data consumption forms and further amplify the data reproduction and governance capabilities we accumulated in the previous era."

Dora differs from personal agents: personal agents solve daily office efficiency; enterprise Data Agents connect massive databases and must consider data permissions, security, result accuracy . Wang Ziru: "But if an enterprise Data Agent messes up the entire company's database, that's a huge accident." Therefore, enterprise agents must be built on reliable data models, business rules, permission management, and verification mechanisms.

04 Project Launch Treated as End — No One Owns Ongoing Operations

Construction responsibility is clear: IT selects tech, business defines requirements, vendor delivers, management approves budget. But post-launch: who continuously optimizes? Who drives adoption? Who checks data quality? Who coordinates fixes? Often no clear answers.

Shen Tao shared a manufacturing firm's operational analysis practice. The hard part wasn't data preparation but after the analysis meeting: who tracks issues, drives rectification, establishes assessment mechanisms .

Enterprises need not just project mechanisms but continuous operations mechanisms :

Define business owner and management responsibility for each system.

Establish data quality checks and indicator maintenance processes.

Regularly collect business department usage feedback.

Link system usage with operational analysis and business improvement.

Assign responsible person, corrective action, and deadline for every discovered issue.

Digital projects don't end at acceptance; the real test is how many people still use it, how well, and whether business actually changed.

05 Where Should Digital Transformation Truly Start?

Failure isn't a software problem; it's that business, data, and organization never formed a closed loop . To avoid repeated waste, re-examine from three directions:

First, start from business pain points, not technology procurement.

Don't begin with building data mid-platforms or deploying large models. First find a concrete business pain point — inventory, order delivery, cost, profit. Choose systems and tools around those; digitalization then has clear value targets.

Second, start from data connection and governance, not system count.

Enterprises may not need a brand-new platform. First inventory existing systems and data assets , connect data, unify indicators, let scattered departmental data actually work.

Third, turn digitalization from project construction into continuous operations.

Clarify business owners, data owners, project management responsibilities. Tie system usage to actual operational improvement.

Fanruan explores delivery models adapted to different enterprise foundations: for large firms with mature data and complex business, build enterprise Data Agents via Dora and on-premise products; for resource-constrained SMEs, use standardized cloud products cutting in from operational and attribution analysis. Shen Tao outlined three service modes: self-service with standardized products, combined solutions for complex business, co-innovation with leading firms. The core isn't vendor doing everything, but helping enterprises gradually build their own digital capabilities.

This aligns with Fanruan's AI-era layout: let existing data assets keep delivering value, and let enterprises adopt AI at lower thresholds in ways that fit them. AI isn't a magic fix for digitalization gaps. Enterprises still need trustworthy data, clear business scenarios, and necessary management mechanisms .

Looking back: why do firms spend millions yet end with shelfware? Because they may have only completed software procurement without changing processes; only collected data without governing it; only delivered projects without establishing continuous operations.

The true endpoint of digital transformation is never system launch, but making business more efficient, data more trustworthy, and management decisions better grounded. What enterprises buy shouldn't be just software suites, but a capability to continuously improve operations .