01
How many steps does an MES implementation take?
Survey (walk the real process, define station checkpoints and barcode rules), configuration (mostly productised modules), trial running (manual and system checked in parallel), then acceptance item by item against the contract list.
- Survey first
- Walk the process on site (people, machine, material, method, environment, measurement), define station checkpoints, barcode rules and defect codes. Time saved here is paid back twice later.
- Configuration-led
- A productised MES configures proven modules with limited custom work. Involving key users during configuration reduces resistance after go-live.
- Trial running
- Run manual and system in parallel for a few days and reconcile the data before cutting over, then accept item by item against the contract list. Baichuan implements by process stage, accepting each stage separately.
02
What data has to be prepared before go-live?
Three groups: material master and BOM imported and validated, barcode and SN rules with an SN block pool, and accounts, permissions, stations, operations and defect codes.
- Material and BOM
- Import material master from Excel and BOMs with automatic matching and validation, cleaning slot, quantity and description discrepancies first.
- Barcode and SN blocks
- Material barcode rules, SN barcode rules, the customer number pool where one exists, and the SN block pool configuration.
- People and stations
- User accounts, role permissions and definitions for stations, operations and defect codes. Solid initialisation is what makes trial running smooth.
Baichuan makes data initialisation the first milestone of implementation.
03
How long does staff training take, and is it easy to learn?
Training is layered: operators follow the PDA and station screen, team leaders focus on exception handling and changeover verification, and parallel running with on-site support consolidates it in the first weeks.
- Layered training
- The skills and training module builds training plans and records results, keeping training records online.
- Designed to be simple
- PDA prompts guide each step, alarms pop up automatically and SOPs display at the station, so new staff follow the system rather than memorise it.
- Reinforcement
- Parallel running and a few days of on-site support at the start. The system only takes hold once the habit forms.
Baichuan trains by role and provides on-site support during implementation.
04
Does going live require stopping production?
No. Training, initialisation and cutover are scheduled in line-stop windows, one stage at a time, with paper records kept during parallel running as a fallback.
- Off-peak implementation
- Training, initialisation and cutover happen in line-stop windows, leaving normal production unaffected.
- One stage at a time
- Switching one stage at a time keeps risk contained. Cutting the whole plant over at once has a well-known failure mode.
- Rollback plan
- Paper records remain during parallel running, so a system fault does not block shipment. Writing the rollback mechanism into the go-live plan is what lets production cooperate.
Baichuan writes the rollback mechanism into the go-live plan.
05
How long does the vendor stay on site during implementation?
At key points the team must be there: process survey, trial-running support, cutover standby and first acceptance. During those visits the customer's administrator is trained so self-maintenance becomes realistic.
- Key points
- Process survey, trial-running support, cutover standby and first acceptance all need people on site.
- Service network
- 18 service points respond locally, with a handling plan within 48 working hours for standard tickets. More reliable than a vendor in a tier-one city serving a plant in a tier-three one.
- Knowledge transfer
- Bring the administrator up during those visits (configuration, accounts, basic troubleshooting), which is what makes later self-maintenance realistic.
Baichuan is on site at four points: survey, trial-running support, cutover standby and acceptance.
06
How should acceptance criteria be written to avoid disputes?
Three categories are quantified: functions demonstrated item by item against the module list, performance expressed as specific values (scan response, alarm timeliness, concurrent stations), and data verified by sampling with a stated tolerance.
- Functional items
- Each function on the module list is demonstrated during acceptance, so whether it exists is obvious.
- Performance items
- Scan response, error-proofing alarm timeliness, query speed and concurrent station count, quantified as specific values.
- Data items
- Report data is reconciled against physical samples (output counts, stock counts) with the tolerance written into the acceptance criteria. The finer the criteria, the smoother the final payment.
Baichuan accepts item by item against the module list, with performance quantified as values.
07
How are MES projects usually paid for?
By milestone, each tied to an acceptance outcome, with retention covering the warranty period. The line-item quote is attached to the contract and changes are added or removed against it.
- Tied to milestones
- Each payment node corresponds to an acceptable deliverable, avoiding the situation where payment is made and attention moves elsewhere.
- Retention
- Retention covers the warranty period (commonly one year), which gives the response commitment teeth.
- Line-item quote
- A quote listing modules, implementation scope and acceptance criteria is attached, and changes are added or removed against the list. There is no industry standard for payment terms, but milestone payment is fair to both sides.
Baichuan attaches a line-item quote to the contract, with changes adjusted against the list.
08
How is custom development kept under control?
Configure before developing, keep custom items in a signed schedule, and put the code under version control. A controlled customisation ratio is the precondition of the productised route.
- Configure before customising
- If productised configuration can meet the requirement, do not develop. Custom code is a long-term maintenance burden.
- Schedule management
- Custom items go into a separate schedule (requirement, quote, duration, acceptance criteria), signed by both sides.
- Version control
- Custom code enters version control and is assessed at upgrade time. A productised MES such as Baichuan's is configuration-led with custom work as the exception, keeping the ratio controlled.
09
How fast is after-sales response?
Tickets are prioritised with line-stop issues first. Remote response is immediate, regional engineers reach the site within 24 hours, and standard tickets receive a handling plan within 48 working hours.
- Tiered response
- Line-stop emergencies are handled first, general issues queue by ticket, and progress is trackable.
- Local service
- Service points are spread across regions, so on-site support does not require crossing provinces.
- Warranty
- Maintenance is free during warranty and continues under a maintenance agreement afterwards. Ask whether there is somebody local before choosing a vendor; that is more concrete than a feature list.
Baichuan responds remotely immediately, with regional engineers on site within 24 hours.
10
How much is annual maintenance?
It covers upgrades, bug fixes and remote and phone support, excluding new function development. It is priced by module with nothing charged for what is not deployed, and multi-year agreements are negotiable.
- What is covered
- System upgrades, bug fixes, remote support and phone consultation, generally excluding new function development.
- Renewal negotiation
- Multi-year agreements can be discounted. The system keeps running without maintenance, but upgrades and response are no longer guaranteed.
- Cost perspective
- Maintenance is insurance, and small next to the cost of one line stop. Where modules were bought in stages, maintenance follows the same modules.
Baichuan prices maintenance by module, with nothing charged for modules not deployed.
11
Will an upgrade overwrite our customisations?
In a productised MES, customisations run through extension points, so standard module upgrades do not touch custom code. Upgrades are regression-tested in a test environment before production.
- Architectural separation
- Customisations in a productised MES use extension points, so standard module upgrades do not touch custom code.
- Upgrade process
- Upgrade and verify in a test environment first, including regression of custom functions, then move to production. Upgrading production directly is a serious mistake.
- Contractual responsibility
- Write the upgrade support obligation and compatibility commitment into the maintenance agreement, so there is recourse if something is overwritten.
Baichuan runs customisation regression in a test environment before upgrading.
12
How is system data secured?
On-premise deployment keeps data inside the plant, accounts are authorised by level with complete operation logs, and backups are taken regularly with at least one restore drill. An unverified backup is not a backup.
- Deployment
- On-premise deployment keeps data inside the plant, and delivery includes the schema documentation, so the vendor cannot lock the customer in.
- Permissions and audit
- Accounts are authorised by level with configurable role permissions, and login and operation logs are complete, so changes are attributable.
- Backup and restore
- A regular backup policy plus one restore drill. A backup that has never been restored is not a backup.
Baichuan supports on-premise deployment and performs scheduled backup and restore drills.
13
If the network drops or the server goes down, does the line stop?
Define the degraded-mode procedure in advance (which steps are recorded on paper, how they are backfilled), and consider dual-server failover for critical lines. Network drops, UPS and server-room conditions are infrastructure to confirm before implementation.
- Degraded-mode plan
- Agree what is done by hand and how it is backfilled when the network or server is down, with a simple, reliable switchover.
- Optional high availability
- Dual-server failover and regular database backups for critical lines, with investment sized against the cost of a line stop.
- Implementation prerequisites
- Network drops, UPS and server-room conditions are confirmed before implementation. Many cases of system instability are really infrastructure debt.
Baichuan documents the degraded-mode procedure for network and server outages in the go-live plan.
14
Where do MES projects most often fail?
Three places: no plant-manager-level sponsor, installing a system without reshaping the process, and going live with BOMs and barcode rules already broken.
- Sponsorship
- The plant manager leads, milestones carry rewards and penalties, and the shop manager drives execution. MES without management sponsorship goes live and then sits idle.
- Process reshaping
- MES fixes a process into a system, so change the process first. Forcing old habits into a new system leaves both sides awkward.
- Data quality
- Going live with broken BOMs, material master and barcode rules only digitises the chaos. Data cleansing beforehand is not optional.
Baichuan puts process analysis first in implementation: change the process, then install the system.
15
What if nobody uses it after go-live?
Three measures: reduce the number of steps to the minimum, enforce hard constraints on key actions (no scan, no loading; no inspection, no stocking), and put usage into team leader reviews.
- Experience matters
- Minimum steps, fast response and a dashboard that gives team leaders real value. A system that is pleasant to use gets used.
- Hard constraints
- Key actions are locked by the system (no scan, no loading; no inspection, no stocking), so there is no way around them.
- Audit mechanism
- Spot-check physical versus system consistency and include usage in team leader appraisal. Auditing works better than exhortation.
Baichuan makes key actions system checkpoints: no scan, no loading; no inspection, no stocking.
16
How is return on investment assessed?
Wrong-material rework cost multiplied by annual occurrences, labour saved on statistics and traceability, and the order opportunities that audits and reliable delivery bring. Three accounts calculated together.
- Wrong-material account
- Cost per wrong-material rework (labour, material, delivery impact) multiplied by annual occurrences; error-proofing compresses this sharply.
- Labour account
- Clerical hours spent on statistics, reports and digging through records; automating them is a net saving.
- Intangible account
- Order opportunities from passing audits, customer confidence and stable delivery. Hard to quantify, but real.
Baichuan suggests proving one stage's numbers in a pilot before discussing expansion.
17
Can the system still be used after the maintenance period ends?
Yes. An on-premise system belongs to the customer after delivery and cannot be switched off remotely. But system and database upgrades and new equipment integration lose support, and reinstating maintenance may carry a fee.
- Right of use
- After acceptance the system belongs to the customer, and on-premise data sits on the customer's server, so it cannot be switched off.
- Risk of lapsing
- When a Windows or database upgrade or a new equipment integration appears, there is no support, and reinstating maintenance may cost extra.
- Sensible policy
- Once operation is stable, lapsing can be evaluated, but critical systems are better kept under maintenance. One year of fees is small next to a production stoppage.
Baichuan's on-premise deployment cannot be switched off remotely.
18
How is data migration done, and can old system data be brought over?
Master data must migrate, opening stock is calibrated by stocktake, and historical detail is migrated as needed. Clean and map the codes first, then sample-verify after migration before cutting over.
- Migration tiers
- Master data (material master, BOM, suppliers) must migrate; opening stock is imported and calibrated by stocktake; historical detail as required.
- Cleansing
- Old data rarely uses consistent codes, so run a cleansing and mapping pass first. Migration projects are usually delayed by data cleansing.
- Parallel verification
- Sample-verify after migration (quantity, lot, value) and cut over only after it passes. Baichuan supports Excel import of material master and stocktake sheets, so the path exists.
19
What qualifications should a vendor be judged on?
Three: whether a model factory can be visited to see the system running, the degree of productisation and the pace of iteration, and whether service point coverage and response commitments are written into the contract.
- Case verification
- Ask for a visitable model factory (Baichuan has six open by appointment) and see the system running rather than a slide deck.
- Product maturity
- How productised the 15 modules are and how frequently versions iterate. The cost difference between productised and project-based delivery is large.
- Service delivery
- Service point coverage, response commitments (handling plan within 48 working hours) and warranty terms. Pre-contract service promises belong in the contract text.
20
Can we trial or run a pilot first?
Yes. Pick one stage with a clear pain point (error-proofing or traceability) and run a 4–8 week pilot, accepted on wrong-material incidents and traceability completeness, then expand modules.
- Choosing the pilot
- Choose a stage with a clear pain point and visible results; error-proofing and traceability are the usual first choices, with a short cycle, small investment and quantifiable outcome.
- Replication after acceptance
- Agree acceptance criteria in advance (wrong-material incidents, traceability completeness, adoption rate) and expand by module once they are met.
- Controlled investment
- Single stages start around CNY 100k and show results in 4–8 weeks, so the pilot is itself the best feasibility test. Many of Baichuan's plant-wide customers grew out of a single-stage pilot.