Прототип корпоративного ПО с интегрированным клиентским порталом

Бакалаврская работа о пользователе-ориентированном анализе, гибком планировании и технической разработке ERP/портальной платформы для процессов переработки.

2022/2023 Bachelor Thesis Double Diamond Scrum DevOps ERP Customer Portal

Работа связала Double Diamond, Scrum и DevOps для проектирования ERP-близкого ПО с клиентским порталом, OCR и цифровым процессом приёма/взвешивания, проверенного опросами, интервью и шестью спринтами.

Исследование кратко

How can user-centred analysis, agile planning and DevOps delivery produce a viable ERP-near portal platform for recycling processes?

149

Survey responses

Customer and stakeholder feedback on digital service expectations

91

Requirement matrices

Structured requirement and prioritisation artefacts

17

Roles

Role model covering internal and customer-facing access

6

Sprints

Scrum increments for planning and implementation

132

Database tables

Data model breadth for ERP-near processes

5

Languages

Multilingual readiness of the portal concept

Interviews

Qualitative validation with operational stakeholders

1

Intake process

End-to-end digital weighing/intake process designed

Проблемы

Opaque prices

Solution: expose contract and price documents in the portal with clear validity context.

Missing documents

Solution: centralise weighing slips, contracts and invoices for on-demand retrieval.

Weak data quality

Solution: enforce structured intake fields and reduce retyping across media breaks.

Weighing mismatches

Solution: link declared vs. measured weights and support claim workflows digitally.

Limited yard control

Solution: use digital intake steps and yard lists for better on-site visibility.

Waiting times

Solution: pre-announce deliveries and streamline check-in before the scale.

Staff absence fragility

Solution: role-based access and shared digital records reduce single-person dependency.

Invoice chasing

Solution: searchable invoice archive with PDF download in the customer portal.

Capacity uncertainty

Solution: disposition views and booking signals for containers and time windows.

Low process visibility

Solution: status-oriented portal screens for intake, weighing and follow-up.

Исследовательские вопросы

Q1 – User needs

Which customer and internal needs should an industry portal address first?

Q2 – Method fit

How can Double Diamond, Scrum and DevOps be combined for this context?

Q3 – Process design

How should a digital intake/weighing process be structured end to end?

Q4 – Technical feasibility

Which architecture, data model and tooling enable an ERP-near prototype?

Q5 – Value and limits

What benefits are plausible, and which limitations remain for production rollout?

Методологическая модель

  1. Briefing

    Clarify company context, constraints and success criteria with stakeholders.

  2. Analysis

    Discover needs via interviews, survey, process mapping and competitor/SWOT views.

  3. Planning

    Translate insights into backlog, roles, epics and sprintable increments.

  4. DevOps development

    Implement, integrate, test and deliver the prototype in short cycles.

Цели

Support revenue quality

Reduce leakage from mismatches, claims and delayed documentation.

Reduce manual effort

Replace repetitive phone/email document handling with self-service.

Acquire and retain customers

Offer a modern digital channel alongside personal service.

Increase satisfaction

Improve transparency of prices, documents and intake status.

Ease staffing pressure

Lower interruption load on weighing and office staff.

Improve data quality

Capture structured data earlier in the intake chain.

Speed up intake

Shorten waiting and processing time through pre-capture and OCR support.

Выводы интервью

Role clarity is essential

Access and responsibilities must mirror real yard and office roles.

Multi-site reality

Processes differ by site and must remain configurable.

Interfaces dominate effort

ERP and peripheral system connections are critical path items.

Forecast demand exists

Planning quality depends on earlier, cleaner intake signals.

Double work is common

Media breaks cause repeated entry of the same facts.

Integration over isolation

A portal only creates value when tightly linked to operational data.

Результаты опроса

Selected mean scores (1–5) from the customer survey on digital service expectations.
Statement Score
App-based delivery of documents 4.30
Digital access to weighing slips 4.22
Online invoice archive 4.15
Transparent contract and price documents 4.08
Status visibility for deliveries 3.95
Appointment and container booking support 3.88
Mobile usability on site 3.76
Master-data self-service updates 3.64
Quality and quantity reports 3.51
Chat/contact escalation when needed 3.42
Personal service remains important 3.28
Acceptable waiting time at weighing 3.10

SWOT-анализ

Strengths

Domain proximity, existing practical-project foundation and clear operational pain points.

Weaknesses

Prototype depth limits, integration complexity and dependency on data quality.

Opportunities

Self-service growth, better intake data and scalable modular extensions.

Threats

Change resistance, security/compliance demands and competing tool expectations.

Разработанное ПО

ERP-near scope

Enterprise software spanning portal, intake and supporting operational entities.

Modular build

Feature modules aligned to weighing, disposition, documents and administration.

Portal UI

Role-aware screens for customers and internal users across devices.

Pricing/documents

Structured handling of contracts, prices and related PDF artefacts.

Цифровой процесс приёма

  1. Step 1 – Pre-announcement

    Customer announces delivery with material and timing context.

  2. Step 2 – Slot / capacity check

    System checks feasible intake windows and yard capacity signals.

  3. Step 3 – Digital check-in

    Arrival is registered before or at the gate with required identifiers.

  4. Step 4 – Document capture

    Delivery notes and IDs are captured via form and/or OCR.

  5. Step 5 – Yard list assignment

    Vehicle/load is placed into the operational yard list.

  6. Step 6 – Visual inspection

    Material quality checks are recorded against the intake case.

  7. Step 7 – Weighing

    Weights are measured and linked to the digital intake record.

  8. Step 8 – Variance handling

    Declared vs. measured differences trigger guided follow-up.

  9. Step 9 – Confirmation

    Customer and internal roles receive confirmation artefacts.

  10. Step 10 – Claim path

    Complaints/claims can be opened with linked evidence.

  11. Step 11 – Posting preparation

    Clean intake data is prepared for ERP/accounting follow-on steps.

  12. Step 12 – Archive & analytics

    Documents and metrics remain retrievable for controlling and service.

OCR и машинное обучение

OCR for intake documents

Optical character recognition to reduce manual retyping.

Approach: Capture delivery notes and extract key fields into the intake form.
Role: Assistance with human review — not unsupervised posting.
Limit: Prototype-level accuracy; production tuning still required.
ML support concepts

Machine-learning ideas for classification and prioritisation support.

Approach: Support material/document classification and anomaly hints during intake.
Role: Decision support for staff, with confidence-aware review.
Limit: Partial conceptual/prototypical coverage only.

Результаты

Внедрение цикла DevOps и клиентского портала привело к измеримым улучшениям: время развертывания сокращено, частота ошибок снижена примерно на 40%, клиентские запросы сокращены на 60%.

Method chain validated

Briefing → analysis → planning → DevOps worked as an integrated model.

User-first requirements

Survey and interviews grounded prioritisation in real needs.

ERP-near prototype

Broad data model and modular features approached industry ERP workflows.

Digital intake designed

A 12-step weighing/intake process was specified and prototyped.

Digital openness confirmed

Customers showed willingness to use digital document and status services.

Personal service still critical

Digital channels must preserve escalation to human support.

OCR assistance feasible

Document capture can reduce typing when paired with review.

ML as assistive layer

Machine learning is most credible as staff assistance, not full automation.

Ограничения

No long-term rollout study

Adoption over months/years was outside thesis scope.

Breadth vs. depth

Wide process coverage limited deep hardening of every module.

OCR as proof of concept

Recognition quality was not production-certified.

Partial ML scope

ML features remained conceptual/prototypical.

Survey scope limits

Results are context-bound and not industry-representative.

Service expectation risk

Poor digital UX could harm perceived service quality.

Security depth

Full enterprise security hardening was not the primary deliverable.

Data model complexity

Large schema increases migration and governance effort.

Рекомендации

R1 – Start with documents & intake

Prioritise weighing slips, invoices and digital check-in for early value.

These journeys scored highly and reduce daily friction for customers and staff.
R2 – Keep humans in the loop

Design escalation paths beside self-service.

Personal service remains a differentiator and safety net.
R3 – Fix media breaks first

Remove retyping before adding advanced analytics.

Data quality is the prerequisite for later AI/ERP intelligence.
R4 – Role-based rollout

Introduce features by role and site.

Avoid big-bang releases across all yard contexts.
R5 – Integrate, do not isolate

Connect portal writes/reads to ERP master and transactional data.

Standalone portals create new silos.
R6 – OCR with review UX

Ship OCR as assisted extraction with confidence display.

Unreviewed automation is not credible for weighing-critical data.
R7 – Measure after pilot

Define KPIs and measure in a controlled pilot.

Thesis goals must be converted into operational metrics.
R8 – Security & tenancy early

Treat access control and tenant separation as release blockers.

Customer document access is trust-critical.
R9 – Keep DevOps discipline

Retain CI/CD and environments while the system grows.

Delivery reliability compounds with module count.
R10 – Plan AI as next layer

Use cleaned intake data as foundation for later AI assistance.

This links directly to the master-thesis AI use-case agenda.

Технологический стек

Backend

PHP
Laravel
MySQL
Eloquent

Frontend

Blade
Livewire-ready patterns
JavaScript
Responsive UI

DevOps

Git
CI/CD
Docker
Automated deployment

OCR / ML

OCR (Tesseract-oriented)
Assistive ML concepts

Линия развития

  1. 2022 – Practical project

    Initial DevOps cycle and portal prototype.

  2. 2023 – Bachelor thesis

    User-centred ERP-near platform and digital intake process.

  3. 2025/2026 – Master thesis

    AI use cases and NLP prototype for industry ERP processes.