| Choosing a System | Needs Analysis | Software Capabilities | Why Minerva? |
- Foundational
- Advanced
- Leading-edge
Needs are not a list. They are a ladder you climb.
| Companies usually buy the upper layer and live in the lower one. That is where things break. |
The wish lists brought to a software selection meeting are almost always the same: "we want business intelligence, we want mobile, we want artificial intelligence". Yet if that same company has an 8% stock count variance, takes three weeks to close its account reconciliations, and its sales team looks up prices in Excel, artificial intelligence will add nothing — it will simply turn bad data into bad decisions faster.
This is why needs have to be read in three layers. Each layer asks its own question:
Foundational needs — "What happened?"
Recording, legal compliance, accuracy. Holding a complete picture of what the business has actually done, in one place. This layer is not a competitive advantage; it is your licence to operate.
Advanced needs — "How is it running?"
Process, control, planning, integration. Work flowing end to end; information not stopping at departmental borders. Efficiency and margin are won here.
Leading-edge needs — "What is coming, and what should I do?"
Decision support, autonomy, artificial intelligence, resilience, speed of adaptation. This is where the company's future room to manoeuvre is decided.
The most common mistake
Level 3 promises (analytics, artificial intelligence, automation) are what influence the purchase decision most; yet most projects fail to produce value because of poor data quality at Level 1 and a lack of process discipline at Level 2. Industry research commonly reports that between 55% and 75% of ERP projects fail to meet their objectives — and the leading causes are not the software itself, but data migration, process misalignment and change management.
Foundational needs: recording, compliance and one version of the truth
| The goal here is not intelligence, it is reliability. If this layer does not work, everything above it is guesswork. |
1.1 Recorded once, and only once
Every transaction captured once, at its source, by the right person. The same customer opened three times under three different spellings, or stock items carrying one code in the warehouse and another in accounting, undermines every layer built on top.
- Master data: customer/supplier (business partner), product, service, price list, unit, warehouse, chart of accounts, cost centre — one definition, one code, no duplicates.
- Transaction data: quotation, order, delivery note, invoice, collection, payment, stock movements, production entries.
- Document chain: unbroken traceability from quotation to order, order to shipment, shipment to invoice and collection.
1.2 Legal and fiscal compliance
In Türkiye this is not a matter for negotiation, and it changes constantly. Keeping up with those changes on time is not a "feature"; it is a matter of business continuity.
- e-Invoice, e-Archive, e-Delivery Note, e-Self-Employment Receipt, e-Producer Receipt and e-Ledger processes, plus integrator connections
- VAT, withholding tax, income tax deductions and special consumption tax; import/export transactions and foreign-currency documents
- Reporting under Turkish tax law and TFRS/IFRS; periodic obligations such as inflation accounting
- Payroll, social security filings, employment agency notifications, incentive calculations, severance and notice liabilities
- Classification of personal data, access restrictions, retention and deletion policies under Turkish data protection law (KVKK)
- Standard trial balances, ledgers and statements for auditors and accountants
1.3 A working accounting and finance core
- General ledger, automatic posting (no journal entries typed by hand), cost centres
- Account tracking, ageing, reconciliation, risk and credit limit control
- Bank, cheques and notes, loans, card payments, cash; cash flow visible in real time
- Receivables and payables ageing and collection follow-up; overdue receivables surfacing without anyone having to ask
1.4 Stock, warehouse and purchasing fundamentals
- Real-time stock balances, distinguishing reserved, in transit, quarantined and consignment goods
- Barcode and handheld terminal movements, stock counts and variance handling
- The flow from purchase requisition to order and from goods receipt to invoice, with three-way matching
- Lot/batch, serial number and expiry date traceability (mandatory in food, pharmaceuticals, automotive and chemicals)
1.5 Authorisation, audit trail and continuity
- Role-based authorisation: who sees what, who changes what, who approves what
- Change history (audit trail): who altered a record, and when
- Backups and restore testing; not merely taking backups, but proving you can come back from them
- Keeping the system running: availability commitments, disaster scenarios, data centre security
The common assumption
"Our basics are already in place; we are looking at the next level up."
What it looks like in practice
If month-end close takes longer than 10 days, if sales and accounting report different numbers, if stock count variances are a recurring topic and critical reports are stitched together by hand in Excel, then the foundational layer is not in place. These symptoms are the strongest indicator that an investment in the upper layers will not pay back.
What happens if this layer is missing?
- Management runs the business by looking backwards; every intervention arrives late.
- Every report starts with an argument: "where did this number come from?"
- Institutional memory sits with individuals; one resignation becomes a loss of knowledge.
- The company struggles in audits, bank lending, investor reviews and acquisition processes.
Advanced needs: process, control and integration
| Recording is not enough; the work has to flow. This is the layer where margin, speed and customer satisfaction are won. |
2.1 End-to-end processes
Companies do not run departments, they run processes. The maturity of a software system is measured by whether these chains run uninterrupted from start to finish:
| Process | Chain | Symptom when it breaks |
|---|---|---|
| Order to cash | Quotation → order → allocation → shipment → invoice → collection | Backlog of unshipped orders, delayed invoicing, lengthening receivable terms |
| Purchase to pay | Requisition → approval → order → goods receipt → invoice matching → payment | Unapproved purchases, price discrepancies, supplier reconciliation workload |
| Plan to product | Demand forecast → master production plan → material requirements → work order → cost | Line stoppages, emergency purchasing, delivery dates that are never met |
| Request to resolution | Call → work order → field/service → spare parts → invoicing | Calls that never close, warranty leakage, dissatisfied customers |
| Hire to exit | Request → candidate → contract → payroll → performance → departure | Manual payroll corrections, disputes over leave and overtime |
| Record to report | Transaction → posting → reconciliation → close → consolidated report | Month-end dragging on, a flood of correcting entries |
2.2 Planning and foresight
- Demand planning: forecasts fed by seasonality, campaigns and contracts
- Material requirements planning (MRP): bills of material, lead times, minimum stock, batch sizes
- Capacity and scheduling: machine, tooling, shift and staffing constraints calculated together
- Supply chain planning: multi-warehouse distribution, transfers, import lead times, alternative suppliers
- Cash planning: forward cash flow derived from order and purchase commitments
2.3 Cost, profitability and budgeting
What most companies do not know is not how much they sell, but which business actually makes them money.
- Standard and actual costing; variance analysis (material, labour, overhead)
- Profitability broken down by product, customer, channel, region, project and order
- Budget preparation, revision, budget-versus-actual comparison and spend control
- Investment and project cost tracking; progress payment management
2.4 Workflow, approval and rule management
- Approval hierarchies that vary by amount, category and organisational level
- Exception handling: anything outside the rules caught automatically and routed to the right person
- Notifications and alerts: critical stock, overdue receivables, late orders, expiring contracts
- Task assignment and follow-up; seeing whose desk the work is waiting on
2.5 The customer and sales side
- CRM: lead–opportunity–quotation pipeline, sales forecasting, visit and activity tracking
- Pricing and discount discipline; contract prices, campaigns, commission and bonus calculation
- After sales: warranty, service, maintenance contracts, spare parts, field team management
- Channel management: dealers, franchises, stores, marketplaces and e-commerce
2.6 Integration and the outside world
A modern company is only as fast as the systems it connects to.
- Bank integration, virtual POS, payment providers, credit and factoring
- E-commerce platforms, marketplaces, courier and logistics companies
- EDI connections to retail chains and OEM customers
- The production floor: PLC/SCADA, terminals, scales, barcode and labelling systems
- Government systems: tax authority, e-government, customs and trade registry
- Open API: the company's own in-house applications able to connect to the system
2.7 Portals and self-service
- B2B portal: dealers and customers enter their own orders and see their balances and shipments
- Supplier portal: order confirmation, shipment notification, invoice status
- B2E (employee portal): leave, advances, expenses, payslips, requests and approvals
- Self-service permanently reduces the load on the call centre and the back office
2.8 Multiple companies, countries and currencies
- Intercompany trading, elimination and consolidated reporting
- Different tax regimes, local regulation and reporting formats
- Multi-currency handling, revaluation and periodic exchange differences
- Multilingual interface and document output
How to measure this layer
The advanced layer is not measured by "number of modules" but by this question: does information travel from where it is created to where it is needed without being carried by hand? Exporting to Excel, copying by hand, "could you email me that file" — every one of these is the symptom of a break in this layer.
Leading-edge needs: decisions, autonomy and resilience
| This layer determines the company's room to manoeuvre tomorrow. But it only truly works if the two layers beneath it are sound. |
3.1 Real-time decision support
- Dashboards running on live transaction data, without waiting for a transfer to a data warehouse
- Role-based indicator sets: within the same company, the CFO, the production manager and the sales representative each need a different screen
- Users able to ask their own questions (ad-hoc queries, user-defined reports) — no queue at the IT department for every question
- Scenario analysis: putting numbers to "what if the exchange rate rises 10%", "what if we lose this customer", "what if we add a shift"
3.2 Artificial intelligence: from assistant to agent
Through 2025–2026 this became the main axis of enterprise software. Gartner expects a significant share of enterprise applications to be integrated with task-specific AI agents, and describes a shift in which ERP moves from a system of record to a system of execution. In practice this appears at three levels:
| Level | What it does | Example |
|---|---|---|
| Assistant (copilot) | Finds what is asked for, summarises it, drafts documents | "Show me this customer's delays over the last 6 months and their open balance" |
| Automation | Runs rule-based work without human involvement | Invoice–order–delivery note matching, collection reminders, document classification |
| Agent | Runs multi-step work towards a goal, returning to a human at exceptions | Detecting critical stock → comparing suppliers → creating a requisition → submitting it for approval |
At all three levels, human approval (human-in-the-loop) must be part of the architecture. Autonomy without approval produces errors in enterprise data that are hard to undo.
The precondition for artificial intelligence to be useful is not technological but structural: a single, consistent data model. An agent working across fragmented systems, trying to reconcile data that does not agree, only accelerates the error. Artificial intelligence is therefore not the cure for a scattered system landscape; it is the return on an integrated one.
3.3 Forecasting and anomaly detection
- Forecasts for demand, cash flow, collection risk and supply delays
- Automatic flagging of unusual transactions, prices or stock movements (error and fraud detection)
- Predictive maintenance: intervening before the breakdown
- Price and discount optimisation; customer churn signals
3.4 Data governance
The least visible but most decisive topic at this level. Data is a corporate asset and has to be managed as one.
- Master data management: one definition, clear ownership, approved changes
- Data quality rules and regular measurement
- Data classification, access policies, retention and disposal
- A data dictionary and shared definitions for indicators — "revenue" must mean the same thing to everyone
3.5 Architectural flexibility: composable and event-driven
- Composability: adding or removing a capability without disturbing the whole
- Event-driven flow: the event "order approved" triggering every related process by itself
- API-first design: the capability behind every screen also being available externally
- Phased go-live: controlled transition instead of a "big bang" that changes everything at once
3.6 A link to the physical world: IoT and the field
- Automatic data capture from machines, meters, sensors and vehicles
- Production efficiency (OEE), energy consumption, cold chain monitoring
- Fleet, vehicle and field team location management
- Instant recording of warehouse and field transactions on mobile and handheld devices
3.7 Sustainability and compliance reporting
- Tracking carbon footprint and resource consumption at product and order level
- Carbon border adjustment mechanism (CBAM) data requirements for exporters
- Supply chain due diligence and supplier compliance documentation
- Occupational health and safety, environmental and quality management system records
3.8 Security, continuity and risk
- Authentication, multi-factor login, privileged access management
- Segregation of duties: the same person unable to both raise an order and make the payment
- Ransomware scenarios; recovery objectives (RPO/RTO) and drills to test them
- Logging, monitoring and an incident response plan
3.9 Speed of adaptation
The last and perhaps most important item at this level: how long the software takes to adapt when the company changes. If the answer to a new legal entity, expansion into a new country, an acquisition, a new business model (subscription, rental, marketplace) or a new legal obligation is "months and a serious budget", then that software has stopped being an asset and become a constraint.
The expectation
"Artificial intelligence will connect our scattered systems and solve our reporting problem."
The reality
Artificial intelligence inherits the quality of the data beneath it. A model working on inconsistent, fragmented data with conflicting definitions will present a wrong answer in convincing language — which is riskier than producing no report at all. The order is this: first a single data model, then process discipline, then autonomy.
The additional load carried by a company operating in Türkiye
| The three layers above are universal. The list for a company operating in Türkiye is a little longer. |
Frequent, short-notice change
Exchange rates, inflation and pricing discipline
Banks, couriers, marketplaces, government
Support and access to knowledge
Which layer is your company in? An 18-question check
| To find out where the need really sits, before you start looking at software. |
Level 1 — Foundational
- How many days does your month-end close take? (Longer than 10 days means there is a problem in the foundational layer.)
- Is your stock count variance a regular topic of conversation?
- Does the same customer or product exist in the system under more than one record?
- Are your critical reports assembled by hand in Excel?
- When did you last successfully test a restore from your backup?
- Can the system tell you who changed what, and when?
Level 2 — Advanced
- Can you see what stage an order is at from a single screen, without asking anyone?
- Are purchase approvals handled in the system, or over email and WhatsApp?
- Which product and which customer earns you how much — does the system tell you?
- Do you set delivery dates based on planning, or on experience?
- Can dealers and customers see their own balances and orders themselves?
- Is there a person or a file moving data between two systems?
Level 3 — Leading-edge
- Can you see today's figures today, or are you looking at yesterday's closed data?
- Can a user produce their own report without asking IT?
- Do critical events (stock-outs, delays, limit breaches) reach you without you going to look for them?
- Is the definition of "revenue" the same across every department?
- How long does it take you to add a new company, country or business model?
- When you discuss using artificial intelligence, do you also discuss which data it will work on?
How to read the results
If three or more items are a problem at Level 1, your priority is not analytics or artificial intelligence; it is a single data model and clean processes. If Level 1 is clean and the breaks are at Level 2, that is where the returns come fastest. If both levels are sound, an investment at Level 3 will genuinely compound.