Section 02 - Evaluation Criteria
What capabilities should business software have?
A feature list is the most misleading instrument in a purchase decision. Almost every vendor says "yes" to almost every item. The difference does not lie in whether a capability exists, but in what benefit it delivers, how long that takes, and at what cost.
In this section we have grouped capabilities under six headings, and tied each heading to one question: What does this change in the company? At the end, we set out the measurable equivalent of every expected benefit.

The value of a capability is measured by the difference it makes

Instead of "do you have it?", ask "in how many days, with how many people, at what cost?"

Every screen you see in a demo works. The problem is that the setup, data, integration and consulting effort needed to reach that screen stays invisible. Industry data shows this gap clearly: ERP project failure is commonly reported in the 55–75% range, average implementation takes around 17 months against the roughly 12 months planned, and the main causes of delay are not the capability of the software but data migration, scope creep and change management.

55–75%
Share of ERP projects that fail to meet their objectives (the range commonly reported in Gartner and Panorama Consulting analyses)
~17 months
Average actual implementation time; the planned duration is usually 12 months
30%
The faster financial close Gartner predicts for 2028 through embedded artificial intelligence
40%
Share of enterprise applications Gartner expects to include task-specific AI agents by the end of 2026 (under 5% in 2025)

Functional capabilities

The capabilities that carry the work itself. They are measured not by breadth, but by whether the coverage is unbroken.

1.1 A single data model

Every module using the same master data and the same database. The word "integrated" has two very different meanings here, and the difference is critical:

Connected by interfaces

Separate products talk to each other through transfer routines. Data is delayed, duplicated, contradicted. Every new release means retesting the connections. The question "which one is right?" never goes away.

Integrated by design

Modules share one data model. The moment a sales order is entered, stock allocation, credit limit control, planning and the accounting effect all happen at once. There is no transfer — and therefore no transfer error.

1.2 End-to-end process coverage

Order to cash, purchase to pay, plan to product, request to resolution — every chain completed within one system. The test is simple: if any step of a process leaves the system for Excel or another program, the coverage is incomplete.

1.3 Configurable flexibility — adapting without writing code

Every company's processes are different; but that difference should not require software development every time. In a good system, the following are settings, not development work:

  • Approval hierarchies, workflow steps, rules and limits
  • Screen fields, mandatory entries, default values, user-specific views
  • Document layouts, numbering series, multilingual texts
  • User-defined fields, and the ability to use those fields in reports
  • Reports and analysis dashboards

1.4 Industry depth

80% of a general ERP fits every sector; the 20% that keeps a company standing is specific to its industry. Bills of material and costing in manufacturing, campaigns and stores in retail, contracts and work orders in services, traceability in healthcare. That depth is either already in the product or written again at every customer — and the second option is the most common reason projects run late.

1.5 Traceability and auditability

  • Where each document originates and what it connects to (the document chain)
  • Backward and forward tracing by lot, serial number and batch
  • Change history and approval trail
  • Segregation of duties enforced by the system itself

1.6 Role-based design and proactive information

Users should not have to search for information; information should come to them. A screen that opens according to their role, the tasks waiting on them, the alert that appears when a threshold is crossed. This is the technical answer to what is usually called "user adoption" — the point where projects most often stall.

1.7 Localisation and multi-entity structures

  • Turkish regulation being inside the product, requiring no add-on or local partner
  • Multiple companies, warehouses, currencies and languages
  • Intercompany transactions and consolidation

Technical capabilities

The layer users never see, but will feel every day for five years.

2.1 The age and type of the architecture

The question to ask is not "is it in the cloud?" but "when was it designed, and for what?" The core of many products on the market is 1980s–90s architecture with a modern-looking interface fitted over it. In those products, web, mobile, real-time analytics and artificial intelligence are each an additional layer, and each one creates its own cost and its own delay.

  • Genuine web architecture: runs in the browser, needs no client installation, no remote desktop or terminal server workaround
  • Multi-tenant cloud / SaaS: updates, scaling and security are the vendor's responsibility
  • Deployment options: cloud, SaaS, on-premise or hybrid — and the right to move between them

2.2 Database and data integrity

Consistency in enterprise data is not negotiable. Transaction integrity (ACID), running without locking under hundreds of concurrent users, performance at high data volumes and mature backup and recovery tooling all depend on the quality of the database layer. An enterprise-class database (Oracle, for example) provides proven ground on all of these.

2.3 Performance and scalability

  • Response times holding steady as user numbers and data volumes grow
  • Capacity flexing during peak periods (month-end, campaigns, stock counts)
  • Heavy reports not slowing down day-to-day operations
  • The test: response times for the 10 most frequently used screens on real data — not on demo data

2.4 Integration capability

  • Documented, open APIs with version guarantees
  • Event-based notifications (webhooks) able to trigger external systems
  • Ready-built connections: banks, e-document integrators, e-commerce, marketplaces, couriers, EDI
  • Bulk data transfer tools and repeatable transfer templates
  • Standard interfaces for secure access by AI agents

2.5 Security

  • Role- and data-based authorisation; restrictions down to field level
  • Multi-factor authentication, single sign-on (SSO)
  • Encryption in transit and at rest; retention and deletion compliant with KVKK/GDPR
  • Logging, monitoring and an incident response process; a history of penetration testing
  • Data centre certifications and physical security

2.6 Continuity: backup, recovery, availability

The concrete questions to ask: How often are backups taken? How much data loss is acceptable at most (RPO)? How long does the system take to come back up (RTO)? When was a restore last tested? Is the availability commitment written into the contract?

2.7 Upgradability — the source of the silent cost

The most expensive concept in enterprise software: customisation debt. Every line of code written for one customer creates retesting and rework at every subsequent release. In time the company ends up locked into a version too customised to update; new capabilities — artificial intelligence included — never reach it.

Ask this during evaluation

"The development you will do for us — does it go into the standard product, or does it create a version specific to us?" If the answer is "specific to you", have the cost of future upgrades written into the contract as well.

2.8 Extensibility and observability

  • Fields, rules and reports the company's own team can add (low-code development)
  • Separate development, test and live environments, with version management
  • Visibility of system health, error and performance indicators

Next-generation capabilities

The move from a system that records to a system that executes. This is where today's differences are made.

Gartner's 2026 assessment describes ERP moving beyond its classic definition as a "system of record" towards an event-driven, composable execution model with a layer of intelligence on top. In the same direction, task-specific AI agents in enterprise applications are expected to reach 40% by the end of 2026. Over the five-year horizon of a software decision, this is a decisive break.

01

An embedded AI assistant

Users asking questions in their own words, the system finding and summarising the data, drafting documents and correspondence. The test: the assistant working on the company's own data and within its authorisation limits.
02

Agents and autonomous tasks

Multi-step work carried out towards a goal: matching, reminders, procurement suggestions, exception detection. The non-negotiable condition: human approval at critical steps and a complete audit trail.
03

Real-time embedded analytics

Dashboards running on transaction data, without a separate business intelligence product or an overnight transfer. Having analytics inside the ERP removes both the cost and the delay.
04

Event-driven automation

Events such as "order approved", "stock fell below threshold" or "invoice received" triggering the related processes by themselves. The human role shifts from starting the work to resolving the exception.
05

Composable architecture

Capabilities that can be added or removed without disturbing the whole; phased go-live. The most practical capability for removing the risk of a "big bang" transition.
06

IoT and a link to the physical world

Connecting data from machines, meters, sensors and vehicles to business processes: OEE, energy, cold chain, predictive maintenance.
Red flags in next-generation capabilities
  • The artificial intelligence exists only in the demo video; there is no live customer reference.
  • The AI module is sold separately at a high additional licence fee and cannot be used in core processes.
  • The agent creates transactions with no approval step and no audit trail.
  • Analytics only works through a separate business intelligence product and an overnight data transfer.
  • Nobody can explain clearly where the data goes and which model processes it.

User experience and adoption

Software that goes unused costs more than software never bought.
  • Learning curve: how many hours of training does a new user need before they can do their basic job? This feeds directly into go-live time and into resistance.
  • Screen economics: how many clicks a daily task takes. On work repeated 200 times a day, a difference of 3 clicks means weeks of time over a year.
  • Personalisation: users able to set up their own lists, filters and shortcuts.
  • User-defined reports: not having to raise a job with the vendor for every new question.
  • Mobile and field use: genuinely usable in the field, in the warehouse, on the road — not merely looking "mobile friendly".
  • Accessibility and speed: working on low bandwidth; full operation from the keyboard.
  • Clarity of error messages: messages that tell the user what to do next reduce the support load noticeably.

Commercial and relationship characteristics

The contract and the vendor are themselves a "capability" — and the longest-lived one at that.

5.1 Transparency of total cost of ownership (TCO)

The licence or subscription fee is often less than half of the real cost. When comparing options, all of the following items have to be counted:

Cost itemThe detail most often missed
Licence / subscriptionUser types, module-based extra charges, annual increase rate
Implementation and consultingEstimated number of days, pricing of out-of-scope work, travel and accommodation
Data migrationThe data cleansing effort usually stays with the customer
Integration and middlewareA separate product licence + ongoing maintenance
Business intelligence / reporting toolDoes it need a separate licence and separate expertise?
InfrastructureServers, database licences, backup, system administration
Training and change managementContinuing training needs after the first year
Maintenance, support, version upgradesReworking customisations at every release
Internal resourcesThe time your own employees give to the project — the largest and most invisible item

Compare on a five-year total; a difference in a single year's licence fee is usually misleading as a basis for the decision.

5.2 Data ownership and the right to exit

The clauses to look for in the contract: that the data belongs to you, that you can take all of it in a standard format whenever you wish, that it will be handed over to you when the contract ends and retained for a reasonable period. A system with no guaranteed exit is not a supplier relationship; it is a dependency.

5.3 The support model

  • Does the person supporting you understand the business, or only open a ticket?
  • Are response and resolution times defined in the contract? (They are two different commitments.)
  • Is there enhanced support during critical periods (month-end, stock counts, closing)?
  • How much distance is there between the team that builds the product and the team that supports it?

5.4 Vendor continuity and roadmap

  • How many years the company has been doing this, and how many live customers run the product
  • A published roadmap for the product's next 24 months
  • How many releases a year, and how updates are applied
  • Reference calls: being able to speak to a customer of similar size chosen by you, not by the vendor

The benefits a company should get from its software

Every benefit has three parts: the mechanism that creates it, the indicator that measures it, and the condition under which it actually happens.

6.1 Operational benefits

BenefitMechanismMeasurement indicator
Shorter order cycle timeManual transfers and approval waits disappearingAverage days from order → shipment
Better delivery performancePlanning, stock allocation and capacity calculated togetherOn time in full (OTIF) %
Lower stock investmentReal-time balances + demand forecasting + safety stock disciplineStock turnover, dead stock ratio, days of inventory
Fewer errors and less wasteBarcoding, system controls, no need for double entryStock count variance %, number of returns and misdeliveries
Purchasing savingsRequisition consolidation, contract price control, approval disciplineOff-contract purchasing ratio, average unit cost
Higher production efficiencyScheduling, downtime tracking, material readinessOEE, plan adherence, unit production cost
Service qualityIntegration of the call–work order–parts chainFirst-time fix rate, average resolution time

6.2 Financial benefits

BenefitMechanismMeasurement indicator
Shorter closeAutomatic posting, reconciliation tools, embedded artificial intelligenceDays to complete month-end close
Faster collection of receivablesAgeing, automatic reminders, limit and risk controlDSO (days sales outstanding)
Cash visibilityForward cash flow derived from orders and commitmentsVariance in the 13-week cash forecast
Knowing true profitabilityCost allocation by product, customer and channelNumber and share of loss-making products and customers
Lower cost of financingAccurate, timely financial statements; credible reporting to banksCost of credit, collateral required
Lower IT costConsolidating many point solutions, middleware and BI tools onto one platformAnnual IT spend per user

6.3 Management and decision-making benefits

  • A single source of truth: meetings starting with the decision at hand, not with an argument about the numbers.
  • Speed of intervention: seeing a problem on the day it arises, not at month-end. This is the most underrated benefit and the one with the highest return.
  • Management capacity: running a bigger business with the same team — headcount growing more slowly than revenue as you scale.
  • Institutional memory: knowledge held in the system rather than in individuals, so handovers and resignations do not disrupt continuity.
  • Auditability: being ready for tax audits, independent audits, bank and investor reviews.
  • Company value: orderly data and transparent processes have a direct effect on valuation in acquisition and partnership discussions.

6.4 People and culture benefits

  • Less repetitive manual work; qualified staff analysing rather than copying data
  • Breaking the cycle of blame between departments — everyone looks at the same data
  • New employees getting up to speed quickly; work tied to a process rather than to a person
  • Employee self-service reducing the routine request load on HR and accounting

6.5 Strategic benefits

  • Capacity to scale: a new company, warehouse, country or channel opened within weeks
  • Trying new business models: subscription, rental, marketplace or service sales having an equivalent in the system
  • Supply chain resilience: alternative suppliers, multiple sourcing and scenario readiness
  • Compliance readiness: being able to meet sustainability and reporting obligations with real data
  • Being ready for artificial intelligence: clean, integrated data is the precondition for every future automation investment
The only honest way to measure the benefit

Before the project starts, write down today's value for 8–10 indicators (days to close, DSO, stock turnover, OTIF, stock count variance, order cycle time, number of open service calls, IT cost per user). Measure the same list at 6 and 12 months after go-live. That list is both the clearest contract between you and your vendor and the real report card for the investment.

20 questions to ask every vendor

Write the answers side by side. The difference shows up in these answers, not in feature lists.
  1. In what year, and on what architecture, was the core of your product written?
  2. Do all modules use the same database, or is there a transfer between them?
  3. How many live customers do you have at our size and in our industry? May we speak to one — chosen by us?
  4. What is your estimate for time to go live, and what is that estimate based on?
  5. Who performs the data migration, with which tools, and in how many days?
  6. Will the development done for us go into the standard product, or create a version specific to us?
  7. How often and how are version updates applied? How much work does that create on our side?
  8. Can we change approval flows, screen fields and reports ourselves?
  9. What level of knowledge does it take to produce a user-defined report?
  10. Is Turkish regulation (e-documents, payroll, inflation accounting) inside the product, or an add-on?
  11. How many days does it take for a change in legislation to reach the system? Is there an extra charge?
  12. At which live customer, and in which process, are your AI capabilities running today?
  13. Does artificial intelligence carry an additional licence? Where is our data processed?
  14. Is analytics inside the product, or does it require a separate business intelligence tool?
  15. Are your APIs documented? Will our integrations break when the version changes?
  16. What are your availability commitment, RPO and RTO values? Are they in the contract?
  17. How close is the support team to the team that builds the product? Do they solve the problem or open a ticket?
  18. Can you give us the five-year total cost in writing, with every item included?
  19. If we decide to leave, in what format and within what period do we get our data?
  20. What is on the product roadmap for the next 24 months?
So where does Minerva stand against these criteria?
The section where we answer every one of the questions above with evidence rather than claims.
Contact Us