How to choose HR software (for teams in Turkey)

Start from your own process inventory, not from a product list. The core modules are well known: employee records (Turkish employers are legally required to keep a personnel file for every employee under Labor Law No. 4857, Article 75), leave tracking, payroll, recruiting, attendance/PDKS data, and employee self-service. For each one, answer two questions: how do we do this today, and how many hours does it cost per month? The processes whose honest answer is "Excel, email and memory" are where software pays for itself fastest.

Priorities shift with headcount. At 10-25 employees the core need is records plus leave tracking. Payroll usually runs at an external accountant, so the critical question is not whether a payroll module exists but whether the data that feeds payroll (leave, overtime, attendance) flows out cleanly. Between 25 and 100, self-service and approval flows start to matter: HR's day should not be spent answering "how many leave days do I have left". Above 100, org structure, permissions and reporting become the deciding factors.

The goal is not the product with every module — it is the product that does your three core processes end to end. Do not pay today for modules you will not use, but do check that the needs of two years from now (say, recruiting or performance) exist on the product's roadmap.

Payroll and leave legislation is a pass/fail filter, not a scoring criterion. The question is not "does the product support Turkey" but whether Turkish legislation is native to it: how do SGK processes and filings run? When minimum wage or premium parameters change, who applies the update and how fast? Are the statutory annual-leave tiers (minimums of 14, 20 and 26 days by seniority under Labor Law No. 4857, Article 53) built into the arithmetic? In global products, Turkish payroll often runs through a local partner — not a flaw in itself, but make the contract state clearly who is liable when a filing goes wrong.

KVKK is the second pass/fail set. Employee data is personal data under Turkey's Data Protection Law No. 6698, and records such as health reports count as special categories of personal data subject to stricter processing conditions (KVKK Article 6). Ask: in which country is the data hosted? If it crosses borders, how are the transfer conditions of KVKK Article 9 met? Does the vendor sign a data-processor agreement, and who are its sub-processors?

Do not confuse Turkish support with a translated interface. The real test is whether the support team works in Turkish and can answer a legislation-adjacent question such as "why was this employee's leave balance calculated this way". The language of the documentation and the contract belongs under this heading too.

Evaluate data migration in both directions. On the way in: is there bulk import from Excel/CSV, can leave entitlements earned before the system be entered as opening balances, and how many fields will need manual fixing? On the way out: can you export your data in a standard format? A product that is hard to leave also weakens your hand in every future price negotiation.

Read pricing as 12-month total cost of ownership, not list price. Per-user pricing looks cheap for a small team but grows with every hire. Flat packages are predictable but can charge you for capacity you never use. Ask explicitly about the hidden lines: setup fees, training, add-on modules, annual commitments, user-tier overages, support levels. For integrations, list the tools you already run (accounting, calendar, attendance hardware) and have each one demonstrated — "we have an integration" and a working connection are not the same thing.

AI is no longer a checkbox: it is its own evaluation axis, because agentic HR software has started doing the work rather than merely recording it. The right question is not "does it have AI" but two questions. First, authority: which tasks does the AI take on, and where does a person approve? In well-designed agentic HR the agent screens, prepares and drafts, and publishing and approval stay with a person. Second, data: which employee data can the AI reach, and is your data used to train models? A product without clear written answers to both should count AI as a risk item, not a feature.

Build a shortlist of two or three products from those that pass the eliminating filters (native legislation support, KVKK, budget ceiling). A list longer than five makes every comparison shallow, and a list of one erases your negotiating power.

Do not watch a demo — run a trial. The difference is large. Work with your own data: import 10-20 real employee records, push a real leave request through the approval flow, and simulate one payroll period end to end. Measure three things while you do: how many hours setup took, whether an employee outside HR could log into the portal and request leave without help, and how fast, and how well, support answered your question.

Decide with a scorecard and more than one pair of eyes: HR judges process fit, finance judges total cost, and IT (if you have it) judges security and KVKK. Before signing an annual commitment, read the exit clauses and data-return terms. For most companies it is healthier to start monthly and move to annual once the product has proven itself.