
Updated 16 October 2026. A practical Zama Engineering Systems guide.

Biometric attendance and payroll integration should turn recorded attendance into reviewed payroll inputs. It should not turn every clock-in directly into a salary decision. Devices record events, attendance software interprets them against schedules, supervisors investigate exceptions, and authorised payroll staff decide what enters the payroll process. A dependable project makes each of those steps visible and assigns responsibility for it.
For a Kenyan employer replacing spreadsheets or connecting existing systems, the first question is not whether two products both mention integration. The question is whether the exact attendance and payroll products provide accessible, supported interfaces for the required information. Those interfaces must then be mapped, tested, reconciled, and operated by named people. A controlled export can be appropriate; a direct connector can also be appropriate when its prerequisites are established.
ZES can assess the devices, software, available interfaces, and reporting workflow before proposing a defined integration scope. No particular product pairing should be treated as supported until access and compatibility have been demonstrated. This guide explains how to prepare that assessment and evaluate the result, without inventing payroll formulas or treating a biometric record as an automatic instruction to pay or deduct money.
Define the Boundary Between Attendance and Payroll
An attendance record describes an interpreted work period supported by events and approved exceptions. Payroll applies the organisation’s authorised pay rules and other relevant inputs. These responsibilities overlap, but they are not interchangeable. A person clocking out late does not automatically establish approved overtime, and a missing clock-out does not automatically establish unpaid absence. The workflow must allow those circumstances to be reviewed.
Write down what the attendance system is expected to supply. It might provide approved days, hours, absence categories, or other agreed units. Payroll should confirm what each field means and how it will be used. Do not assume that a field called overtime has the same meaning in both systems. One product may represent observed time, while another expects an already authorised payroll category.
The ZES biometric attendance system guide explains the earlier journey from enrolment to dependable records. Integration begins from that foundation. Connecting unreliable identities or unexplained attendance calculations to payroll does not remove the underlying problem; it moves the problem into a more consequential process.
Discover the Exact Products and Interfaces
Record the attendance terminal models, software product, version, deployment arrangement, and responsible supplier. Do the same for payroll and any HR system that owns employee information. Ask the vendors what supported export, import, or API options are available for these exact products. A feature offered by a different edition or version is not evidence that your installation has it.
Confirm access before committing to a connector. The assessment may need documentation, a test environment, a sample export, permission from the system owner, or vendor assistance. If an interface requires an additional licence or service, include that dependency in the scope. An integration proposal should state which prerequisites are confirmed and which still need resolution.
Avoid promising direct database integration simply because a database exists. Access rights, vendor support, data meaning, and change behaviour all need assessment. The project should use an authorised, supported route appropriate to the systems. When that route is unavailable, a controlled file exchange may be more realistic than an unsupported workaround that nobody can maintain.
Identify Who Owns Employee Information
Choose an authoritative source for employee identifiers and employment status. This may be the HR or payroll system, depending on the organisation. The important point is that the responsibility is explicit. If separate teams create people independently in attendance and payroll, duplicate identities and unmatched records can arise even when names appear correct.
Define who can create an employee, change an identifier, transfer a person, and deactivate a record. Explain how those changes reach the other systems and who checks that they arrived. A new employee should not be left clocking under a temporary shared profile because nobody agreed the registration process. A departed employee’s history should not be erased merely to stop future use.
Use the approved employee identifier as the mapping key wherever the design supports it. Names remain useful for review, but they are not reliable unique keys: two people can share a name, and names can change. Preserve leading zeros and other meaningful identifier formatting. A spreadsheet that silently turns an identifier into a different value can create an integration error before any file is imported.
Prepare a Field Mapping That Payroll Can Read
A mapping document should show the source field, destination field, meaning, format, and validation rule. Include ownership of the decision where a field needs interpretation. The document should be understandable to HR and payroll reviewers, not only the person writing the connector. If a reviewer cannot explain a field, it is not ready to influence a live payroll run.
For a hypothetical example, an attendance export might contain employee reference EMP-0042, period reference P-2026-10-A, reviewed ordinary hours 160.00, and approved additional-work units 6.00. The payroll import might require worker code 0042, a payroll period, and separately defined input categories. These labels and figures are illustrative, not a claim about any product’s actual interface or an instruction about pay entitlement.
The mapping must explain how EMP-0042 relates to 0042, which period it belongs to, what the two quantities mean, and who approved them. Do not strip the EMP prefix or rename additional-work units as payable overtime without an agreed rule. The receiving payroll team must approve the interpretation before the example becomes a production transformation.
Specify Units, Dates, and Blank Values
Confirm whether a quantity represents minutes, decimal hours, days, events, or another unit. For example, 7.50 decimal hours means seven hours and thirty minutes; it is not the same duration as seven hours and fifty minutes written as 7:50 in an hours-and-minutes field. The interface contract should state the representation and the approved handling of precision. Payroll and HR must authorise any rounding rules rather than leaving them to an implementation default.
Dates and period boundaries also need explicit treatment. Agree the time zone, date format, period identifier, and handling of work across midnight. If devices or locations record different time settings, investigate that before export. A technically valid date can still belong to the wrong work period if the shift interpretation is incorrect.
Define the meaning of zero, blank, and missing fields. Zero may represent a reviewed quantity, while a blank may indicate that review is incomplete. An absent employee row may mean no eligible input, a mapping failure, or a failed extraction. The integration must not treat these possibilities as equivalent without an approved business rule and a visible exception process.
Document Rules Before Configuring Them
HR should provide the approved treatment of working days, breaks, grace periods, leave, rest days, public holidays, additional work, and absences. The system should implement authorised organisational rules within the agreed scope. It should not create employment policy or infer a legal formula from the name of an attendance category. Where a rule needs specialist advice, the responsible organisation should obtain it before configuration.
Keep the rules readable and versioned. For each rule, identify its owner, effective date, affected employees, and a simple example. If a policy changes, the team must know whether it applies only to future periods or requires a controlled adjustment to earlier records. Untracked rule changes can make two reports for the same period disagree without an obvious explanation.
The ZES time attendance guide covers shifts, exceptions, and reviewed reporting. The integration project should use the approved result of that workflow rather than duplicate the rules independently in several spreadsheets. Repeated implementations of the same policy create more opportunities for inconsistent interpretation.
Assign Overnight Work to the Correct Period
An overnight shift can begin on one date and end on another. Agree how the attendance application associates its events with the scheduled work period and how the resulting quantities are assigned to payroll. Test an ordinary overnight shift, a missing final event, an additional punch, and work that extends beyond the expected finish. Do not assume that a calendar-day report answers the payroll question correctly.
Pay particular attention to the end of a payroll period. A shift crossing that boundary needs an approved treatment that both teams understand. The integration should not split or move quantities simply because a file is generated at midnight. Ask HR and payroll to review representative examples and record the accepted outcome before the connector is approved.
Keep original timestamps available to authorised reviewers, while sending only the agreed payroll inputs through the interface. This allows an exception to be explained without forcing payroll to reinterpret raw device events for every employee. The distinction between original events, reviewed attendance, and exported inputs should remain visible throughout the workflow.
Create an Exception Queue With Named Reviewers
Missing punches, duplicate events, unmatched identifiers, unusual locations, and incomplete approvals should enter a defined review process. Each exception needs a responsible reviewer and a deadline appropriate to the payroll timetable. A dashboard is useful only if someone knows what to do with it. An unresolved queue should not quietly become an approved export when the deadline arrives.
Supervisors can confirm operational facts, such as an authorised off-site assignment or an omitted clock-out. HR applies the organisation’s approved policy. Payroll confirms the receiving input and its treatment within the payroll process. Small businesses may combine some roles, but the responsibilities should still be distinguishable and changes traceable.
Give employees a documented route for raising an attendance discrepancy. Do not treat a missing biometric event as the sole basis for a financial or disciplinary decision. Device faults, enrolment issues, interrupted connectivity, or an incorrect schedule can all require investigation. Human review is part of reliable integration, not a temporary inconvenience to remove after installation.
Preserve Original Records When Correcting Attendance
A correction should retain the original record and show the revised value, reason, author, approver where required, and time of change. The exact features depend on the software, so confirm what the proposed system actually supports. Avoid a process where administrators repeatedly overwrite a spreadsheet until the totals look plausible and no earlier version remains available.
Separate correcting a factual record from authorising a payroll input. A supervisor may confirm that an employee worked at another site, while HR still needs to determine the approved attendance category. One action should not silently imply the other. The workflow should make the sequence clear enough for a later reviewer to reconstruct the decision.
Restrict correction permissions to appropriate roles. Access should reflect responsibility rather than convenience. A branch supervisor may need to review their own team without changing another branch’s data. A technical support user may need diagnostic access without authority to approve payroll inputs. Confirm these boundaries in the actual system, rather than assuming a role name guarantees them.
Choose a Controlled File Exchange When It Fits
A CSV or spreadsheet export can be a sensible integration method when the volume, timetable, and receiving system make it practical. The file still needs a documented format, approved source, secure handling, validation, and reconciliation. Manual transfer is not automatically unreliable; uncontrolled transfer is. The process should identify who generates, reviews, imports, and confirms the batch.
Use a consistent naming and version scheme that includes the period and batch reference. Keep the approved file separate from working copies. If an error requires regeneration, make the superseded version clearly identifiable and prevent it from being imported accidentally. Do not rely on filenames such as final-new-latest to communicate which data has actually been authorised.
Test how the receiving product treats delimiters, text identifiers, date formats, decimal representation, and empty values. Confirm whether it rejects the whole file or only particular rows when errors occur. A file opening correctly in a spreadsheet does not prove that payroll interpreted it correctly. The receiving-system result must be checked.
Consider an API Only After Proving Access
An API-based exchange can reduce repeated manual handling where a suitable interface exists, but it introduces its own responsibilities. Confirm the supported operations, required permissions, authentication arrangements, available documentation, and any relevant limits with the vendor. Do not invent endpoint names or assume that a product’s marketing reference to an API includes the required payroll operation.
The design must explain how requests are recorded, how responses are interpreted, and what happens when a request times out or is partly accepted. A network error does not always prove that the receiving system did nothing. Before retrying, the process needs a way to establish the batch or record status and avoid duplicate application.
Protect credentials through an approved secret-management arrangement and restrict access to the minimum necessary scope. Do not place credentials in exported attendance files or broadly shared implementation notes. Assign responsibility for renewal, revocation, and vendor changes. A connector that works once with one person’s account is not yet an operational integration.
Give Every Export a Batch Identity
Assign a unique reference to each approved export or transfer. Record the period, source version, creation time, reviewer, expected employee count, and relevant control totals. The receiving result should be associated with the same reference where the systems allow it. This makes a support conversation about a specific transfer rather than a vague report that payroll received the wrong file.
Define how a corrected batch relates to the original. It might replace a rejected batch or provide an approved adjustment, depending on the receiving product and process. Do not assume that resending a file replaces earlier values. Some imports add records, while others update them, and the actual behaviour must be established through supported documentation and testing.
Keep the batch register available to authorised operational staff. It should show prepared, approved, transmitted, accepted, rejected, or otherwise relevant states without misleading users about completion. Transmitted is not the same as accepted, and accepted by an interface is not necessarily the same as approved for payroll release.
Reconcile More Than a Grand Total
Reconciliation should compare the approved source with the receiving result. Check employee counts, unmatched identifiers, rejected rows, period references, and totals by relevant category or department. A matching overall total can conceal one employee receiving another person’s input. Include employee-level checks and an agreed sample of detailed records, especially during testing and early operation.
Use a control report that records expected, accepted, rejected, and unresolved quantities. If the receiving system provides an import log, retain it with the batch record. Where it does not provide enough information, define another supported verification method with the vendor and payroll team. Do not call a transfer successful solely because no error message appeared.
Assign a person to sign off reconciliation before the inputs are relied on. The technical implementer can investigate transmission problems, but payroll must confirm the business result. A clear division prevents a developer’s successful API response from becoming an implicit approval of salaries that nobody in payroll has reviewed.
Make Reprocessing Safe and Understandable
Failures happen at different points: extraction, mapping, approval, transfer, or receiving-system processing. The recovery procedure should identify the failed stage before repeating work. If a file failed validation before import, the next step differs from a transfer whose result is uncertain. A universal retry button can be dangerous when its effects are not understood.
Test duplicate submission explicitly. Ask the receiving vendor how repeated batches or records are treated and verify that behaviour in the agreed test setting. Where the interface supports a unique batch or record reference, use it appropriately. Where it does not, establish a controlled operational check. Never promise duplicate prevention that the chosen products have not demonstrated.
Record who authorised reprocessing and what happened to the earlier batch. If a correction must be handled as a later adjustment, follow the receiving team’s approved procedure. The goal is a traceable result, not simply making an error disappear from a dashboard. Recovery is part of the delivery scope and should be documented before go-live.
Build a Test Pack From Representative Scenarios
Prepare test cases with expected outcomes agreed by HR and payroll. Include an ordinary day, an overnight shift, approved leave, a missed punch, duplicate events, authorised additional work, and a rejected request. Add new hires, transfers, inactive employees, and unmatched identifiers. Use anonymised or synthetic information where practical and limit access to any personal information required for testing.
For each case, record the source events or reviewed input, applicable rule, expected export, expected receiving result, and actual result. This makes disagreements specific. If HR expects one treatment and payroll expects another, resolve the policy or mapping question before blaming the connector. The test pack should reveal those differences while they are still safe to correct.
Include invalid data and interrupted transfers, not only ideal examples. A project demonstrated using three perfect records may fail on ordinary monthly exceptions. Ask how the system reports a missing identifier, a malformed quantity, a duplicate batch, and an unavailable destination. The quality of failure handling matters as much as the successful path.
Use a Concrete Trial Workflow
Consider a hypothetical employer testing a small, representative group before a full rollout. HR confirms the employee mapping register. Supervisors review the sample period and resolve exceptions. The attendance administrator generates batch TEST-01, recording the period, employee count, and approved category totals. Payroll checks the mapping and imports the batch into the agreed test environment.
Payroll then compares the receiving records with the approved source, investigates any rejected row, and records the outcome. The team deliberately submits an agreed duplicate test to confirm the receiving behaviour. A separate correction case demonstrates how an authorised change is handled after an earlier export. These steps test the workflow rather than merely proving that a file can move between computers.
The example is not a claim that every payroll product provides a test environment or identical controls. Discovery must confirm what is available. If a vendor cannot support a safe trial method, that limitation must be resolved before the organisation relies on a new direct integration for live payroll processing.
Run in Parallel Before Replacing the Existing Process
A parallel period compares the proposed workflow with the organisation’s controlled existing process. Agree its duration, employee coverage, expected reports, and decision criteria. The purpose is to identify differences and explain them, not to produce two competing payroll instructions. Keep authority over the live payroll process explicit throughout the trial.
Investigate discrepancies by category. They may come from employee mapping, shift interpretation, approval timing, units, or the receiving product’s import behaviour. Record the cause and resolution. Do not force totals to match through unexplained adjustments, because that hides the very issues the trial is intended to expose.
The organisation should decide when the evidence is sufficient to proceed. Sign-off should include the responsible attendance, HR, payroll, and technical representatives. Unresolved material differences should remain visible with an agreed action. A project deadline is not evidence that the integration is ready, and a successful demonstration is not a substitute for representative operational testing.
Agree Cut-Offs and Late Corrections
Set deadlines for attendance review, HR approval, export generation, import, reconciliation, and payroll release. Give each deadline an owner and explain the dependencies. Supervisors need enough time to investigate exceptions, while payroll needs a stable approved input. An integration cannot compensate for a process where every team expects another team to approve the same unresolved record.
Define what happens after the attendance period is closed. A late correction should follow an authorised route and preserve the history of what was previously exported. Depending on the organisation’s process and the receiving system, it may require reopening a controlled period or creating a later adjustment. The responsible payroll team must approve the treatment.
Notify affected reviewers when a closed result changes. Otherwise payroll may reconcile one version while a supervisor sees another. Version references and batch records make this discrepancy easier to detect. The aim is not to prevent legitimate corrections; it is to ensure that corrections remain visible, authorised, and associated with the correct payroll action.
Protect Information Without Moving More Than Necessary
The payroll interface should carry the information needed for its defined purpose. It should not routinely include biometric templates or unrelated employee details simply because they are available in the attendance platform. Ask which fields are required and remove unnecessary exports from the approved design. Separating purposes reduces avoidable exposure and makes the mapping easier to review.
Agree access roles, storage locations, retention arrangements, and approved transfer methods with the organisation’s responsible teams. Avoid uncontrolled copies of attendance and payroll files across personal devices or informal messaging channels. Keep diagnostic logs useful without exposing more personal information than support needs. These are implementation decisions that should be reviewed against the organisation’s applicable obligations and policies.
For Kenya-specific privacy requirements, consult the ODPC guidance library and obtain appropriate advice for the actual processing arrangement. A configured attendance system or encrypted transfer does not, by itself, establish compliance. The organisation remains responsible for determining its lawful and appropriate handling of employee information.
Plan Multi-Branch Ownership and Local Exceptions
Distributed workplaces need a clear distinction between local review and central payroll approval. Branch supervisors may confirm events for their teams, while a central HR or payroll team controls the final export. Define how transfers between branches affect identifiers, schedules, approval responsibility, and period allocation. Do not create a second active employee merely because a person changes location.
Test delayed synchronisation and branch connectivity interruptions where relevant to the actual devices and software. Establish how recovered events are reviewed before a period closes. Product capabilities differ, so do not assume every terminal stores, synchronises, or deduplicates records in the same way. Discovery and acceptance testing must confirm the proposed arrangement.
Employers with distributed sites can also review the ZES multi-branch attendance guide. Manufacturing teams may find the factory biometric attendance guide useful for shift and clocking-point planning. Those operational decisions feed the integration; they should not be postponed until payroll discovers unexplained branch totals.
Define Support and Change Responsibilities
Name the owner of each stage after handover. Attendance support handles device and source-data issues within its scope. HR owns policy interpretation and authorised employee changes. Payroll owns receiving rules and release decisions. The integration support provider investigates mapping and transfer behaviour. Vendor involvement may be needed for product defects or interface changes. The contact plan should reflect these boundaries.
Agree how software upgrades, licence changes, authentication changes, and new payroll fields are reviewed. A connector tested against one version may need reassessment when a product changes. Keep interface documents and mapping versions current. Do not assume that a working integration can be left unchanged indefinitely while both systems evolve around it.
Train a primary and backup operator. Give them a runbook covering normal processing, validation, reconciliation, common failures, and escalation. Support should not depend on one person’s memory of how a monthly file was prepared. Clear operational records help the organisation continue when staff are absent or responsibilities change.
Use an Acceptance Checklist Before Go-Live
Confirm that the exact interfaces are accessible and supported; employee identifiers map correctly; units and dates are defined; shift and exception rules have authorised owners; and the export contains approved inputs. Verify that invalid records are visible, receiving results can be reconciled, and repeated submissions have a tested treatment. Each check should have evidence rather than a verbal assurance.
Then confirm the operational side: cut-off dates, correction procedures, role permissions, secure handling, backup operators, vendor contacts, and the support runbook. Ensure that the final mapping and test results describe the configuration actually delivered. A last-minute product substitution or format change must not bypass the checks applied to the original proposal.
Finally, obtain business sign-off from the people responsible for attendance, HR, and payroll. The technical team should demonstrate the transfer, but the organisation must approve the meaning and use of the inputs. This is the point where the project becomes an accountable operating process rather than a connector that happened to pass a demonstration.
Request a Scoped Integration Assessment From ZES
Prepare the terminal models, attendance software, payroll product and version, available export or import documentation, employee counts, branch structure, and approved shift rules. Include an anonymised sample of the required output and describe the current monthly process. Identify who can authorise access to vendor documentation or a test environment.
Ask the proposal to separate discovery, mapping, implementation, testing, training, and ongoing support. Costs depend on the available interfaces, data quality, rule complexity, locations, historical migration, and vendor assistance required. A fixed promise based only on two product names is not a reliable basis for a consequential payroll workflow.
To begin, contact ZES for a biometric attendance and payroll integration assessment. Request a demonstrated data path, an approved mapping, a representative trial, and a reconciliation process. Those deliverables help turn clocking records into dependable, reviewed payroll inputs while keeping financial decisions with the people authorised to make them.