All resourcesPractical guide

Excel or a training system: selection and a small pilot

A spreadsheet is a useful starting tool and remains valuable for analysis. Problems appear when it must simultaneously store requirements, version history, results, reminders and ownership of overdue actions. Look at what maintaining it takes from a leader's time for explanation, feedback and work alongside the team.

Author: Piotr Adamski · Updated:

Choose the tool around the work it must control

A spreadsheet can be a suitable training register when the process is simple, ownership is clear and records remain reliable. Moving to a dedicated system is justified when it reduces identifiable control or administration problems, including repeated status reconciliation that interrupts a leader's work with employees. The number of employees alone does not decide the issue. Versions, shifts, substitutions, assessment types and reporting needs can create more complexity than headcount.

ISO's public guidance on documented information for the 2015 edition recognises different media [1]. It does not impose one training software product. EDPB guidance also makes clear that personal data obligations apply to structured paper and electronic records, including employee identifiers [2]. Buying software therefore does not create a privacy exemption or automatically make an unreliable process compliant.

Examine the process before the product

Begin with a process review. Where do requirements originate? Who assigns them? How are people identified? Where are results kept? What happens after a procedure revision, an unsuccessful assessment or a transfer? The answers reveal whether the current difficulty is a tool limitation, an unclear rule or missing responsibility.

The comparison below is a proposed decision method. It uses your own observed workload and evidence needs rather than invented industry benchmarks. A system should earn its place by handling the relevant cases consistently and preserving access to the records the organisation needs.

Identify the spreadsheet's actual failure modes

Look at recent work, not just the workbook's appearance. Count instances where the owner had to reconcile names, correct status formulas, locate a source version or combine several files. Record what caused each correction. A mistake in an employee roster needs a different solution from a formula copied into the wrong range. Check whether there is one authoritative file. Email attachments, local copies and separate shift workbooks can create competing versions. If managers update their own copies, the coordinator must reconcile events before the matrix can be trusted. A shared file can reduce duplication, but it still needs ownership, access controls and a clear correction process.

Review historical evidence. Can you identify the material used for a completion six months ago? Does a new course version overwrite the old title or result? Can you distinguish an unsuccessful attempt from a later success? A workbook containing only one current status may be sufficient for a narrow planning view while insufficient for evidence reconstruction. Inspect overlooked populations. Workers without assignments do not necessarily appear as overdue. Substitute operators, agency staff and people returning from leave can fall outside a register built from an old list. A new tool will reproduce the gap unless the assignment rule and roster source are corrected too.

Measure administration without inventing savings

Use a short observation period covering ordinary activity and at least one change. Record time spent creating assignments, chasing completions, correcting data, compiling reports and answering evidence requests. Include team leaders' time as well as the coordinator's, and note which coaching or follow-up work gets delayed. Identify whether the work is recurring or a one-off cleanup. A migration business case should not count the same temporary effort as permanent monthly savings. Build an illustrative calculation using your own inputs. If a coordinator records four hours of report preparation in a month and their agreed internal hourly cost is known, that gives a local administration cost. Do not present a hypothetical number as a typical result for other companies. Include supervision and technical review that will remain necessary after automation.

Account for subscription, setup, migration, training, support and exit work. A system may reduce reconciliation while introducing user administration and supplier management. Consider who will own those tasks and whether the organisation has time to implement the workflow properly. Define the improvement you expect to observe. It might be fewer manual reconciliations, faster retrieval of a selected historical result or fewer requirements missed after role changes. Use measurable local criteria. Claims such as "eliminates administration" are implausible because the organisation still needs to decide requirements, approve content and assess exceptions.

Test representative cases before buying or migrating

Select one real procedure and a controlled sample of records. Include a normal completion, an unsuccessful attempt, a repeat, an absent worker, a substitute role and a new release. If practical authorisation is relevant, include that separate evidence too. These cases reveal more than a demonstration built entirely from uncomplicated courses. Ask the proposed tool to preserve the relationship between the person, requirement, source release, learning material and result. Verify which decisions it supports directly and which records remain elsewhere. A dashboard may summarise activity clearly while leaving version history or assessment evidence difficult to retrieve.

Test assignments from the intended population. Check whether unassigned people remain visible. Ask how a joiner, transfer or departure changes the current view without rewriting history. Confirm how the process handles people who need assistance with access or language. Evaluate the export with the actual recipient. A coordinator and an auditor may need different fields. Check whether the file identifies scope, date and source versions. Avoid assuming that the existence of an export button means the report can answer your evidence question. Open the exported file and reconstruct one selected result. Ask a leader to use it to prepare an employee's next learning step, and observe where they still need to chase information.

Compare access and data protection in practice

A training register contains personal data, even if it records only names and statuses. Define the purpose, necessary fields, access and retention in both options. EDPB guidance emphasises minimisation and storage limitation [2], while its security guidance discusses access controls and protection of confidentiality, integrity and availability [3]. For the spreadsheet, review the shared folder, permissions, backups, change history and downloaded copies. For a system, review user roles, authentication, supplier terms, data location, support access and available export or deletion workflows. Assess the actual configuration rather than treating a familiar file as harmless or a specialised product as inherently safe.

Identify who can edit evidence and how corrections are recorded. Managers may need to view task status without being able to alter assessment results. A technical reviewer may need content access without seeing unrelated employee data. Choose permissions that reflect these duties. Check availability and recovery. What happens if the workbook is damaged or the service is unavailable before a shift? Retain an appropriate contingency method for operational decisions. The system does not remove the need to know which people can perform critical tasks when the usual screen cannot be opened.

Plan migration as a reconciliation exercise

Do not import every historical cell without checking its meaning. A green cell might represent attendance, a passed test, a manager's judgement or an expired approval. Map old fields to the new model and label uncertainty. Importing ambiguous data under a stronger label can create an unsupported claim of competence. Identify the source and owner for each dataset. Reconcile employee identities, current roles, requirement versions and result dates. Keep a record of transformations and unresolved issues. Where evidence is missing, record the gap rather than inventing an event from a vague status.

Decide what historical detail needs migration and what can remain in a controlled archive. The current operational view and long-term evidence need not contain identical detail. Ensure archived records remain retrievable for their defined retention period and that references from the new system make sense. Use a planned cutover. Specify the last update in the old register, the reconciliation check and the first controlled update in the new workflow. Running two editable registers indefinitely creates the same duplication problem the migration was meant to solve. Keep the old file as an appropriately labelled snapshot where necessary.

Preserve an exit route

Before committing to a service, examine how the organisation can retrieve its records. Export should include enough information to identify people, requirements, source versions, attempts and assessment decisions. A collection of current green statuses may be insufficient to rebuild the evidence trace elsewhere. Check what happens to linked attachments and historical material. An export that points to inaccessible supplier URLs can leave the organisation with names and dates but no source evidence. Clarify formats, timing and access during termination through the applicable product information and agreement.

Document the internal owner of portability. Someone should know where exports are stored, what they contain and how their completeness is checked. Regularly creating unnecessary copies can increase exposure; never testing export can leave an unexamined dependency. Choose a proportionate approach based on record importance and recovery needs. For a spreadsheet, the exit issue looks different but still exists. Macros, undocumented formulas and one person's knowledge can make a workbook difficult to maintain. Use a clear data dictionary and written rules. Independence depends on interpretable records, not solely on whether the file opens in familiar software.

Make the decision after a limited pilot

Keep the spreadsheet if it handles the representative cases reliably and its maintenance cost remains acceptable. Improve its ownership, definitions and access before replacing it solely because a dedicated system appears more professional. A small, well-controlled register can be the sensible choice.

Choose a system when the pilot demonstrates a material improvement in your process. CrewShift can support procedure assignments, completion scores, history and exports. Create an account to evaluate those functions with your own scope and verify the current offer, workflow and support arrangements. Before subscribing, compare the observed administrative effort and the time available for team support with your starting point. A product cannot define competence criteria or approve operational exceptions on the organisation's behalf.

Record the decision and assumptions

Record the decision and its assumptions. Include expected local benefits, total cost, migration scope, data protection considerations and criteria for reviewing the choice. If the pilot fails to preserve history or makes assignments harder, that evidence should affect the decision even if the presentation looks good.

The useful comparison is practical: which option gives the responsible people a reliable current view and a reconstructable history at an acceptable cost? Answer it with the same process sample in both tools, including whether the leader can get back to supporting the team after checking the record. Familiarity, dashboard polish and employee count alone do not answer that question.

Questions worth asking

Must Excel be removed completely?

No. A spreadsheet remains useful for one-off analysis and exports. It should not be maintained manually as a second source of current status.

How many employees justify a system?

There is no universal threshold. Change frequency, risk, number of requirements and reconciliation cost matter more than headcount alone.

Is employee headcount a sufficient buying criterion?

No. Assess versions, task differences, substitutions, reporting and reconciliation effort. A smaller complex process can need more control than a larger simple one.

What should an exit test examine?

Check whether exports retain identities, source versions, attempts, decisions and usable references to supporting evidence. Current statuses alone may not allow reconstruction elsewhere.

Sources and scope

  1. ISO / TC176 — guidance on documented information (ISO 9001:2015)
  2. European Data Protection Board — data protection basics
  3. European Data Protection Board — secure personal data

This informational material does not replace licensed standards or individual legal advice.