PRACTICAL GUIDE FOR SWISS SMEs
A five-step method for moving from scattered files to faster, more reliable and more profitable decisions.
| IN BRIEF Start with a business question that matters enough to justify change, is specific enough to measure and is simple enough to produce an early result. |
A data project starts with the business
When an SME asks where to start with data, we begin with its day-to-day operations. Which number comes too late? Which decision still relies on intuition because the information is unreliable? Which report keeps your team busy when it should be helping them? This friction provides the starting point for a useful first project.
Think about your next management meeting. Which question almost always sets off a search for a file, a call to Finance or twenty minutes of discussion about which figure is right? That is often your first use case. We look for a project that removes an obvious frustration, creates a shared language and makes your teams want to continue.

We start with a business question and connect the data that matters to a first shared dashboard.
A concrete frustration often prompts action. Executives wait days for a figure. Two departments give different results for the same indicator. The monthly report takes someone half a day. A manager finds out about a stockout or a margin drop too late. These symptoms usually indicate that data is spread out, poorly structured or hard to turn into decisions.
A successful data initiative therefore starts with a business objective, then organises the processes, data, tools and responsibilities needed to meet it. This avoids two common pitfalls: buying a solution before clarifying the need and trying to solve every problem at once.
Before you begin, choose a first use case
In workshops, we often ask each person which decision they are putting off for lack of reliable information. The answers tell us a lot. Executives mention margin, operations teams talk about workload and sales point to the order book. We choose a question that combines a visible need, accessible data and someone who is ready to use the result. This compromise is more useful than a long wish list.
A good first use case meets three criteria. It addresses a real priority, most of the necessary data already exists and you can see the result quickly. Choose the one that creates enough value and confidence to move forward.
In sales, get revenue, margin and order book figures every morning. In finance, automate a monthly report currently consolidated by hand. In operations, track lead times, defects, downtime or team workload. In logistics, monitor stockouts, inventory coverage and turnover. In customer service, identify common reasons people get in touch and how long requests take to resolve.
| QUICK TEST If the result would change no decision, action or behaviour, the use case is probably too vague. Rework it until you can name who will use the information, how often and to make which decision. |
The roadmap we recommend
Datavizin’s editorial plan organises the first project around five complementary steps. Some steps overlap. Governance and user support, for instance, need consideration from the start. Following this sequence keeps the project grounded in how the business works.
1. Link data to a business objective
Describe the problem in business terms. Your goal might be to reduce reporting preparation time, protect margin, cut stockouts, improve forecasts or increase the conversion rate. Then link a small number of indicators to that goal.
A useful indicator is understandable, measurable and actionable. “Understand sales better” is broad. “Track revenue, gross margin and conversion rate by segment every week” provides a specific direction. For each KPI, record its definition, source, update frequency, owner and the decision it informs.
Deliver a one-page summary of the problem, users, three to five KPIs, expected value and a measurable success criterion.
2. Make collection and processes reliable
Data reflects the processes that produce it. When staff enter customer references differently in the ERP and CRM, a date means something different in each department or certain rows are never completed, a dashboard will not offer a lasting fix.
Map the useful sources, such as Excel files, ERP, CRM, accounting software, production tools, forms or databases. For each critical field, identify the owner, how it is entered and which checks are already in place. Start with problems that can change decisions: duplicates, missing values, inconsistent units, invalid dates and conflicting definitions.
Name an authoritative source for each key data item. Standardise formats, lists of allowed values and entry rules. Document the transformations applied between the source and the report. Add a few automatic checks and track errors.
3. Choose architecture that fits
An SME does not need to reproduce a large group’s architecture. For a simple first use case with one source, a structured file may be enough. A database, data mart or warehouse makes sense when you need to combine systems, keep a history or use shared indicator definitions.
The architecture should reduce manual work and leave room for growth. A common design extracts data from source systems, consolidates it, cleans and models it, and makes it available in a BI tool. Cloud services and standard connectors often make it possible to start in stages without excessive upfront investment.
| KEEP IT SIMPLE Choose the simplest solution that provides the reliability, historical data, security and automation your use case needs. Add complexity to address a real constraint. |
4. Establish clear, lightweight governance
Early data governance is mainly about making a few rules and responsibilities explicit. Decide who can view or change data, who approves a KPI definition, how errors are corrected and how long information is kept.
For personal data, the revised Swiss Federal Act on Data Protection calls for particular attention to purpose, proportionality, security and transparency. Incorporate data protection by design and by default into the project. Collect what is necessary, restrict access, document processing and plan how to handle incidents.
A business lead can act as the data owner for their domain. The role involves deciding definitions, approving access and keeping the data fit for business use, while technical work is shared with others.
5. Support adoption of the new workflow
We build training into the project from the beginning. Users should see intermediate versions, challenge definitions and test the dashboard with their own scenarios. Trust grows when someone sees a variance they already know about and understands why the report shows it.
A dashboard delivers value when it becomes part of regular decision-making. Prepare for adoption with future users. Involve them in prototypes, observe their questions, simplify the navigation and explain the definitions. Brief, role-specific training is more effective than a long demonstration of every feature.
Set a regular review rhythm. Who uses the dashboard and in which meeting? What happens when a threshold is crossed? How are requests for changes prioritised? These answers turn a technical deliverable into a management practice.
A realistic 90-day action plan
| Period | Objective | Main deliverables |
| Days 1–30 | Define the scope | Use case, KPIs, source map, data quality assessment, risks and responsibilities. |
| Days 31–60 | Build | Minimal data workflow, reliable model, first dashboard, checks and a tested prototype. |
| Days 61–90 | Roll out | Automation, training, documentation, regular reviews and measurement of early benefits. |
This timetable assumes a limited scope, accessible data and decision-makers who are available. It provides a useful guide for keeping a first project from getting stuck in analysis.
The most costly mistakes
Buying the tool before defining the problem.
A licence alone cannot create shared definitions, ensure quality or drive adoption. First get agreement on the use case and the decisions it should support.
Tracking too many indicators.
An overloaded dashboard distracts users. Start with a few KPIs linked to objectives and add details when requested.
Underestimating cleanup and change management.
Collection, quality and user support often require more work than visualisation. Set aside time and assign ownership for each.
Scaling too early.
A prototype should prove value and settle definitions. Standardise and scale later, once security, performance and maintenance requirements are clear.
Overlooking compliance.
Access, data minimisation, retention and traceability belong in the design of the data service from the start.
How do you know the first project succeeded?
The best sign is hearing something like “We saw the variance on Tuesday and could act before Friday.” A project succeeds when decisions happen faster and your teams can discuss a figure without having to rebuild it first.
Measure success by operational change. A report now takes twenty minutes instead of half a day. A meeting finally uses shared figures. A team acts on an alert before a stockout or answers its usual questions on its own.
Track adoption through usage frequency and the number of decisions the tool supports. Assess efficiency through time saved, manual tasks removed and how quickly information becomes available. Measure quality through fewer errors, duplicates and discrepancies. Look for business impact in protected margins, lower inventory, improved sales or better-managed risks. Gauge self-service skills by whether teams understand the indicators and can adapt simple analyses.
Start small and build a sound foundation
Data maturity grows through a series of useful projects that improve data quality, skills and trust. For an SME, the strongest first project addresses a recognised problem and lays foundations you can reuse.
Datavizin’s Data Starter Pack supports the entire process, from a maturity assessment, selection of the first use case and data structuring to dashboard development and skills transfer. It gives the business a well-defined, compliant foundation that can grow and that its own teams can manage.
| NEXT STEP Choose one critical process, list the decisions that currently lack reliable information and gather three examples of existing reports. This is enough to organise an initial scoping workshop with Datavizin. |