A dashboard is easy to buy and hard to make true. The demonstration is clean, the charts respond instantly, and every number looks authoritative. Your own data is spread across an accounting file, a spreadsheet, a point-of-sale system, a courier portal and a few inboxes, and none of it agrees.
An integrated operations dashboard is therefore not a screen with many charts. It is a small number of trusted numbers, refreshed on a known schedule, tied to decisions that someone actually makes. Buying it well means buying the definitions, the pipeline and the ownership as much as the visual layer.
Begin with the decisions the dashboard must support
List the decisions that currently wait for a person to assemble a report. For each, record who decides, how often, what they look at and what they do next. A useful first list is short:
- Daily cash position across branches and payment channels.
- Stock cover and pending transfers between locations.
- Sales against target by channel, with returns separated.
- Order backlog and dispatch exceptions.
- Outstanding receivables by customer and age.
If a chart does not change a decision, it is decoration. If a decision has no owner, the dashboard will become a monthly argument about whose number is right.
What the dashboard brief must specify
Write requirements in terms a vendor can test, not in terms of charts:
- Sources and owners: every system, file and manual input, with a named owner for each.
- Grain: whether each metric is per order, per day, per branch, per customer or per invoice line.
- Refresh and cut-off: how often data updates and which transactions belong to which day.
- Definitions: formula, exclusions and version for every metric on the first screen.
- Access: what a branch manager, finance user, director and auditor can see.
- Drill-through: the ability to move from a total to the underlying records.
- Alerts: thresholds that notify an owner, not a channel nobody reads.
- Export and audit: who extracted what, and whether the export reconciles to the source.
- Performance: response time at month-end and during peak trading.
The vendor landscape in plain terms
A business intelligence platform with an implementation partner gives strong visual tools and a partner who builds the first models. The risk is that the partner owns the semantic layer and the client owns only a licence.
An ERP or accounting reporting module is already connected to one source and usually cheaper to start. It sees its own data well and other systems poorly, so cross-functional questions often return to spreadsheets.
An integration specialist is strongest when the hard part is moving and reconciling data, not drawing charts. Confirm that the visual layer, definitions and support are included rather than assumed.
A custom build fits an operating model that no product expresses without heavy workarounds. It also creates a maintenance obligation, so require documentation and a named internal owner.
An internal build can work when a capable analyst already owns the warehouse. Be honest about key-person risk and the time required to maintain connectors and definitions.
A scorecard for shortlisting
| Criterion | Weight | Evidence to request |
|---|---|---|
| Source coverage | 20% | Named connectors and a live test on your data |
| Metric governance | 15% | Definition register with owners and versions |
| Usability for decision owners | 15% | Task-based demo with a branch manager |
| Performance at peak | 10% | Test at month-end data volume |
| Access and security | 10% | Role matrix, audit log and data handling terms |
| Implementation approach | 10% | Phased plan with acceptance gates |
| Total cost of ownership | 10% | Three-year view including connectors and support |
| Exit and portability | 10% | Export format, documentation and termination terms |
Score each criterion from one to five against evidence. Treat source coverage, access control and a documented exit as must-haves; a strong visual demo cannot compensate for a failed must-have.
Questions that change the three-year total
- Is licensing based on named users, concurrent users, data volume or a capacity tier?
- Which connectors cost extra, and what happens when a source API changes?
- Is the implementation priced by days, by deliverable or by phase?
- Who pays for data cleaning that the source system should have done?
- What is included in support, and what counts as a change request?
- What does training cost for branch users who will not attend a long course?
- What is the exit assistance fee, if any?
Ask for a three-year total, not a first-year subscription. The cheapest year one can be the most expensive by year three.
Design the pilot around one decision
Pick a decision with a manual comparison available, such as the daily cash position across two branches. Build only the sources and definitions that decision needs. Run the dashboard beside the current process for two weeks and reconcile every difference. The pilot passes when the owner uses it to make the decision and the numbers can be traced to source records.
Exit terms and ownership
The contract should state who owns the data model, the metric definitions, the queries and the dashboards. Require an export of the definitions and a documented data dictionary, not only a CSV of results. Confirm what happens to your data at termination, how long it is retained and what assistance the vendor provides during a transition. A dashboard you cannot leave is a subscription with extra steps.
Red flags in the sales process
- The demo uses a curated sample that looks nothing like your source data.
- The vendor promises unlimited dashboards but cannot describe a metric owner.
- No one can show lineage from a chart to a source field.
- The licence is quoted before the scope is agreed.
- Acceptance is defined as "dashboard delivered" rather than a decision supported.
- There is no exit clause or export commitment.
Where this fits in our delivery work
We build integrated dashboards and the data layer beneath them inside Custom Software & Integrations. The pricing page explains how discovery, integration and support are scoped. For the conceptual model behind a good dashboard, read how a business KPI dashboard is put together.
Sources
- DAMA International, Data Management Body of Knowledge, accessed 17 November 2026.
- Kimball Group, Dimensional modeling techniques, accessed 17 November 2026.
- GOV.UK Service Manual, Measuring success, accessed 17 November 2026.
- Microsoft Learn, Power BI guidance, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.