BI without a data team: dashboards people actually open
Most companies do not need a data warehouse. They need three numbers they currently assemble by hand, updated automatically, in a place where the management team will look at them. That is a two-week BI project, not a two-year programme.
Work backwards from decisions. Which meeting happens weekly, and which number is argued about in it? Revenue against plan by channel, margin by product group, receivables ageing, warehouse turnover, sales funnel conversion by stage. Each of those is a report with a named owner and a decision attached. Everything else is a chart nobody will maintain.
Then check the sources. A modern BI platform connects directly to your accounting database, CRM, spreadsheets and the warehouse system; the hard part is agreeing which one is authoritative when two disagree by 3%. Resolve that once, in writing, and the dashboards stop being a debate about methodology.
Three reports that usually pay for the platform
- 01Plan versus actual, daily. Because a shortfall found on the 5th can still be fixed; one found on the 30th cannot.
- 02Margin by product and client. The usual outcome: 20% of the range earns everything, and two large clients are sold at a loss.
- 03Funnel conversion by stage. Shows where deals stall, which is where the sales process needs changing rather than the team.
Adoption depends on delivery, not on visualisation. Push the dashboard where people already work - email digest on Monday morning, a screen in the sales room, a link pinned in the messenger. A beautiful report behind two logins and a VPN is a report that gets replaced by a spreadsheet within a month.
In our catalog, BI is one of the categories where the pilot is decisive: we connect one real data source, build the two reports your board argues about, and let you judge on your own numbers. If the platform cannot answer the questions you actually ask, you find out in week two and pay nothing for a licence you would not have used.
Start with the decision, not the dataset
A dashboard exists to change an action. Before building anything, write the sentence: "every Monday the commercial director decides X, and needs Y to decide it." One decision, one owner, one cadence. If nobody would act differently based on the numbers, you are building a report for the archive.
Three numbers beat thirty. A board with three metrics and a clear owner gets opened weekly. A board with thirty gets opened during the review that commissioned it and never again. Extra detail belongs one drill-down deeper, not on the first screen.
Data quality is a process, not a project
Analytics inherits whatever discipline the operational systems have. A one-off cleanup improves the numbers for a quarter; ownership keeps them right.
- 01An owner per field. Someone is accountable for deal stage being accurate, someone else for product category. Unowned fields decay within weeks.
- 02Validation at entry. Required fields, controlled lists and format checks in the source system are cheaper than corrections in the warehouse.
- 03An exception report. A short weekly list of records that break the rules, reviewed for twenty minutes by the people who created them.
Making dashboards stick
Put the numbers where the work happens: a link in the weekly meeting agenda, a scheduled PDF for the people who will never log in, a threshold alert in the messenger the team already uses. Review usage quarterly and retire boards nobody opens - an unused dashboard is not neutral, it dilutes trust in the ones that matter.
Certify the definitions. One agreed formula for revenue, active customer and margin, published where anyone can read it. Most reporting arguments are definition arguments in disguise.