Business intelligence consulting services, from the systems you run to the reports you read

Who are business intelligence consulting services for?

Business intelligence work fits a business whose numbers live in several systems and reach the people who need them too late to act on. Someone exports a report from accounting, another from the CRM, pastes both into a spreadsheet every Monday, and the meeting opens with a debate about whose total is right. What the reporting has to do changes as a business grows:

• Established local businesses: one location, an owner who decides, and a handful of systems such as accounting, scheduling and a CRM. The first job can be one weekly view of sales, work booked and cash that nobody has to assemble by hand.
• Multi-location and regional businesses: several branches or a wide service area, with managers who each need their own numbers. Reports have to compare locations on the same terms, with each manager seeing their branch and the owner seeing all of them.
• National businesses: many systems, often several brands, and a finance or operations team that already reports. The work is a shared set of measures across the business, access rules that hold up when someone checks who saw what, and reports that keep pace with the volume.

Midnite Systems was formed in 2023, and we build reporting from our office in Valencia for businesses across the US. If a report inside a tool you already pay for would answer the question, we’ll say so on the call. Our custom software page covers the rest of what we build.

What problem does business intelligence solve?

IBM describes business intelligence as the technology and processes a business uses to gather, manage and analyze its data, so that what the data shows can shape strategy and operations. For an operating business, the problem it solves is simpler to state: the facts a decision needs are spread across systems that don’t talk to each other.

Invoices sit in accounting, deals in the CRM, jobs or orders in an operations system, hours in payroll. Each system reports on itself and answers questions about its own records. Many of the questions that matter cross them. Which customers are profitable once the cost of serving them is counted? Which branch is slipping before the month closes? Which salesperson’s deals turn into late payers?

IBM notes that BI presents its results as reports, charts and maps, which might be shown on a dashboard, and that BI platforms have traditionally relied on a data warehouse, which gathers data from several sources into one system. The dashboard is the part people see. Much of the work sits underneath it: agreeing what each number means, getting the records out of each system, and cleaning them so the totals hold.

What does business intelligence consulting involve?

A business intelligence consultant works out what a business needs to measure, then builds the path from its systems to the reports. IBM sets the work out in steps: identify the data to review, collect it and clean it, look for trends, present it in dashboards people can drill into, then weigh the results against past data and key performance indicators to plan what to do next.

What we would build follows that order:

• The measures: the few numbers each team decides by, each written down with how it is counted, so that “revenue” or “on time” means one thing in every report.
• The data map: which system holds each number, which field it sits in, and where two systems disagree about the same customer or order.
• The pipeline: code that pulls records from each system on a schedule, cleans them and loads them into one reporting database.
• The reports: dashboards and scheduled reports laid out for the people who read them, with a way to open any total and see the records behind it.

We build it in-house, with the code kept in a repository of its own. Much of the effort goes into the first three steps, since a dashboard can only be as good as the records it reads.

Where do the numbers on a dashboard come from?

A number on a dashboard usually starts in a system someone already uses. There are three ways to get it out, and a project can mix them:

• An API: many current accounting, CRM and scheduling systems let other software read their records through an API. Where a system has one, it is the direct route, and it can run as often as the system allows.
• A scheduled export: some older systems offer a nightly file in place of an API. The pipeline collects the file, checks it arrived whole and loads it.
• The database itself: a system you host may let the reporting side read its tables directly, with read-only access so nothing in reporting can change the source.

Microsoft says Power BI Desktop connects to more than 100 data sources, from databases and cloud services to files and web sources. A ready connector shortens the work. The time goes on the system with no connector, the field whose meaning changed three years ago, and the customer who appears under four spellings. Our page on API integration services covers the connections themselves, including what happens when a source system changes or stops answering.

Should reports live in a BI tool or inside your own system?

Reports can live in two places, and the choice shapes what they cost to run.

A business intelligence tool such as Power BI or Tableau is made for analysis. Microsoft lists scheduled refresh, email subscriptions, alerts and row-level security among the Power BI service’s features, and an analyst can build a new report without waiting for a developer. Microsoft and Tableau license the people who build reports per user, and sell access for the people who read them either per user or by capacity, so price the tool against the number of people who will use it. If you already pay for one of these tools and your team uses it, we build on it.

Reports can also be screens inside a system we build, such as a custom CRM or a web application your team already logs in to. The numbers sit beside the work they describe: a manager sees the late jobs on the screen where jobs are assigned. No reporting license is needed, and each new report becomes build work. Low-code tools can carry simple reporting screens too, and our page on no-code and low-code development covers when they fit.

Some businesses end up with both: operating numbers inside the system people work in, and a BI tool for the analysis that crosses systems. The proposal says which we would build, and why.

How do you know the numbers on a dashboard are right?

A dashboard caught in one wrong total can lose its readers’ trust, so the checks belong in the build from the start. What we would build carries these:

• Totals that match the source: before launch, each report’s totals are compared with the systems they came from and with the figures your team already trusts, and every gap is explained before anyone decides from the number.
• One definition per measure: each measure is defined once, with how it’s counted, and every report uses that definition.
• A visible refresh time: each report shows when its data last updated, so nobody reads yesterday’s figures as today’s.
• An alert on a failed load: when a source system doesn’t answer or sends half a file, the load stops and tells someone, and the report keeps the last good figures.
• A path to the records: any total someone doubts opens to the records behind it.

That last check earns its keep in a review. When the owner asks why a branch’s margin dropped, the answer should be a click or two away, in the same report.

When is a spreadsheet enough?

For plenty of businesses, it is. A business with one location, one system that holds most of its records and a weekly review that takes an hour to prepare may not need business intelligence work yet. The reports built into your accounting system or CRM, or a well-kept spreadsheet, can carry a business a long way.

The signs that a spreadsheet has run its course are practical:

• The same report is rebuilt by hand: someone spends part of every week copying exports into a workbook.
• Two people get two totals: the sales figure depends on who pulled it and when.
• Numbers arrive after the decision: the report shows last month when the decision is about this week.
• The file has one keeper: only one person understands the formulas, and their week off stops the reporting.

If you recognize one of these, our guide on moving from a spreadsheet to a dashboard walks through the options. Whether to buy a reporting tool or have one built is a wider question, covered in our guide to building or buying software.

How do business intelligence companies differ?

Business intelligence companies fall into three broad groups, and the right one depends on what you already have:

• Platform partners: firms that specialize in one vendor’s tool, such as Power BI or Tableau, and implement and support it. They suit a business that has chosen its platform and needs it set up well.
• Analytics firms: data teams that build warehouses, models and statistical analysis. They suit large volumes of data and questions that need a statistician.
• Software builders: firms that build the systems a business runs on, and the reporting with them. They suit a business whose numbers are hard to get because its work runs on spreadsheets and tools that don’t connect.

We are a software builder, and we are not a partner of any BI vendor. When your systems already hold clean records and you only need reports in one tool, a platform partner may be the better hire, and we’ll tell you so. When the records themselves are the problem, the reporting starts with the systems, which is the work our process automation page describes.

Where does AI fit in business intelligence?

BI tools now carry AI features. In Power BI, Copilot answers questions about a report’s data asked in plain words and can summarize a report. Microsoft says it needs a paid Fabric or Premium capacity; a Pro or Premium Per User license alone isn’t enough. A question in plain words can help someone who doesn’t know a report’s layout find a figure. These features are only as good as the data underneath, and Microsoft notes that Copilot can misread data nobody has prepared for it. Microsoft also says that when a question isn’t about the data in the report’s semantic model, Copilot answers from the language model’s general knowledge.

AI also has a role before the dashboard. Many numbers a business wants are trapped in documents: order emails, supplier PDFs, signed forms. Reading those into records usually takes a model, which is the work on our AI integration page. Once a model has turned them into records, they feed the same reporting database as everything else.

Our advice is to settle the measures and the data first, and to add AI features once the numbers underneath them hold.

What does reporting need after launch?

Reports drift as the business changes. A new product line, a new branch, a renamed field in the CRM or an update to a vendor’s API can each break a total without anyone noticing. Looking after reporting means watching every scheduled load, fixing a connection when a source system changes, adding a measure when the business starts tracking a new decision, and retiring the reports nobody opens.

BD Emerson’s guide puts maintenance and support for custom software at 15 to 25% of the build each year, and reporting carries the same kind of upkeep. If the reports run in a BI tool, its licenses sit on top. Reporting we build runs on a monthly plan, and your proposal sets what that plan covers. The market’s figures for building and running software at each level are on our page on custom software cost.

If your Monday report takes a morning to put together, book a 15-minute call. Bring the report and a list of the systems it draws from, and we’ll tell you whether reporting work is worth doing or whether the tools you have can already do the job.