Most mid-market companies need business intelligence and cannot justify a BI team. A full team, meaning a data engineer, someone to model the data, a report developer and an administrator, is a real payroll commitment for a business with a few hundred employees. So reporting ends up with whoever is best at Excel, and the cycle of manual exports and disputed numbers continues.
You do not have to choose between that and a full team. The work a BI team does can be split into parts, and most of the parts can be bought instead of hired.
What a BI team actually does
Five jobs get lumped together under BI. Someone moves data out of your source systems on a schedule, which is data engineering. Someone turns raw tables into clean, consistent definitions of the things you measure, which is modeling. Someone designs the dashboards and reports people use. Someone keeps the platform running, manages access and fixes what breaks. And someone in the business decides what matters and acts on the answers.
The first four are technical and can be done by a partner or a managed platform. The fifth cannot be outsourced, and it is the one that decides whether anything works.
Four ways to cover the work
Hire one or two generalists. A single capable analytics engineer can cover a surprising amount in a mid-market business, but you are exposed when they leave, and it is hard to hire someone who is good at all four technical jobs.
Use a consultancy for a project. This is a good way to build the foundation, the data layer and the first set of reports. The risk is the handover: if nobody inside the company understands what was built, it decays.
Use a managed BI platform, where a partner builds the reports, hosts the system and keeps it running, and your people open dashboards. This is how Horizon works. You give up the ability to build anything yourself in exchange for not needing to.
Do it yourself with a self-service tool. This works when someone in finance or operations genuinely has the time and the appetite. It fails when it is a side project for someone with a full-time job.
What you still have to own
Whichever route you take, four things stay with you. The first is deciding what matters: the ten or so numbers you would want to see every week. The second is agreeing on what each one means, because finance and operations often count the same thing differently. We wrote about how to run that conversation. The third is granting access to the source systems, ideally read-only. The fourth is having someone who acts on what the numbers show.
None of those take a team. They take a few hours of a senior person's attention at the right moments.
How to start small
Do not begin with a platform decision. Begin with one decision your business makes badly because the numbers are late or disputed. Maybe it is cash forecasting, production scheduling or sales pipeline review. Build the one set of reports that serves that decision, on top of a clean read-only copy of the data it needs. When it works, add the next one.
This approach also tells you quickly whether a partner or tool is a good fit, before you have committed to anything large.
What a sensible first 90 days looks like
In the first month, choose the decision you want to improve, agree the handful of numbers that serve it, and get read-only access to the source data. Nothing is built yet, and that is the point: this month is where most projects quietly succeed or fail.
In the second month, build the data layer and the first set of reports for that one decision, and let a few real users use them. Expect to find data problems. Finding them now, with a small audience, is cheap.
In the third month, fix what the users found, add the checks that will tell you when data goes stale, and decide what comes next based on what people actually asked for, not on a roadmap written in advance.
By the end you should have one set of reports people use and trust, a clear owner for each number, and an honest view of whether your chosen route is working.
What to budget for
Whatever route you choose, budget for four things, not only the software: building the data layer, building the reports, running and maintaining it afterward, and your own people's time to define the numbers and review the results. The last one is the item most often left out, and the one that most often decides how it goes.
Warning signs
Be cautious of any plan that starts with buying software before anyone has defined the numbers, any proposal that cannot say who maintains the reports after launch, and any arrangement where only one person understands how it works. Each of those turns into the same problem later: a dashboard nobody trusts.
If you want business intelligence but do not want to build a team to get it, reach out. We can scope what a managed setup would look like for your business, or help you build the foundation and hand it over properly.
Have a question about this topic? We work with businesses worldwide.
Get in Touch