If you run a growing South African business, you already know the month-end ritual. Someone exports from Xero or Sage. Someone else pastes it into last month’s workbook. A tab breaks. VAT does not quite match the accountant’s pack. The owner gets a PDF on day eight, asks three questions, and those answers take another afternoon.
That is not a data strategy problem. It is a reporting problem, and you can fix it without hiring a team or buying an enterprise stack.
The job is a pack you can trust on Monday
Xero’s 2026 State of South African Small Business report asked 427 businesses how they actually run the numbers. Most owners are already watching the money: about two thirds track cash flow, a similar share check the bank daily or weekly, and more than six in ten had cash-flow trouble in the past year. They are not short of figures. They are short of a pack they can trust when they sit down on Monday.
A useful reporting layer is boring in the best way. “Revenue” and “overdue” mean the same thing every time. Cash and debtors show up daily, the P&L when the books have been touched, not every five minutes. And you and the accountant are looking at the same page, not reconciling a side workbook against Sage after the fact.
If a dashboard does not change what you do that week (chase a debtor, delay a hire, leave a supplier) it is decoration.
Excel is fine. Excel as the system of record is not.
I am not anti-spreadsheet. Excel is the right place for a one-off. It becomes a liability when it is how the business remembers what happened last month.
The pattern I keep seeing around Cape Town is simple. Xero or Sage holds the books. Shopify, HubSpot or a job-card system holds the work. A “master” file stitches them together with VLOOKUPs and one person who knows which tab is current. That person goes on leave. The pack is late. Trust drops.
Microsoft retired the native Xero connector for Power BI years ago. Sage 200 and Evolution often sit on SQL, which is easier to read, but only if someone knows the tables and is not writing over the ERP. “Just connect it” is the sales pitch. The real work is mapping the chart of accounts once, cleaning customers and items, and a refresh that does not hammer the API until it falls over and everyone goes back to CSV.
You do not have to become a data engineer. You do need someone who has done this on an SME budget, not a six-month warehouse programme.
What I actually build
Keep the architecture dull on purpose.
Xero and Sage stay the books. The dashboard only reads. Nobody posts journals from Power BI. Your accountant’s close does not change. They just stop being a copy-paste factory.
For Xero, do not hang a live dashboard off the API all day. You will get rate-limited, the refresh will fail, and people will go back to CSV. Land invoices, payments, contacts and the P&L in a small database (PostgreSQL is enough for most SMEs) or a warehouse you already pay for. Then Power BI or Looker Studio reads the copy. For Sage Evolution or 200, a read-only SQL connection is often enough. Stage it if you have more than one company and want a single P&L.
Then build a small model, not a 40-page pack. Cash and the next 30 days of expected receipts, from unpaid invoices, not from hope. Revenue and gross margin versus last year and versus budget, in rand. Debtor days, and the ten invoices that actually matter this week. How expenses sit: salaries, cost of sales, the rest, so a spike is obvious. If you also run jobs, stock or a pipeline, add one operational view after finance is trusted. Mixing a messy CRM into a half-trusted P&L is how dashboards die.
Refresh to match the decision. Overnight for cash and debtors is fine. P&L after the books have been touched. Shopify daily. “Real-time” is a vendor word. For most local SMEs, yesterday’s close of business on a phone beats a live feed nobody opens.
Pick the tool for the team, not the brochure
I am not tied to a vendor. Power BI is the default if you already pay for Microsoft 365 and finance lives in Excel. Watch the licence count; do not buy Premium because a demo said AI. Looker Studio is the default if the business is Google-heavy and the first dashboard is sales plus a simple P&L. Cheap to start, awkward when the model grows. Metabase is good if you already have PostgreSQL and a technical owner who is happy asking questions rather than pixel-perfect packs. Tableau is rarely the first buy here unless you inherited it.
The charting layer is the cheap part. Mapping Xero accounts onto the way you actually talk about the business is the expensive part. Do that once.
A short POPIA note
A management pack is mostly company financials. Add customer names, ID numbers, staff detail or a HubSpot “customer 360” and you are processing personal information.
You do not need a 40-page policy to start. You do need a named place the data lives, not five workbooks on laptops. Limit access to people who need it. The whole company does not need the debtor listing with cellphone numbers. And do not dump a full Xero contact export into an open Google Sheet “just for the dashboard.”
The Information Regulator is no longer theoretical. Administrative fines under POPIA go up to R10 million. For an SME the practical risk is usually a leaked spreadsheet and an unhappy client, not a headline. Build access in from day one. It is cheaper than retrofitting.
How you know it worked
Skip vanity metrics. After a month with a live pack, the owner should be able to answer “how is cash and margin?” without waiting on a person. Finance should have stopped rebuilding the same six tabs. The accountant’s numbers and the dashboard should match, or the gap should be a documented timing issue, not a mystery. And at least one real decision should have moved because the number was in front of you in time.
If none of that is true, kill the dashboard or rebuild it around the actual question.
Two weeks, not two quarters
A 25-person company does not start with Snowflake, dbt and a semantic layer. Pick one source of truth for money, Xero or Sage, not both mashed badly. Write down the five numbers the owner asks for every week. Stage that data, define the numbers once, put them on one page. Sit with finance and check the first month against the signed-off pack. Only then add Shopify, HubSpot or jobs.
That is the same sequence I use with clients: a quick scan of the current setup, the three opportunities that would pay off first, and a next-step plan you can run with or without me.
If the monthly pack is still a shuffle, book a free 30-minute strategy session. Video or phone, no pitch. You leave with a read on the reporting you have, where automation would actually help, and an honest view of whether I am the right person to build it.