Why we asked
At DataFest Johannesburg in August, we asked leaders how quickly they could get a trusted answer to a business question. Only one in five could do it on the same day.
At DataFest Cape Town, we asked something simpler. Everyone who stopped by our stand got to make one data wish. They came from banks, insurers, development finance institutions, government departments, retailers and consumer brands. They work in the roles that sit between the source systems and the numbers that reach the CFO's desk: analysts, BI managers, heads of data, risk managers and finance system owners.
We read their wishes as a message to finance leadership. These are the people who prepare, reconcile and defend the numbers you sign off, and what they wish for shows where the friction sits in your own reporting chain.
Very few of them asked for AI. Most asked for things a CFO would recognise immediately: one place where the numbers live, data that can be trusted on first read, and reports that are current without someone chasing them.
At a glance
Five wishes that kept coming up
1
One version of the truth
(the most common wish, from roughly one in five people)
2
Numbers you can trust on first read
3
Reports that are current without chasing
4
Numbers that don't need defending
5
Shared ownership of data
What was missing: Only a handful of wishes mentioned AI, and those asked for AI that is reliable and doesn't add complexity.
One version of the truth
What they wished for
- "To have it in one place and trusted."
- "One dashboard that connects to the many data sources to produce one metrics report."
- "An integrated, centralised data environment in our department."
- "A single source of reliable, real-time data that moves us from reporting on what has happened to predicting what could happen."
What it costs finance
When the same number exists in three places, month-end turns into reconciling the reconciliations. Your team spends its best hours agreeing which figure is right before anyone can say what it means, and exco discussions open with a debate about whose number is correct.
Our view
Most organisations have tried to fix this with a platform, such as a warehouse, a data lake or a new BI tool. The platform usually works. What breaks is agreement on definitions, so "revenue" or "arrears" still means something slightly different in each team that touches it. A single source of truth comes from agreeing who owns each definition, and holding that line when a business unit wants its own version. The technology follows that decision. It can't make the decision for you.
Numbers you can trust on first read
What they wished for
- "Clean, curated and trustworthy data to make informed decisions."
- "I wish all data would be fully filled out with no NULL values."
- "I wish with every bone in my body that data products were properly documented."
- "Data to self-tag and classify itself, so lineage is like a DNA test."
What it costs finance
Every number nobody has checked is a risk that someone in your team absorbs by hand. Analysts rebuild checks in spreadsheets, quietly fix the same errors every month, and carry the knowledge of where the data is weak in their heads. When they leave, that control leaves with them.
Our view
Data quality is usually run as a clean-up project. It lasts when the checks are built into the process where the data is created, with clear lineage from source to report, so anyone can trace a figure back without asking the person who built it. In regulated environments, that traceability is also what auditors want to see. Automation helps because the checks then run every time, in the same way.
Reports that are current without chasing
What they wished for
- "Not to have to follow up with anyone to update their sheets manually."
- "Not having to pull a report with outdated information again."
- "Real-time data for reporting and quick decision-making."
- "That all data is live, open and actually usable, not trapped in a PDF."
What it costs finance
Behind many on-time reports sits someone chasing inputs by email and pasting them into a template. The report arrives, but it describes a position that has already moved. The cost shows up as decisions made on last month's picture, and as skilled people spending their week on collection.
Our view
Self-service analytics gets a lot of attention, and it only pays off once the underlying data refreshes reliably on its own. The order matters. First automate the collection and preparation steps people do by hand, then give business users access to data that is already current. Doing it the other way round gives more people faster access to stale numbers.
Numbers that don't need defending
What they wished for
- "Never having to be called into a meeting to explain a number."
- "A dashboard that doesn't need explaining."
- "Data that's information, not just data."
- "Not being told your data is wrong when there is no proper context. I want to be able to solve issues before they happen."
What it costs finance
Every exco question about where a number came from costs time twice: once in the meeting, and again when the team goes back to rebuild the explanation. It also weakens confidence in the next number, even when that one is right.
Our view
Reports that need explaining usually have one of two problems. Either the logic behind the number isn't visible to the person reading it, or the report was designed around the data that was available instead of the decision being made. Both are design problems. When reports are built with the people who use them, and the logic is shown in plain language next to the number, most follow-up questions go away.
Shared ownership of data
What they wished for
- "Data ownership and stewardship."
- "Collaboration instead of resistance from business partners. View things as 'ours', not 'mine' or 'yours'."
- "Data is not a scapegoat for organisational issues."
- "For everyone to be on the same page with their different expertise."
What it costs finance
When nobody clearly owns the data, problems land with whoever is closest to the report, usually finance or the data team. Issues get escalated instead of fixed, and the same arguments come back every quarter.
Our view
When IT owns the data, it tends to stay a technical problem. The organisations we see making progress give ownership to the business people who use the numbers and are accountable for them, with the data team supporting them. That is the thinking behind our business-led approach: the people who own the process shape the solution, and they keep running it after go-live.
Very few people wished for AI
The few who did asked for:
- "a reliable AI agent"
- "AI-built systems not resulting in increased complexity"
That matches what we see inside finance and risk functions. Interest in AI is high, and the teams closest to the data know it runs on the same foundations as every report:
Four foundations
One trusted source
(Wish 1)
Clear lineage
(Wish 2)
Current data
(Wish 3)
Clear ownership
(Wish 5)
Where those are in place, AI can add value quickly. Where they aren't, AI adds another layer that someone has to explain.