The quality of our analytical thinking
Framing the question properly, choosing the method it needs, and testing the answer before anyone relies on it.
About SA Informatics
An analytics company built around analytical thinking, and the ability to turn it into products.
SA Informatics is an African data analytics and data science firm. We combine data science, analytics engineering and product development around one analytical problem at a time.
Why we exist
Most organisations don't have a shortage of data. They have a shortage of analysis they can rely on.
When an important question has to be answered, the work is still manual, fragmented and dependent on one or two people.
The same questions are rebuilt from scratch. Models are built and never reach a decision. Useful findings disappear into a slide deck when the engagement ends.
What we're building. Known for two things.
Consulting lets us take on new and difficult problems. Data science gives the depth to solve them rigorously. Product engineering turns the methods that recur into capabilities organisations keep.
Turning that thinking into products
Some problems stay bespoke. Others show patterns that justify a product: a method that reruns on new data, without being rebuilt each time.
How engagements are led. By the people doing the work.
SA Informatics is founder-led. The founder is accountable for the quality of every analysis we deliver. You work directly with the people building it, from the first conversation to the final handover, with no account-management layer in between.
The expertise behind the work
Applied AI applications
Applications that read enquiries and documents, apply business rules and product knowledge, and prepare the work for review.
See the applicationsForecasting and demand planning
Outlooks built on a business's own history, compared against simpler methods on unseen months, with ranges and scenarios.
See a forecast evaluationMarket and competitor analysis
Share tracking across competitors and periods, with every movement traced to who drove it.
See the vehicle-market analysisPerformance analysis
Price separated from volume, contributions that reconcile to the total, and unusual movements checked before they are acted on.
See the retail analysisAnalytics engineering
Pipelines pinned to the data release they came from, so every result reproduces and every refresh runs the same way.
See how a result reproduces
How we work
The problems we're built for. Complex, and recurring.
When a problem is simple, off-the-shelf tools are usually right. When it is complex but genuinely once-off, a focused study may be all you need. When it is both complex and recurring, that is where we do our best work.
- The data is difficult, fragmented or incomplete
- Routine reporting doesn't answer the question
- The method matters: comparison or prediction is needed
- You'll need the answer again as new data arrives
- Generic software doesn't capture the problem
- Rebuilding the analysis by hand each time is slow or risky
Answer needed once
Answer needed repeatedly →
Routine question
Answer needed once
An analyst and a spreadsheet
A one-off question with clean data rarely needs outside help.
Answer needed repeatedly
Your BI tool or off-the-shelf software
A standard report that recurs is what reporting tools are built for.
Complex question
Answer needed once
A focused analytical study
A hard question you'll ask once needs rigour, not a system. We do these too.
Answer needed repeatedly
Our strongest fit
A hard question you'll keep asking: we answer it, then build the method into something your team keeps using.
How an engagement runs. The stages are fixed; how many a problem needs is not.
Some engagements end in an explanation, some in a forecast. Only the ones that recur become a system.
- 1
Frame
We agree the decision the analysis serves, what counts as an answer, and whether the question will recur.
- 2
Structure
We assemble the data, reconcile definitions across sources, validate inputs and record where each value came from.
- 3
Analyse or model
Descriptive, diagnostic or predictive methods, as the question requires, and only the ones it requires.
- 4
Communicate
You get the answer with its uncertainty and assumptions, in a form built for the decision.
- 5
OperationaliseWhere justified
If the question recurs and the value justifies it, we build the method into a system that reruns on new data.
Analytics engineering happens at the Structure stage, as part of the work rather than a separate project. Product engineering happens at the last stage, when a recurring question justifies it.
What we believe about analytical work.
Reporting is not analytics.
A dashboard shows what happened. It doesn't explain why, what is likely next, or which factors matter.
A model has to earn its place.
If it can't be validated, explained or used in a real decision, its sophistication is beside the point.
The question comes before the technology.
AI, machine learning, a data lake or a dashboard may be part of the answer. They are not the starting point.
Repeatable analysis should become repeatable capability.
If an important question is answered every month, rebuilding the analysis by hand each time is not sustainable.
Important results should be defensible.
You should be able to see what the data represents, what was assumed and how it was validated. False precision damages decisions.
What you can expect from our work.
- A clear answer to the question you asked
- Methods matched to the question, simplest first
- Forecasts proven on months they haven't seen
- A range to plan around, not just a number
- Every figure traceable to its source
- Work your team can rerun without us
Where AI fits
We use AI in two ways: as methods within data science where a question needs them, and in applications that take on work your team repeats, such as turning enquiries into quotations. Either way we start from the task and the result, and your team stays in charge of what goes out.
- Based in
- South Africa
- Billing
- South African rand
- Data handling
- In line with POPIA; inside your environment where data can't leave it
Tell us the question you need to keep answering.
Describe the question, the data you have, and whether you'll need the answer again. We'll tell you whether we're the right fit.