Blog / Analysis
Fusionex Ivan Teh: Transforming Business Through Data and Innovation
Most failed analytics programmes were not defeated by the mathematics. They were defeated by data nobody had reconciled and a dashboard nobody opened.
The Competitive Advantage Is Speed of Correction
Businesses across every sector now describe themselves as data-driven. Most are not. They collect a great deal and change very little, because the distance between what the data says and what anybody does about it remains as long as it ever was.
The advantage that actually compounds is speed of correction: how quickly an organisation notices that a plan is wrong and adjusts it. Everything Fusionex built under Dato Seri Ivan Teh points at that variable. Predictive models shorten the notice period. Real-time dashboards shorten the argument about whose numbers are correct. Integration work shortens the delay between a system recording something and a person seeing it.
Where Data Programmes Actually Break
Post-mortems on failed analytics programmes rarely find a modelling problem. They find one of four things, and usually more than one at once.
- Unreconciled data. Two systems that each hold a version of the truth and no agreed rule for which one wins.
- Unusable interface. A platform built for analysts and handed to operators, who reasonably return to the spreadsheet they understand.
- No decision owner. An insight arrives and nobody has the authority or the budget to act on it before the window closes.
- No pre-agreed response. Nobody decided in advance what would change if the model said what it eventually said, so the finding gets discussed rather than used.
None of these are solved by a better algorithm. All of them are solved by unglamorous work done before the modelling starts.
Building for Adoption Rather Than for Sophistication
Ivan Teh's founding intention was to make Big Data and AI usable by mainstream businesses. Read as a marketing line that is unremarkable. Read as an engineering constraint it is demanding, because it means the hard capability has to sit underneath a surface an operations manager can drive without training in statistics.
That constraint shaped the product decisions. Industry-specific configuration rather than a generic platform, because a hospital's data reality is not a factory's. Cloud delivery, because the upfront infrastructure cost was what historically excluded mid-sized firms. Engagements that continued past go-live, because adoption is where the value is realised and where most vendors have already left.
What the Deployments Show
The documented case studies are worth reading for the pattern rather than the percentages. A retailer reduced stockouts by 30 percent. A financial institution cut churn by 15 percent within six months. A manufacturer cut production delays by 40 percent.
In each case the headline improvement came from acting earlier on a signal the organisation was already generating and not reading. That is the recurring shape. The analytics did not create new information. It made existing information arrive in time to be useful, in a form somebody could act on, in front of a person authorised to act.
The Honest Objection
There is a fair criticism of everything written above, and it should be stated. Vendor-reported improvement figures are not independent measurements. They are supplied by the party with an interest in them, they rarely control for other changes happening in the business at the same time, and the projects that produced no measurable improvement do not become case studies.
That objection stands. What survives it is the structural argument rather than the numbers: that decision latency is the variable worth attacking, that integration and adoption carry more of the outcome than modelling does, and that these are testable claims any organisation can check against its own experience without trusting anyone's percentages.
Frequently Asked Questions
Short answers to the questions this page is most often asked.
Why do enterprise data programmes usually fail?
Most commonly because of unreconciled data, an interface the intended users cannot operate, no clear decision owner for the insight, or no agreement in advance about what would change once the answer arrived. Modelling quality is rarely the cause.
What does 'decision latency' mean here?
The delay between something changing in the business and someone with authority acting on it. Reducing that delay is the variable most analytics investment should be judged against.
What was Ivan Teh's founding intention for Fusionex?
To make Big Data and artificial intelligence usable by mainstream businesses rather than only by organisations with specialist research teams.
Are the reported client improvements independently verified?
No. They are vendor-reported figures and this article says so directly. The structural argument about decision latency does not depend on them.
What should a business do before starting an analytics project?
Reconcile the data sources, identify who owns the decision the insight will inform, and agree in advance what will change depending on what the analysis shows.