ERP Modernisation in Local Government: Why Projects Slip
The failure pattern is consistent across entities of very different sizes, and it is established during procurement rather than during implementation.
Twenty-five publications on financial oversight in the public sector — method, practice and the places where both go wrong. Open to everyone, without registration, permanently.
Showing all 25 publications
The failure pattern is consistent across entities of very different sizes, and it is established during procurement rather than during implementation.
Most of the value from machine learning in local government finance is not in prediction. It is in triage — deciding which of ten thousand transactions a human should look at first.
A balanced budget and a structurally balanced budget are different claims. Telling them apart takes about twenty minutes and four adjustments.
Purchase cards concentrate a disproportionate share of detected public-sector fraud in a small share of spending. The controls that work are unglamorous.
The comparison is only as good as the risk quantification, and the risk quantification is where the analysis is most often engineered toward a conclusion.
A workable policy fits on six pages and answers five questions. Anything longer tends to describe aspirations rather than constraints.
A five-year forecast that nobody believes is worse than no forecast. The credibility problem is usually structural, not statistical.
The standard is settled. The operational problem it exposed — that nobody owned a complete contract inventory — mostly is not.
The label is voluntary, the pricing benefit is modest and contested, and the reporting obligation is real. Issuers should be clear about which of the three they are buying.
Banking supervisors have spent fifteen years formalising model risk management. Most of it transfers to public finance at a fraction of the cost, if you drop the parts built for capital adequacy.
Allocate each risk to the party that can best control it, price it, or absorb it — and accept that some risks satisfy none of those tests.
In a department of three, ideal segregation is arithmetically impossible. Compensating controls are not a lesser substitute — but they only work if they are specified precisely.
The accounting is a close cousin of GASB 87. The hard part is deciding which of four hundred software contracts is actually in scope.
Programme inventories are the easy part. The reason most implementations fade after two cycles is that nothing was allowed to end.
Analytics projects in public finance fail on data quality far more often than on methodology. The remedies are boring, cheap, and rarely funded.
Most disbursement fraud starts with a record, not a payment. The vendor master file is where the control is cheapest and least often applied.
The standard for a public entity is not that the model is interpretable. It is that the decision is defensible to the person it affected.
Investors and rating agencies are asking. The difficulty is that the honest answer for most issuers is a range, and disclosure practice is built for point estimates.
Two sets of statements measuring two different things is the design, not a redundancy. Reading them as alternatives is how officials reach wrong conclusions confidently.
Two months of expenditure is the most-cited target and one of the least-justified. The right level depends on four measurable properties of the entity.
It is a screening tool for selecting samples, not evidence of anything. Treating a conformity test as a finding is the most common misuse.
The transaction team disperses, the contract runs for thirty years, and institutional memory has a half-life of about four.
Most of the delay in a public sector close is waiting, not working. The fix is scheduling, not effort.
Outputs are cheap to count and easy to defend. Outcomes require attribution, and attribution is where most public measurement quietly gives up.
Publishing everything produces a portal nobody uses. Four design decisions determine whether a transparency site informs anyone.
Walk through the platform with your own chart of accounts, or start with the research library. Both routes are free to begin.