Golden Pod Cocoa Processing Ltd: ERP go-live that nobody trusts
Golden Pod Cocoa Processing Ltd is a Takoradi-based processor with 480 staff and revenue of GH₵210 million. Five months ago it went live with the finance, procurement and inventory modules of a mid-tier ERP, replacing a 12-year-old system. The plan was to add payroll and fixed assets in the next quarter.
Since go-live, month-end close has stretched from 5 to 12 working days; a stock count recorded a GH₵1.4 million variance that nobody can explain; and three system reconciliation reports do not agree with the manual stock cards that stores clerks still keep. Twenty-two staff maintain shadow spreadsheets, and the finance manager signs off those spreadsheets rather than the ERP reports. The vendor's implementation team left two weeks after go-live; only one internal accountant attended the report-writing course, and he resigned last month. The project sponsor, the Finance Director, says the problems are teething and wants Phase 2 approved at the next board meeting.
1. As an independent adviser, set out how you would decide whether this is a failed implementation, and what you would do in the next 30 days.
Read the worked answer
Separate three questions. Did the project deliver scope, cost and time (delivery)? Does the software work (technical)? Does the organisation get information it can act on (benefits)? On these facts the modules are running, so delivery and technical performance look broadly acceptable. What has failed is benefits realisation: the ledger is not trusted, so the business has quietly reverted to spreadsheets. That is the finding to report.
Root causes, ranked by evidential strength.
- Data migration was never reconciled. An unexplained GH₵1.4 million variance points to opening stock quantities loaded from legacy records without a physical count and sign-off.
- User acceptance testing was ineffective. The reconciliation reports were never proven against the stock cards before go-live.
- Knowledge transfer was not an exit condition. One trained user, now gone, and the vendor demobilised two weeks after go-live.
- Sponsorship is weak. The Finance Director is declaring success without any defined success measure to test the claim against.
Next 30 days. Freeze Phase 2; put the outstanding hypercare items to the vendor in writing. Run a full physical stock count, reconcile line by line, post adjusting entries with an audit trail, and have internal audit observe. Build a register of every shadow spreadsheet with an owner, purpose, data source and a decision to retire or formally control it. Retrain two named users per process and obtain the vendor's data dictionary and report specifications. Agree three measurable targets with the board: close within seven days by month three; no unexplained variance above GH₵50,000; zero sign-offs from shadow spreadsheets by month four.
2. Advise the board whether to approve Phase 2 (payroll and fixed assets) at the next meeting.
Read the worked answer
Recommendation: defer, do not cancel, and replace the vote with a gate and a re-decision date in the following quarter.
Why deferring beats approving now. Phase 2 is not more of the same. Payroll touches every employee and every statutory filing, and errors are publicly visible within one pay cycle; fixed assets depends on the same clean supplier and purchase-order data that the team currently distrusts. Approving Phase 2 into an untrusted general ledger multiplies the reconciliation workload at exactly the point where capacity is thinnest - one certified accountant has already left.
Gate criteria the board should require before release of funds.
- Close within seven working days for two consecutive cycles.
- Stock variance explained and reconciled.
- At least two trained users per module and a named process owner.
- Shadow spreadsheets retired or formally controlled.
- Hypercare exit signed off, with a support SLA and vendor retention money in place.
- An agreed data migration and reconciliation method for payroll master data.
Risks of deferring, and mitigations. The vendor team may demobilise, so keep them on a paid support retainer and book implementation slots now. Staff morale may slip, so publish the gate criteria openly. There may be cost escalation; obtain a fixed price for Phase 2 valid for six months. There is also a sequencing option: fixed assets is lower risk than payroll and depends on data already live, so it could go first if the board wants visible progress. Payroll should go last, not first, whatever the original plan said.
3. Recommend the governance and benefits arrangements that would prevent a repeat on Phase 2.
Read the worked answer
Ownership and decision rights. Reconfirm a project sponsor at executive level with written accountability for benefits, not just budget. Constitute a steering group (Finance Director as chair, operations, IT, internal audit as observer) meeting monthly with a standing agenda of scope, risk, data quality and benefits. Give the board's audit and risk committee a short quarterly dashboard rather than a narrative.
A benefits register with numbers. Record each promised benefit, its baseline, target and measurement method: days to close, stock-count accuracy percentage, number of shadow spreadsheets, percentage of transactions posted in-system. Post-implementation reviews at three and twelve months, reported to the audit committee.
Change and data control. A change control board with authority to refuse scope, and a rule that no change enters the release plan without a data impact assessment. Name a data owner for each master file (supplier, item, employee) who signs off migration reconciliation. Treat migrated opening balances as a formal opening-balance audit.
Vendor management. Contract discipline is the cheapest control available. Payment milestones tied to acceptance criteria, retention of 10-15% until hypercare exit, a knowledge-transfer obligation with named trainees and materials, and a documented exit/transition plan.
Capability. Two trained users per process, with training written into job descriptions. Consider a part-time business systems analyst rather than relying on one accountant who may leave.
Tone. The single most important change is that the sponsor must stop describing failure as teething. A board that accepts unexplained variances of GH₵1.4 million in one site will accept them in Phase 2.