Hiring guide
Can your team maintain the CFO's financial model? A handover test
By CFO Index · Published
A financial model is not fully handed over because someone has emailed the workbook and presented its outputs. Ask an internal operator to run a normal update from approved source files without the author steering each step. The test below produces evidence of what your team can maintain and what still depends on the fractional CFO.
Agree what maintenance actually includes
Choose the tasks your team will own: import the next period, update a commercial assumption, add a customer or produce a management report. Distinguish those from redesigning the model or changing accounting policies. Name the operator and reviewer before the test. If nobody internally has the capacity or relevant skills, the solution may be an ongoing maintenance service rather than a more elaborate handover document. Put that responsibility into the engagement scope.
Require a small, usable handover pack
ICAEW's spreadsheet principles support audience-appropriate documentation, a clear path from inputs to outputs, review, and version control. For this test, ask for the exact workbook version, a list of input sources, refresh instructions, required permissions and a sample accepted output. Record which actions need a specialist. The exercise below is CFO Index's proposed acceptance method, not an ICAEW certification or an independent audit of the model.
Run the baseline before changing anything
Give the operator a test copy and the source files used for the last accepted period. Keep access within company-approved systems; do not share the author's personal credentials. Ask the operator to recreate the summary using the written instructions. Record any difference from the accepted output and whether it comes from data, software settings or an undocumented manual step. Stop if the starting result cannot be reproduced: otherwise later test differences will be hard to interpret.
Use a change with a result you can calculate separately
Hypothetical isolated cash test: move a €25,000 customer receipt from October to November, with all other movements fixed and no interest or financing feedback. October closing cash should fall by €25,000. November opening cash should also fall by €25,000, while November receipts rise by €25,000, leaving November closing cash unchanged. Have the operator explain each movement. This expected result applies only to the stated assumptions; document any genuine model dependencies that make a different result appropriate.
Test one new row and one new reporting period
A successful input change does not prove that an expanding model works. In the test copy, add a hypothetical €12,000 receipt from a new customer in the existing month and check that the detail and summary both increase by that amount under otherwise fixed assumptions. Then roll the calendar forward using an agreed sample period. Confirm that the new month appears in the output and that historical actuals have not been overwritten. These cases expose different handover gaps: incomplete ranges and unclear period-update instructions.
Log failures instead of letting the author silently fix them
Use these fields for each test: starting version, operator, input change, expected output, observed output, help required, defect owner and retest result. If the author intervenes, note exactly why. Classify the issue as a formula defect, missing instruction, unavailable source, permission problem or training need. A polished demonstration can hide those differences; a written result makes it possible to agree which repair is required before ownership transfers.
Make acceptance conditional on the next person being able to operate it
Agree which failures block use for the intended decision. Retest corrected steps with the receiving operator, retain the evidence and identify the production version separately from the test copy. Record unresolved limitations beside the outputs they affect. Acceptance should confirm the agreed maintenance tasks and the available support route, not claim the workbook is error-free. A material change in scope or model logic may require a new review even after this handover passes.