Capstone — the complete worked model

For eleven modules you have built one thing, a piece at a time. A structured assumptions tab. An income statement driven by real revenue, not a lone growth cell. Working-capital, capex and depreciation schedules. A balance sheet and a cash flow that link back to it. A debt and interest schedule. A DCF hanging off the free cash flow, and the sensitivity and error-proofing that let you hand it over without fear. This module assembles those pieces into the deliverable itself: a complete, balancing three-statement model on Sokoni Ltd, and the one-page investment case it produces. From a blank sheet to a number you can defend in a room — that is the whole arc, and by the end of this module you will have walked it end to end.

The whole build, in order

The order is not arbitrary. Each step depends on the one before it, and skipping ahead is how models end up broken in ways that are painful to unpick. You cannot balance a balance sheet before the cash flow exists to feed it. You cannot compute interest before you have a debt schedule, and you cannot finish the income statement until that interest links back in. Build in the sequence below, and confirm the balance check before you add anything clever. A model that balances at step six can absorb a debt schedule, a DCF and a sensitivity table without collapsing; a model that does not balance at step six only gets harder to fix with every row you add.

1. Assumptions tab. Put every input in one place — growth, gross margin, working-capital days, capex, tax rate, WACC, terminal growth — each one labelled, coloured blue, and sourced. From here on, nothing downstream is a hardcoded number typed into a formula.
2. Income statement. Drive revenue from a real driver, not a lone growth cell; take it down through gross profit at the stated margin, operating costs and EBIT. Leave interest as a link you will wire up once the debt schedule exists.
3. Supporting schedules. Turn the days ratios into receivables, inventory and payables, and turn capex into a property-plant-and-equipment roll-forward that produces depreciation. These feed both the balance sheet and the cash flow.
4. Balance sheet. Lay out assets (cash, receivables, inventory, net PP&E) against liabilities and equity (payables, debt, share capital, and retained earnings that rolls in net income less dividends).
5. Cash flow statement. Operating (net income plus D&A less the change in working capital), investing (capex), financing (debt movements and dividends). The closing cash balance links straight back to the balance sheet.
6. Balance check. One cell: total assets minus total liabilities minus equity, which must equal zero in every year. If it is not zero, stop here — do not build another row until it is.
7. Debt and interest schedule. Opening debt, draws and repayments, closing debt, and interest on the balance — then wire that interest back into the income statement, closing the loop (and the circularity it creates, which you resolve with iterative calculation or a clean cash sweep).
8. DCF. Discount the model's own free cash flow — NOPAT plus D&A less capex less the change in working capital — at WACC, add a cross-checked terminal value, and bridge enterprise value to equity value and a per-share number.
9. Sensitivity. Build a two-way table on the two inputs that actually move the answer — typically WACC against terminal growth — and, if you can, a base, bull and bear scenario toggle.
10. Audit and error-proofing. A visible checks row, no hardcodes buried in formulas, consistent units and colours, and a final read of the whole model through a stranger's eyes.

Step six deserves a sentence of its own, because it is the moment the model grades itself. The balance check is a single cell: total assets minus total liabilities minus equity, which must equal zero in every year. When Sokoni's balance sheet ties to the cent across all five forecast years, the three statements are genuinely linked — every shilling of profit has found its way onto the balance sheet, and every movement of cash is accounted for. When the check is not zero, some cash flow is not landing where it should, and no amount of formatting hides it. The discipline is simple and non-negotiable: if it does not balance, stop, and do not build another row until it does. Never force the check to zero with a hardcoded plug — a plug does not fix the error, it conceals it, which is strictly worse.

From a working model to a one-page case

A model nobody reads is worth nothing, however elegant. The point of the whole exercise is a decision, and the decision lives on one page. Sokoni's finished model runs to thousands of linked cells, but the output that matters is small: an enterprise value of about KES 309,790 thousand, less net debt of about 200,000, giving equity value of about 109,790 thousand, or roughly KES 1.10 per share. And crucially, not exactly KES 1.10 — a range. The honest deliverable is a value of roughly KES 1.10 a share within a plausible band, say 0.85 to 1.45, driven mostly by the two inputs that genuinely move the answer: the discount rate and the terminal growth rate. Terminal value is about 64% of that enterprise value, so the reader should know that most of what they would be buying sits beyond the explicit five-year forecast — a fact to disclose, not bury.

• The recommendation and the range at the very top — for Sokoni, a value of roughly KES 1.10 a share within a band of about 0.85 to 1.45, and a clear buy, hold or pass against the market price.
• The two-line thesis: in one breath, why the business is worth this, and what would have to be true for it not to be.
• The handful of numbers that matter: enterprise value, net debt, equity value and value per share; the base-year and terminal growth rates; the WACC; and terminal value as a share of enterprise value.
• The one sensitivity table — WACC against terminal growth — so the reader sees the range, not just the point.
• The three or four assumptions carrying the valuation, stated plainly with their source, so a reader can challenge them directly.
• The swing factors: what you are watching, and what would change the recommendation.

What an employer actually sees

An experienced reviewer can tell an amateur model from a professional one in about thirty seconds, and they look for tells. The amateur tells: hardcoded numbers welded into formulas; no assumptions tab, or one that nothing actually references; a balance sheet that does not balance, or that balances only through a plug that hides the break; every cell the same colour, so you cannot tell an input from a calculation; a single revenue growth cell standing in for a real driver; unresolved circular-reference and #REF! warnings; tabs still named Sheet1 and Sheet2; and a valuation reported as one confident number with no range. The professional tells are the mirror image: a clean assumptions tab with blue inputs; three statements that link and balance to the cent in every year; a visible checks row that reads TRUE all the way across; supporting schedules for working capital, fixed assets and debt; a DCF whose every cash-flow line traces back to the statements; a sensitivity table; consistent units and sign conventions; and a one-page output a busy person can absorb in two minutes. The difference is rarely cleverness. It is discipline.

Presenting and defending the number

When you present, lead with the answer and the range, never the build. Nobody wants a tour of your tabs; they want the decision. Something like: we think Sokoni is worth roughly KES 1.10 a share, in a range of about 0.85 to 1.45, and the swing factor is the terminal growth rate and the discount rate. Then the two-line thesis — why the business is worth this, and what would have to be true for it not to be. Expect the room to attack the two places value hides: the terminal value, which is 64% of enterprise value here, and the WACC. Have the sensitivity table open, and know cold which single assumption carries any disagreement with the market price. Do not defend the point estimate — defend the range and the direction of each input. And if someone changes an assumption live, your model should move correctly in front of them; that is the entire payoff of building it with one source of truth. Know your three biggest assumptions and where each came from, and concede the model's limits before anyone asks — it reads as confidence, not weakness.

The honest limits, and where to go next

Be clear-eyed about what you have built. A model does not predict the future; it makes a particular view of the future explicit, consistent and open to challenge. Garbage in, garbage out — a beautifully engineered model resting on bad assumptions is merely precisely wrong. Sokoni's terminal value, 64% of its enterprise value, rests on a perpetual growth rate and a discount rate that nobody can verify. The model cannot price a change of management, a new entrant, a currency shock, a regulatory shift, or a fraud in the accounts. Treat the output as a disciplined argument and a range, never a fact, and stay more sceptical of your own number than anyone else in the room. That scepticism is not a weakness of the craft; it is the craft.