Forecasting from drivers

The forecast is the foundation the whole model stands on. Every number downstream — the profit, the cash flow, the balance sheet, the valuation — is some consequence of the growth and margin assumptions you type at the top, so this module is about the one skill that separates a forecast a stranger will trust from one they quietly dismiss: building assumptions that survive scrutiny. Start with the distinction that governs everything else. An assumption is a number you can defend — you can point to the company's own history, to what comparable businesses do, or to the market it sells into, and say why this number and not another. A hope is a number you want to be true: it has no source, and it is chosen because it makes the answer come out well. The two look identical on the page — both are just a percentage in a blue cell — and the whole craft of forecasting is refusing to let a hope sit where an assumption belongs.

Top-down and bottom-up

There are two ways to forecast a top line. Top-down starts with the size of the market and the company's share of it: 'the Kenyan packaged-foods market is worth KES 400bn and grows 9% a year; Sokoni holds 4% share, rising to 5%.' It is quick, and it is the right tool for a smell test. It is a poor tool for the primary forecast, because the assumed share gain is almost always more generous than anyone would defend if you made them say it out loud.

Bottom-up starts with the operating drivers — the physical machinery of the business. For a retailer like Sokoni that is number of outlets × revenue per outlet, or transactions × average basket size. For a manufacturer it is tonnes × price per tonne. Bottom-up is harder, and that is precisely its value: it forces you to engage with how the business actually makes money, and it makes a fantasy arithmetically obvious. If your driver build says Sokoni needs to double its basket size and open forty stores it has no capex to fund, the top-line story falls apart in front of you. Build bottom-up as the primary forecast; keep top-down beside it as the reconciliation.

Anchor to history, peers, and the market

A defensible growth or margin number is anchored to three reference points, and each tells you something different. History tells you the base rate: what did this company actually grow at, and earn, over the last three to five years? A forecast that departs sharply from history is not forbidden, but it owes you a specific reason — a new format, an acquisition, a genuine step-change. Peers tell you what is normal for the business model: if comparable retailers run a 38-41% gross margin, a 46% assumption is a claim that Sokoni is structurally better than every competitor, and you had better be able to name why. The market tells you the ceiling: a company cannot outgrow its market forever without taking share from named rivals who will not sit still, and it certainly cannot sell more than the market spends.

Scenario toggles: one cell, three worlds

A forecast is never a single number; it is a range, and the professional way to express that range is a scenario switch. You build one cell — the scenario selector — that holds 1, 2, or 3 for base, bull, and bear, and every driver that matters reads from it, so flipping that one selector reshapes the entire model at once: revenue, margin, profit, cash, and valuation all move together to tell a coherent story. The two mechanics you will use are CHOOSE and INDEX/MATCH. With CHOOSE, the growth cell is =CHOOSE($B$1, 10%, 13%, 6%) — it returns the first, second, or third value according to the selector. INDEX/MATCH is the scalable version: lay the scenarios out as a labelled table and pull the live column with =INDEX(growth_row, MATCH($B$1, scenario_labels, 0)), which lets you add a fourth scenario without rewriting a formula. For Sokoni we make the two drivers that carry the answer — revenue growth and gross margin — scenario-switchable and leave the rest of the model reading off them, so the base case is the anchored central expectation while the bull and bear are coherent, internally consistent worlds rather than a single number nudged up or down. This is where the companion /templates/3-statement-financial-model earns its keep: the selector sits on the assumptions tab, and you watch the balance sheet still balance and the cash flow still tie as you flip between worlds.

• Scenario selector — one labelled cell (say B1) holding 1 = Base, 2 = Bull, 3 = Bear. Colour it as an input; it is the only cell you change to reshape the model.
• Revenue growth driver — Base 10% (Sokoni holds a little above the 9% market), Bull 13% (a real share-gain story, funded by store openings), Bear 6% (below market — share loss in a price war). Cell: =CHOOSE($B$1, 10%, 13%, 6%).
• Gross margin driver — Base 40% (in line with history and peers), Bull 42% (mix shift to higher-margin lines, scale buying), Bear 37% (discounting to defend volume). Cell: =CHOOSE($B$1, 40%, 42%, 37%).
• The wiring rule — every downstream formula references the driver cell, never a typed rate. Because revenue, COGS, working capital, and the DCF all trace back to those two switched cells, one selector move flows correctly through all three statements and the valuation.
• The discipline — only drivers you can tell a coherent three-world story about go on the toggle. Tax rate, depreciation policy, and payables days stay fixed across scenarios; toggling everything at once produces noise, not insight.

The two traps that break long-horizon forecasts

Two errors recur in every junior five-year forecast, and both come from extrapolating a good year forever. The first is the operating-leverage trap. Because fixed costs do not rise with revenue, the spreadsheet makes it seem that every extra shilling of sales drops to profit at a very high rate, so margins expand mechanically year after year. In reality operating leverage is far smaller than the model implies: growth demands real incremental spend on stores, staff, marketing, and systems; wages creep with inflation; and competition forces you to hand some of the gain back in price. Margin expansion is legitimate only when you can name the operational lever delivering it — a mix shift, a supplier renegotiation at scale, SG&A growing at half the pace of revenue. If you cannot write that sentence, hold the margin flat. The second is the mean-reversion trap: a high growth rate held flat across the whole horizon. Every fast-growing business slows as it gets larger and the market saturates; growth fades toward the market rate. A forecast that grows 18% in year one and still 18% in year five is almost always a number that was typed, not built. Build the fade in explicitly.