Investor Model Readiness Checklist

Diligence is not a review of your model. It is a search for the reason not to proceed, conducted by someone who has read several hundred models and knows where the bodies are usually buried.

This is not a list of modelling best practices. It is the sequence a diligence process actually follows, from the first file open to the point where somebody decides the numbers can be relied on.

The order matters more than the contents. Checks 1 to 5 happen in the first few minutes, and a failure there colours everything that follows, because an investor who catches an error early stops assuming competence and starts hunting. Most models that lose a round do not lose it on the forecast. They lose it on a reconciliation break found in minute four.

Work through it on your own model before someone else does. Where a check fails, fix it or prepare the answer, because you will be asked.

23 checks6 stagesFree, no sign-up

Contents

  1. 01The first five minutes1–5
  2. 02Structural integrity6–11
  3. 03The revenue build12–15
  4. 04Costs and margin16–18
  5. 05Cash and working capital19–21
  6. 06Structure and survival22–23
01Minutes 1 to 5

The first five minutes

Before any judgement is formed about your business, the file gets opened and four or five things get checked mechanically. These are cheap to run and devastating to fail, because they test whether anyone on your side has checked the model at all.

01

The file opens and recalculates clean

What they do
Opens the workbook, presses Ctrl+Alt+F9 to force a full recalculation, and watches what changes.
Passes if
Nothing moves. No error values appear, no prompts about broken links, no circular reference warning in the status bar.
Fails if
Numbers shift on recalculation, which means the file you sent was showing stale values. It is the fastest possible way to signal that the model is not under control, and it happens more often than you would think.
02

Historicals tie to the accounts

What they do
Takes revenue, EBITDA and net income for the last two closed years and compares them line by line against your filed or management accounts.
Passes if
Exact agreement, or a visible reconciliation row explaining each difference.
Fails if
A variance of any size with no explanation. The size does not matter much: a $40k unexplained difference and a $4m one both say the model was not built from the accounts. Every projected number descends from these, so a break here invalidates the rest.
03

Closing cash ties to the bank

What they do
Takes the cash balance at the last actual period end and compares it to your bank statement or balance sheet.
Passes if
It matches, and the cash flow statement walks from opening to closing without a plug.
Fails if
A balancing figure labelled "other", "adjustment" or left unlabelled. A plug in the cash flow is the single most common reason a model gets sent back, because it means the three statements do not genuinely articulate.
04

The balance sheet balances in every period

What they do
Finds the check row, or builds one: total assets less total liabilities and equity, across every column including the forecast.
Passes if
Zero in every period, under every scenario, not just the base case.
Fails if
A non-zero in any period, or a check row that only covers the historical columns. Also fails if the check exists but is formatted white on white, which reads as concealment whether or not it was meant that way.
05

An error-check sheet exists and passes

What they do
Looks for a dedicated checks sheet, and tests whether it is live by deliberately breaking an input.
Passes if
A checks sheet aggregating balance-sheet integrity, cash reconciliation, coverage tests and scenario consistency, with a single master flag that visibly turns red when something breaks.
Fails if
No checks sheet, or one that is hardcoded to TRUE. A dead check is worse than no check: it tells the reader the model presents assurance it does not actually have.
02Minutes 5 to 25

Structural integrity

Having established the model ties to reality, the next question is whether it can be trusted to keep doing so when an assumption changes. This is where a model that was edited under deadline pressure usually gives itself away.

06

Inputs are separated from calculations

What they do
Looks for a dedicated inputs sheet, and checks whether changing an assumption there actually flows through.
Passes if
Every assumption lives in one place, is colour-coded distinctly from formulas, and drives the model from there.
Fails if
Assumptions scattered across calculation sheets, so that changing a growth rate requires knowing which of nine cells to edit. This guarantees the model cannot be driven in a live meeting, which is the whole point of bringing it.
07

Formulas are consistent across every projection row

What they do
Selects a projection row, presses Ctrl+\ (or uses Go To Special, Row differences) to highlight any cell whose formula differs from its neighbours.
Passes if
A clean row. One formula, copied across, with any deliberate break explained in an adjacent note.
Fails if
A single cell edited mid-row. This is the commonest silent error in financial modelling, it is nearly invisible to the eye, and it is trivially detectable by anyone who knows the shortcut. Assume it will be found.
08

No hardcoded numbers inside formulas

What they do
Searches formulas for embedded constants, typically anything beyond a 0, 1 or 12.
Passes if
Every number traces to an input cell. A tax rate appears as a reference, not as *0.29 inside a calculation.
Fails if
Constants buried in formulas. The practical damage is that the model cannot be re-run on a different assumption without someone hunting through formulas, and the reputational damage is that it suggests numbers were forced to a desired answer.
09

Circularity is controlled, not switched on and hoped

What they do
Checks whether iterative calculation is enabled, and if so, whether the circularity is deliberate.
Passes if
Either no circularity, or a genuine one (interest on average balances, debt sizing, interest during construction) resolved with a documented breaker and a convergence check.
Fails if
Iteration enabled with no explanation, or a circularity the builder did not know was there. An accidental circular reference silently produces whatever value the last iteration happened to reach.
10

No external links or broken references

What they do
Checks Data, Edit Links, and searches for #REF! across the workbook.
Passes if
No external workbook links, no #REF! anywhere.
Fails if
Links to a file on somebody’s desktop, or to a workbook that no longer exists. Beyond the fragility, it means the model you sent is not self-contained and the recipient cannot reproduce your numbers.
11

Fragile and volatile functions are avoided

What they do
Searches for OFFSET, INDIRECT, and for VLOOKUP with an approximate-match final argument.
Passes if
Structural references use INDEX/MATCH or XLOOKUP with exact match. Volatile functions appear rarely and deliberately.
Fails if
INDIRECT used to assemble references from text, or VLOOKUP defaulting to approximate match, which returns a plausible but wrong row the moment the source is re-sorted. Both are the kind of error that produces a confident wrong answer rather than a visible break.
03The first substantive conversation

The revenue build

Now the commercial interrogation starts. An investor is testing whether the top line describes a business or merely describes a curve, and this is where most models are exposed within a few questions.

12

Revenue builds from operational drivers

What they do
Traces the revenue line back to its inputs and asks what physically produces the growth.
Passes if
Revenue resolves into quantity and price: customers acquired, units sold, capacity utilised, seats retained, average transaction value. Each driver is something your operating team would recognise and could be held to.
Fails if
A growth percentage applied to last year. This is identified in minutes and it is expensive, because the follow-up question is what the percentage is based on, and there is rarely an answer that survives.
13

Volume reconciles to capacity

What they do
Compares projected volume against the physical or operational capacity to deliver it.
Passes if
A visible capacity constraint, with the capex or headcount required to lift it modelled alongside the volume that needs it.
Fails if
Volume growing past what the current asset base, headcount or plant can produce, with no corresponding investment. It is a fast way to demonstrate the model was built from the top line down rather than from operations up.
14

Price assumptions reconcile to realised prices

What they do
Compares the projected average selling price against what you actually realised historically, net of discounts.
Passes if
Projected price starts from realised price, and any increase is attributed to something specific: a mix shift, a contractual escalator, a product change.
Fails if
A price step-up at the start of the forecast with no cause, which usually turns out to be list price rather than realised price. Investors test this by dividing historical revenue by historical volume, and the gap is immediately visible.
15

Retention and cohort logic is internally consistent

What they do
For any recurring-revenue business, checks whether churn in the model reproduces the retention you report elsewhere.
Passes if
Churn applied at the cohort level, consistently across the forecast, reconciling to the retention figures in your deck and your reporting.
Fails if
Retention improving through the forecast with no intervention driving it, or a churn rate in the model that differs from the one in the pitch deck. The second is particularly damaging because it suggests one of the two was constructed for its audience.
04Alongside the revenue discussion

Costs and margin

The cost side gets less attention in most models and correspondingly more scrutiny in diligence, because it is where optimism hides most comfortably.

16

Fixed and variable costs behave correctly under a volume change

What they do
Halves the volume driver and watches the cost lines respond.
Passes if
Variable costs fall proportionally. Fixed costs hold. The gross margin percentage moves in the direction the cost structure implies.
Fails if
Every cost line scaling with revenue, which means the model has no operating leverage and no fixed base, and the downside case will therefore be meaningless. This is the single most common defect in a downside scenario.
17

Headcount build ties to payroll cost

What they do
Divides projected staff cost by projected headcount to get an implied average salary, and tracks it across the forecast.
Passes if
An implied average consistent with what you actually pay, moving only with stated wage inflation and any deliberate seniority shift.
Fails if
An implied average that drifts downward as the company scales, which is the signature of a payroll line grown by a percentage while headcount was grown separately. The two were never connected.
18

Margin improvement has a named cause

What they do
Plots gross and EBITDA margin across the whole forecast and asks what drives each inflection.
Passes if
Margin expansion attributable to something identifiable: scale against a fixed base, a mix shift toward a better product, a specific procurement change, a named efficiency programme with its cost included.
Fails if
Margin rising smoothly year on year with no mechanism. This is the most frequently challenged line in any diligence process, and "operating leverage" without quantification is not an answer.
05Where deals are usually repriced

Cash and working capital

Profit is an opinion and cash is a fact, and the gap between them is where valuations get adjusted. Working capital receives the most forensic attention of anything in the model.

19

Working capital days reconcile to history

What they do
Calculates debtor days, creditor days and inventory days for each historical year, then for each forecast year, and compares the two series.
Passes if
Forecast days sitting at or near historical actuals, with any improvement tied to a specific stated action.
Fails if
Debtor days quietly improving from 74 to 45 across the forecast with no collections programme behind it. This single assumption often carries more cash than the entire profit improvement in the model, and it is the first thing a diligence provider recalculates.
20

Tax is computed, not assumed

What they do
Checks whether the tax line is a rate applied to accounting profit, and whether cash tax differs from the charge.
Passes if
Tax built from taxable profit, with losses carried forward where they exist, the distinction between the charge and cash paid visible, and local rules reflected rather than a generic rate.
Fails if
A flat percentage of profit before tax, with no loss carry-forward despite accumulated losses on the balance sheet. In an emerging-market structure this is frequently wrong by a material amount, and minimum turnover taxes are missed entirely.
21

Capex, depreciation and the asset base are consistent

What they do
Rolls forward fixed assets: opening balance, plus capex, less depreciation, equals closing balance. Then checks whether maintenance capex is sufficient to sustain the asset base.
Passes if
A clean roll-forward, depreciation derived from the asset base and a stated useful life, and maintenance capex at least holding the asset base flat in real terms.
Fails if
Depreciation as a percentage of revenue, or maintenance capex below depreciation indefinitely. The second is the polite version of assuming the assets never need replacing, and it flatters free cash flow throughout the forecast.
06Before the investment committee

Structure and survival

The final two checks decide what the investor can actually pay and what happens if things go wrong. These are run last, and they are the ones taken into the committee room.

22

The cap table ties to the legal documents

What they do
Compares share counts, classes and option pool in the model against the shareholders’ agreement and the option register, then checks dilution at each future round.
Passes if
Exact agreement with the legal documents. The option pool is created before the round where it is promised, preference and liquidation terms are reflected, and per-share outcomes are calculable at each stage.
Fails if
An option pool assumed post-money when the term sheet says pre-money, or share counts that do not tie to the register. Cap table errors are the most expensive kind because they change what each party actually receives, and they surface at the worst possible moment.
23

The downside case is survivable and honest

What they do
Switches the scenario selector to downside and looks for the month cash goes negative, then asks what the response would be.
Passes if
A downside that is genuinely uncomfortable, a visible minimum cash point, and an identified set of levers with quantified effect and a realistic lag before they take hold.
Fails if
A downside case that is the base case less ten per cent, or one in which cash never goes negative. Either tells the investor the risk has not been thought about. The best-received models show cash going negative and then show precisely what gets done about it.

If more than three of these fail

Three or four failures spread across different stages is normal, and fixable in a week. Failures concentrated in stages 1 and 2 are a different matter: they indicate the model has structural problems that will keep producing new errors as you edit it, and patching individual findings tends to move the problem rather than remove it. At that point a rebuild is usually faster than a repair, and it is worth establishing which situation you are in before a term sheet is on the table.

Have us review the modelHave one built instead