Our method

Why five workbooks

The managing director of a small company asks two questions before most decisions. Will the money last over the coming months? And what does it take for the year to turn a profit? The accounts answer neither, because they only show what has already happened.

The questions sound alike, but they need different models. Whether the money lasts depends on when cash comes in and goes out: credit terms, VAT periods, withholding tax, holiday pay and loan repayments. Whether the year turns a profit depends on what drives the result: volume, price, margins, staffing and costs. A model that tries to do both equally well ends up too coarse for one or too heavy for the other.

HULC therefore has two workbooks at the base. The liquidity model follows every krone into and out of the operating account. The budget and operations model builds the result from the bottom up, from revenue streams, positions and cost lines, and derives the balance sheet and cash flow from there. Both read the same accounts, the same answers from the managing director and the same scenarios, and both run 36 months ahead.

Three more questions come up less often, but weigh more when they do. Does the investment pay off? What is the company worth? And which services and customers do we actually make money on? The investment analysis and the valuation build on the budget engine, so the forecast for the company is the same there as in the budget. The profitability map allocates the last twelve months of the accounts to services and customers.

From accounts to model

The work starts with six exports from the accounting and payroll systems at the balance sheet date:

• The trial balance from 1 January to the balance sheet date, with every account.
• Open items for customers and suppliers, with invoice date, due date and outstanding amount.
• Sales per customer for the last twelve months.
• The monthly income statement for the last 24 months, in whole years.
• The payroll list with position, percentage, monthly salary and dates, without names.

The exports are pasted into the RÅDATA sheet in their original form, never converted by hand. No formula reads a fixed row there, so an export with more or fewer accounts works the same way. The New client button also fetches company data from the open services of the Norwegian Register of Business Enterprises. The municipality number gives the zone for employer’s national insurance contributions, and the industry code gives the sector.

In INNGANGSDATA the accounts are summed into the model’s lines using account ranges based on the Norwegian standard chart of accounts. Two mechanisms keep an account from ending up in the wrong place. Account coverage checks that each account falls in exactly one range. The chart-of-accounts check compares each account name with the line it lands on, using more than a hundred keywords. An account called «Husleie» (rent) inside the cost-of-goods range is flagged before it can affect the figures. Finally, the opening balance is reconciled against the latest filed annual accounts.

The drivers are then read from the accounts:

• Credit days for customers and suppliers and inventory days are calculated from the balances against revenue, purchases and cost of goods.
• Seasonality is calculated per calendar month from the income statement, averaging 1.
• Other operating costs are split into a fixed and a variable part by regression against revenue.
• Growth is calculated from two twelve-month windows.
• Salaries, withholding tax and pensions are calculated from the payroll list and the trial balance.

Every driver follows the same pattern. The figure is calculated from the accounts in a white cell, and a blue cell beside it can override it. The formula uses the override when there is one, and the figure from the accounts otherwise. It is always visible what was derived and what was decided.

The conversation with the managing director

The accounts do not know that a fitter leaves in October, that a service van will be bought next spring, or how hard unexpected events could hit the company. We ask about that in a fifteen-minute conversation.

The conversation form is generated by a button in the model and pre-filled with suggestions from the accounts. The managing director therefore does not answer a long list. They confirm or correct what is already there, and add what only they can know. In the form, the adviser’s estimates are yellow and the managing director’s answers are blue. The questions cover:

• Plans: new and discontinued positions, investments, new loans and dividends.
• Targets and limits: the pre-tax profit target, the overdraft facility and how many months of fixed payments the buffer should cover.
• Growth: volume growth, price increases and the wage settlement, with suggestions from history and the national wage framework.
• Unexpected events: how far revenue can fall, how much the margin can be squeezed, and whether the largest customer could disappear.

The Import conversation button writes the answers to the right places. The notes from the conversation carry through to the app, under «Forutsetninger» (assumptions). The investment analysis, the valuation and the profitability map have their own questions in the same form: the assets in the quote and their effect, the owner salary and non-operating assets, and how time is split between the services.

The liquidity model

The liquidity model calculates the operating account with the direct method. Every receipt and every payment is calculated separately, month by month. The calculation runs in a fixed order, where each step only reads from the steps above it.

1. The result for the month: revenue, cost of goods, salaries, holiday pay, employer’s contributions, pensions, other costs, bad debts, depreciation, interest and tax.
2. VAT: output and input VAT per month, collected into VAT periods and paid when due. A negative period gives a refund two months after the period ends.
3. Inventory and purchases: inventory is steered towards the inventory-days target with half of the seasonal pattern, and purchases are whatever it takes to reach the target.
4. Receipts from customers run through a payment profile. Credit days are turned into the share of sales paid in the month of sale and in each of the next four months, less expected bad debts. Customer items open at the balance sheet date are paid on their due dates.
5. Payments: suppliers through a similar profile, net salaries, withholding tax to the tax deduction account, employer’s contributions per period, holiday pay in June, pensions each quarter, advance tax in February and April and residual tax, investments, repayments and interest on each loan, and dividends.
6. The operating account: opening balance plus receipts minus payments. A negative balance is a draw on the overdraft, with interest and commitment fee.

Alongside, the balance sheet and the cash flow are calculated with the indirect method, from the result and the changes in the balance sheet. The two methods must give the same change in the operating account in every forecast month. The balance sheet must balance. Equity must roll forward with profit and dividends, and loans with repayments and new borrowing. Each of these is a separate check.

Out of this come the figures that answer the question:

• The lowest balance ahead, when it occurs, and the first month the facility is breached.
• Available liquidity: deposits plus the unused overdraft facility.
• Runway: how many weeks available liquidity covers fixed payments.
• The buffer is the number of months of fixed payments the managing director wants in reserve. The model counts the months in which available liquidity falls below it.
• Cash burn: the change in the operating account over the next and the last twelve months, and whether the account turns.
• The cash conversion cycle: customer credit days plus inventory days minus supplier credit days.

The budget and operations model

The budget model builds the result from the bottom up. It has four table sheets the managing director will recognise.

INNTEKTER holds up to six revenue streams. Each stream has one of four model types:

• Transaction: volume times price, with volume growth and a price adjustment once a year in a given month.
• Recurring: customers with churn and new customers per month, such as subscriptions and service agreements.
• Project: revenue spread over several months.
• Capacity: the number of fee earners from staffing times hours per month, utilisation and hourly rate.

Each stream has its own variable cost share, so the contribution margin can be read per stream. The suggestions come from the accounts the stream is linked to.

BEMANNING has one row per position, with full-time equivalents, start and end dates, annual salary and wage adjustment. Holiday pay, employer’s contributions by zone and pensions follow automatically. KOSTNADER holds the cost lines with fixed, variable or step behaviour, indexed once a year. INVESTERINGER holds amount, month, depreciation method, useful life and financing.

From this, the income statement, balance sheet and cash flow are calculated for every month. Three parts take more than a simple formula:

• Tax is calculated on accumulated pre-tax profit within each income year, with tax losses carried forward, and paid in two instalments the following year. That gives the right tax even when the year starts with a loss and ends with a profit.
• The overdraft is drawn only after every other payment in the month has been calculated, and repaid when there is room. Interest is calculated on the opening draw, so the model never becomes circular. Whatever the facility does not cover is shown as an uncovered need and is never hidden.
• Dividends are checked against distributable equity. The model never cuts anything by itself. An unlawful dividend becomes a red check that must be resolved with the managing director.

OUTPUT shows the key figures per scenario, with break-even revenue and margin of safety. The five drivers that move the result most, and what must happen to reach the profit target, are shown there too.

The investment analysis

The investment analysis answers three questions in order: does the investment pay off, can the company afford it, and should it wait a year? It builds on the budget engine, so the company without the project is the same forecast as in the budget.

PROSJEKT holds the investment as the managing director described it: the assets from the quote with price and depreciation group, the lifetime, the effect per year in hours, saved costs, extra sales and lost income, new fixed costs, months to full effect, grants and how the investment is paid for. AVKASTNINGSKRAV calculates the hurdle the project must clear: the cost of equity from the risk-free rate, the market risk premium and beta, the WACC with the company’s gearing, and a premium for the risk class, from replacement to new venture.

NÅVERDI calculates the cash flow from the project month by month in three scenarios, both if the investment is made now and if it is postponed a year, with tax after declining-balance depreciation and a residual value at the end. Out come net present value, internal rate of return, modified internal rate of return and payback time, and the break-even: how far the effect can fall before the present value is zero. FINANSIERING calculates the loan or lease and what the payments do to available liquidity, against the buffer the managing director wants. The answer in OUTPUT is one of six sentences, from «pays off and you can afford it» to «does not pay off», with the advice on timing underneath.

The valuation

The valuation answers what the company is worth today, and what could make it worth more. The main method is cash flow: the forecast from the budget engine for five years plus a terminal value, discounted at a hurdle rate for the whole company. Beside it stand two cross-checks, industry multiples on normalised EBITDA and capitalisation of earnings, and a net asset value from the balance sheet with hidden values and deferred tax.

Before anything is calculated, EBITDA is cleaned in NORMALISERING of what does not belong to operations: the owner salary is set to market salary, private costs and one-off items are removed, and income from non-operating assets is moved to the bridge. VERDIDRIVERE scores owner dependence, customer concentration, quality of earnings, track record, margins, key people, systems and the legal side. The points decide where in the industry’s multiple range the company belongs, and the company-specific premium in the hurdle rate.

VERDI goes from enterprise value to the value of the shares through a bridge: minus loans and overdraft, minus debt-like items, plus excess cash, plus non-operating assets, and plus or minus working capital beyond the normal level. The answer is a best estimate, a range from Downside to Base, and the figure if things go wrong. When the forecast gives more than buyers usually pay, the answer says so plainly.

The profitability map

The profitability map allocates the last twelve months of the accounts to services and customers, with no forecast. TJENESTER links up to six services to the revenue accounts, with the share of the fitters’ time, a typical invoice and costs that belong to this service alone. AKTIVITETER allocates the costs in the accounts to the activities that drive them: fitters’ time, orders and invoicing, customer follow-up, and what cannot be allocated.

The result is three contribution margins per service. DB1 is revenue minus cost of goods and own time. DB2 deducts orders and invoicing. DB3 deducts customer follow-up, bad debts and the cost of capital on receivables. The services are ranked on DB3 as a percentage, and the map is reconciled with operating profit in the accounts, so no krone goes missing. KUNDER does the same per customer from sales per customer. Here Downside and Stress are the time estimates 25% and 50% higher, because time is the most uncertain figure in the map.

The scenarios

All the models calculate three scenarios. Base is the drivers from the accounts with the answers from the conversation. Downside and Stress build on what the managing director says about unexpected events, in a mild and a severe version, not on fixed percentages from the template. At Odalsvarme, Downside is a 15% fall in volume and a margin 2 percentage points weaker. In Stress, the largest customer, with 32% of revenue, disappears from the first forecast month, the rest falls 10%, and the margin is squeezed by 4 points.

The model calculates one scenario at a time. Data tables in OUTPUT run all three through the same calculation and collect the key figures, so three scenarios do not mean three copies of the model that can drift apart. Booked months are identical in all scenarios, and a scenario only takes effect from the month after.

Sensitivity uses the same axes in the liquidity model and the budget model: revenue down 20% and 10% and up 10%, gross margin up 2 and 4 points, and customer credit 10 and 20 days longer. The changes apply only to the forecast months. When Stress means losing the largest customer, the liquidity model also calculates the smallest fixed monthly cut that keeps the operating account within the facility. The investment analysis and the valuation each have a tornado of the drivers that move the answer most.

The checks

A model that gives wrong figures without saying so is worse than no model. The liquidity model has 76 checks, the budget model 57, the investment analysis 99, the valuation 90 and the profitability map 22, and they are evaluated in all three scenarios. They come in five kinds:

• Structure: the balance sheet balances, the calendar is continuous, payment profiles sum to 1, and the direct and indirect methods agree.
• Inputs: every mandatory input is filled in. An empty cell means «not asked», and 0 is an answer.
• Reconciliation with the source: the opening balance against the annual accounts, open items against the balance sheet, sales per customer against revenue, and calculated VAT against the VAT account.
• Reasonableness: withholding tax, pensions, cost of goods, interest rates, contribution margin and EBITDA margin within normal ranges for a Norwegian small business, and a worse scenario never above a better one.
• Delivery: the sheet for the app is complete, with no duplicate keys or error values, and the last booked month lies before the delivery date.

The master check is OK only when every check passes. A warning does not stop delivery, but it must be read and assessed. At the bottom of the check sheet is also the list of inputs where the model uses its own estimate because the managing director gave none. That list is shown in the app, so it is never hidden what was estimated.

Every input has a fixed key in the NØKLER sheet, with sheet, address and label. A separate check compares the label at each address with the map, so the map cannot drift without the master check saying so. Every change to the templates is saved as a new version. It is tested against fixed worked examples with known answers, and with the new feature switched off it must produce zero differing cells against the previous version.

When the months are booked

A budget is worth most when it is compared with what actually happens. Each month, the Monthly routine button reads the month-end trial balance into the FAKTISK sheet. The booked months replace the calculation, and the forecast for the months after starts from the actual balances of receivables, payables and the operating account.

AVVIK shows the forecast against what happened, line by line. The budget model splits the revenue variance into a volume variance and a price variance for the two largest streams, and shows a rolling forecast: actual year to date plus the budget for the rest of the year. The liquidity model has a six-month backtest with fixed tolerances. After three booked months, supplier credit days are taken from the booked months instead of the budget.

From workbook to app

The managing director never receives the Excel file. When the master check is OK, the Delivery copy button creates a copy with values only and no macros. The data tables are read before the copy is made, so Downside and Stress keep their own figures.

The app reads only one sheet in the copy, APP. It is flat, with one row per figure and seven columns: key, scenario, period, value, unit, description and source. The liquidity model gives 925 rows, the budget model 1,570, the investment analysis 3,537, the valuation 1,994 and the profitability map 515. The app always reads by key, never by row number. A new version of a model can add keys, but it cannot change the meaning of a key without renaming it.

Before showing the figures, the app runs its own checks. The master check must be OK, and the booked months must lie before the delivery. The monthly series must agree with the key figures, and the booked months must be identical in every scenario. The overdraft must never exceed the facility, and in a profitable year net profit must never exceed pre-tax profit. The app calculates no new figures. If it finds an error, it shows a red banner instead of the figures.

Odalsvarme AS, a worked example

Odalsvarme AS is fictitious: a heat pump company in Skarnes with 16 employees and revenue of NOK 27.5 million in 2025. The accounts were generated to test the models and demonstrate the app, and they are the same accounts in all five.

At the turn of the year the operating account held NOK 3.3 million, and the NOK 2 million overdraft was unused. A dividend of 1.6 million had been provided for, and pre-tax profit for 2025 was 2.5 million. The 2026 budget builds on what the managing director said in the conversation: prices up 3% in January, heat pump sales up about 6%, and a 4% wage settlement in April. A fitter leaves in October, a new one starts in January, and a service van is bought in April 2027. The target is 3 million before tax. The budget falls short, at 2.28 million.

From January to August 2026, things happened that were not in the budget. Service agreements went from NOK 179 to 199 in March, operating supplies worth 185,000 were bought the same month, and the wage settlement came in at 4.5%. In August an advance order of heat pumps was paid in cash, and 1.1 million was drawn on the overdraft. After August the figures look like this:

In Downside and Stress the service van is postponed, as the managing director said in the conversation. The year is tracking 117,000 below target. In Base the operating account is lowest in September, right after the advance order, but well within the facility, and the account turns. Even so, runway is only 3.1 weeks, against 25 at the turn of the year. If Odalsvarme loses its largest customer, the facility holds until August 2027. A fixed cut of about 185,000 a month from September 2026 would keep the operating account within the facility.

The budget model also has a cash side. There the overdraft is repaid after August, and cash is lowest in September 2026, at 322,000. The difference from the liquidity model is that the liquidity model follows payment dates and due dates more closely. When both models have been delivered, the app uses the liquidity model for the question of whether the money will last.

The investment is fictitious: Odalsvarme currently rents out its warehouse and is considering converting it into its own workshop and store for 1.6 million. The rental income of 180,000 a year disappears, the store they rent in Kongsvinger for 260,000 is given up, and the fitters manage about 450,000 more in installations a year. The loan covers 70% over eight years at 7%. Net present value is 303,507 in Base, −323,262 in Downside and −947,462 in Stress, the internal rate of return 15.1% against a hurdle of 10.5%, and the break-even says the effect can fall 16.3% before the investment nets to zero. The answer is «Investeringen lønner seg bare hvis alt går som planlagt» (the investment only pays off if everything goes to plan).

The valuation gives a best estimate of 20.3 million for the shares, with a range from 19.8 million in Downside to 20.3 million in Base, and 8.5 million in Stress. The warehouse sits in the bridge at its market value of 3 million, and the rental income is normalised out of EBITDA so it is not counted twice. The cost of equity is 21.6% and the WACC 16.2%. The cash flow gives 8.6 times normalised EBITDA, against 3.4 in the industry, so the answer says the forecast gives more than buyers usually pay, and that the figures assume the growth in the budget arrives.

The profitability map shows that heat pump sales earn the most, with a DB3 of 7.0 million and 36% of revenue, while the service agreements lose 400,000 a year. Their price must rise 45% for them to break even. The map reconciles with operating profit of 2.63 million, and none of the eleven customers loses money.

What the models do not do

The models do not replace the accounts or the accountant. They calculate what has not happened yet, and the figures are never more certain than the assumptions. That is why the assumptions are written out, the estimates are listed, and the scenarios are the managing director’s own. Downside and Stress are not probabilities, but what the company must be able to withstand for the plan to hold.

Some limits are known and documented in the models. Credit terms beyond four months are paid in the fourth month. Deferred tax is not included in the budget model. Sensitivity becomes almost flat when the low point falls in the first forecast month, because a change in revenue barely reaches the account in the same month. Such limits are listed in the PROSESS sheet, next to the checks that catch them.