What a QDB reviewer actually looks for in your cash flow

Most applicants build a cash flow hoping it will persuade. That is the wrong instinct. A QDB reviewer is not reading your model to be convinced by it. They are reading it to find the point where it stops holding together, and they are running the same small set of tests they ran on the file before yours.

Almost none of those tests are secret. QDB publishes its feasibility study requirements, its financing terms, and a customer journey booklet that sets out how money actually moves once a facility is approved. Read together, they describe the reviewer's desk fairly precisely. What follows is that description, applied to the one document that decides most applications.

This is written for Direct Financing specifically. The application must be submitted under the ownership of a Qatari citizen, the project must fall within a financed sector — manufacturing, education, healthcare, services, agriculture, livestock or fisheries — and the company must be headquartered in Qatar. If your project does not clear that gate, the quality of your cash flow is not the constraint.

The reviewer is answering two questions at once

QDB is a development bank. That is not branding; it changes how the file is read.

The first question is whether the project deserves the mandate. QDB assesses applications against criteria including the demand–supply gap in Qatar, use of locally available raw material, reduction in import reliance, export orientation, technology and IP creation, and labour intensity. There is an internal model that scores this, and the bank has said economic value carries more weight in pricing than risk does.

The second question is the ordinary one every lender asks: can this repay?

Applicants tend to answer the first question in the narrative and the second in the spreadsheet, as though they were separate exercises. The reviewer reads them together. Your local procurement, your export split, your payroll — the qualitative case leaves fingerprints on specific lines in the model, and a study that argues for import substitution while the cost sheet shows everything landing from abroad has contradicted itself in the only place that matters.

Build the model over the life of the facility, not over five years

The most common structural error in Doha is a five-year model.

QDB's feasibility study requirements specify a ten-year forecast horizon, applied across all constituent schedules. The customer journey material puts it differently but to the same effect: projections must cover at least the expected repayment period. Since the maximum term is fifteen years, a five-year model cannot show the facility being repaid at all. It answers a question nobody asked.

This is the fastest credibility signal in the document, positive or negative. A reviewer opening a model that runs to year ten with the debt schedule fully amortised knows within seconds that the person who built it had read the requirements. A model that ends in year five tells them the opposite, before a single number is examined.

Your months are relative. Your market data is not.

Look at the timeline QDB publishes for its own process. Discovery can take up to two months. Application preparation runs four to six weeks to reach an offer approval. The committee takes a further two to three weeks, terms and conditions follow within about two weeks of that, the offer letter is valid for thirty calendar days, and security documentation is expected to be completed within another thirty days.

Add it honestly and you are somewhere between four and eight months from first visit to a facility you can draw on — before construction, before a machine is ordered, before an employee is hired.

None of this means your model should carry dates. Financial models are built in relative periods — month 1, month 2, running out as far as the repayment obligation requires — and that convention is right. Month 1 is the first month of implementation, not a square on a calendar.

The error sits somewhere less obvious. Your market assessment is anchored to a calendar whether you intended it or not. The demand figures, the import statistics, the competitor set and above all the prices in your revenue build are the ones you observed while writing the study. But nothing in the model happens until the facility is available, and revenue does not begin until the project has been built, commissioned and staffed. Work that through against the timeline above and first revenue sits at least ten months after the date you apply, often considerably more.

So the market your model sells into is not the market you researched. Build the market assessment forward from the realistic revenue start date rather than from the day the study was written, and say in the assumptions which date you have used and why. It is a short paragraph that removes a question the reviewer would otherwise have to ask — and it reads as competence, not pessimism.

DSCR is the number — and it is not the only one they have named

Debt service coverage is where the file lives or dies. It compares cash available for debt service against the principal and profit falling due in the same period, and the emphasis is on cash: a project can be profitable in year three and unable to pay in year three, and the reviewer's job is to find out whether yours is one of them.

QDB does not leave the ratio list open. The requirements name debt service coverage, interest cover, total liabilities to net worth, interest-bearing debt to net worth, cash conversion cycle, and performance ratios as items the model must cover, alongside project appraisal indicators — IRR, NPV and payback period.

Compute all of them, on their own clearly labelled schedule, for every year of the forecast. Do not make the reviewer derive a ratio you were asked to provide. The applicant who leaves them out is not usually hiding anything; they simply did not read the list. But the reviewer cannot tell those two cases apart, and does not have to.

Model the month grace ends, when principal and profit both begin, and check that the revenue ramp has actually arrived by then rather than shortly afterwards. The first period of full debt service is the one the reviewer turns to first, because it is where a project that looked fine on paper tends to fail.

One further point on pricing, which almost nobody models. The return rate goes up to 5%, and QDB has said the rate can be reduced by increasing capital contribution or by offering additional collateral. That makes the profit rate an input you have some control over. Run a case at a higher equity contribution and show what it does to your coverage. It is a considered piece of analysis, and it demonstrates you understand the product rather than just applying to it.

The money never arrives in your bank account

This is the single most common structural error in models that are otherwise well built.

Almost every cash flow that crosses my desk shows the facility drawn down as a cash inflow, sitting in the company's account, and then spent on machinery and construction. That is not how QDB Direct Financing works.

For raw material and movable assets, the proforma invoice is issued in QDB's name for Islamic financing, QDB contracts with the supplier through a local purchase contract or a letter of credit, and QDB pays the supplier. For machinery and equipment, the same pattern applies under an Ijara structure. For construction, disbursement runs through the Engineering Projects Department against certified payment applications, and the payment goes to the contractor. In each case the bank pays the vendor and then informs you.

So the facility is not a cash inflow. It settles specific project costs directly, and the only amounts that should ever pass through your operating cash flow are the ones you actually handle.

Your equity goes in transaction by transaction

The corollary is about your own contribution, and it is just as commonly wrong.

Financing covers up to 60% of total project cost, up to 80% of equipment and machinery excluding construction, and up to 100% of raw material. Most models therefore inject the balancing equity as a single lump in month zero and never think about it again.

In practice, contribution is evidenced at the point of each disbursement: you provide a cheque for your share, or proof of payment if you have already paid it. Your equity is therefore staged alongside the facility, transaction by transaction, and the early months of the model should show it that way. The difference is not cosmetic — it changes the shape of the cash line during exactly the period when the reviewer is looking hardest for a month that goes negative.

Your project cost is not fixed

Applicants treat total project cost as a settled number once the study is written. The bank does not.

Quotations must be signed and stamped by the supplier and dated within the last three months, and proforma invoices older than three months are not accepted. Machinery, supplier and cost must match the approved quotations, with any change approved before disbursement.

Read that against the timeline above and the implication is uncomfortable but useful: quotations obtained at the start of a four-to-eight month process will expire during it. When they are refreshed, the total project cost moves — and because financing is capped as a percentage of that cost, your required contribution moves with it.

The applicants who handle this well build a contingency into project cost and say so in the assumptions. The ones who handle it badly discover it at disbursement, when the equity cheque they are asked for is larger than the one they budgeted.

Working capital is where profitable projects fail

QDB asks for the split between cash sales and credit sales in the revenue assumptions. Separately, in the demand and supply section, it asks for supplier and customer credit terms.

Those two disclosures get read against each other, and the contradiction is common enough to be worth naming. A study describes selling to main contractors and government-linked buyers on ninety-day terms, then the model collects every riyal in the month of sale. One of those is wrong, and the reviewer does not need to decide which — the inconsistency is the finding.

Build receivable days, inventory days and payable days as explicit assumptions, drive the working capital movement from them, and report the cash conversion cycle as the named ratio it is. In Qatar this is not a theoretical exercise. Payment terms from large contractors and government entities are long and are not always honoured on schedule, and a manufacturer with a healthy margin and a hundred-and-twenty-day collection cycle can run out of money without ever making a loss.

While you are there, add a note about lead times. Raw material and machinery transactions carry processing time at the bank of several working days each, and letters of credit have their own cycle. Inventory does not appear the day you decide you need it.

Revenue has to reconcile to capacity

The requirements ask for total plant capacity in the project details, and separately for estimated sales volume with a rationale, product-wise selling price, and year-on-year volume growth.

Those are given as separate disclosures, but the reviewer combines them in about thirty seconds: revenue divided by price gives units, and units are compared against stated capacity. Year one revenue implying 85% utilisation of a plant that has not yet been commissioned is where a great many otherwise decent applications lose credibility, and it is entirely self-inflicted. Show utilisation as an explicit assumption that climbs over two or three years, and let the revenue fall out of it.

The same discipline applies to the market case. Growth in the projections has to trace back to something — a demand–supply gap you have quantified, a signed MoU or letter of intent, an import figure you are displacing. A curve that bends upward because businesses are supposed to grow is not an assumption; it is a hope with a formula attached.

The cost lines that get checked, and the ones that get missed

The required breakdown is granular. Direct costs are to include raw material with volumes, stores and spares, direct employee cost, power and utilities, and factory land rent. Indirect costs are to include general and administrative expenses, salaries with a listed roster by category and salary level, and office rent.

The lines that go missing are more revealing, and they tend to be the ones that are invisible until year three rather than the ones you forgot to type. Pre-operative expenses are explicitly required within the capital investment schedule and are routinely left out altogether. Replacement capital expenditure is rarely modelled at all, which is survivable over five years and conspicuous over ten. And a great many models carry no salary for the owner, which does not read as commitment — it reads as a cost the model is hiding, and it makes a reviewer wonder what else is being absorbed silently.

Those are examples rather than a list. The general test is worth more than the specifics: go through the model and ask which costs you have assumed away because they do not fall due in year one. That is the category the reviewer is looking in.

A depreciation schedule including useful life of assets is also required. Beyond the accounting, the reviewer uses it to test tenor against asset life. Asking for fifteen years against equipment with a seven-year life invites a question you would rather not have to answer in the meeting.

The dividend policy tab

The requirements list a dividend policy among the model's schedules. In practice it is the single most frequently omitted item, and its absence is informative.

The reviewer reads it as one question: does the owner intend to take cash out before the bank is repaid? Silence is not neutral. State a policy, make it conservative, and tie distributions to a condition — no distributions until the facility has been servicing for a defined period and coverage is above an agreed level. It takes three lines and it removes a concern the reviewer would otherwise have to raise.

Sensitivity is prescribed, not open-ended

This is not a section where you get to choose the scenarios.

The requirements set out variations of 5%, 10% and 15% in revenue, in direct costs and in indirect costs, with the impact shown on gross profit ratio, net profit ratio, cash flows, IRR and payback period.

Produce exactly that grid. Then go one step further and add the two stresses that matter most in Qatar and are not on the list: a delay in the revenue start date, and a lengthening of collection days. Those are the two variables that actually break projects here, and an applicant who has stress-tested their own file before the bank does is in a materially different position from one who has not.

The point of the exercise is not to prove the project is invulnerable. It is to show you know where it is fragile and have thought about what you would do. A model that stays comfortably above coverage under a 15% revenue reduction is either very robust or not being honest, and reviewers have seen enough of both to tell.

Everything gets cross-checked

The cash flow is not read in isolation. It is read against the rest of the file, and the reviewer's attention goes to the seams:

• Machinery quotations against the capital investment schedule.
• The contractor's bill of quantities and quotation against both construction cost and construction period.
• The implementation plan against the month revenue begins.
• Letters of intent and MoUs against year one revenue.
• Stated capacity against units sold.
• Credit terms in the market section against collection assumptions in the model.
• For an existing company: three years of historic financials, and bank statements, against the base the forecast is built on.

That last one deserves emphasis. For an established business, the historical record is the reviewer's anchor. A forecast that departs sharply from three years of actuals is not disqualifying, but it does need a reason attached to it — a new line, a new contract, a new plant — and the reason has to appear in the model as a driver rather than as a change in slope.

What gets flagged

A short list, in the order a reviewer tends to find them:

• Revenue appearing during the construction period.
• Any month with a negative closing cash balance.
• A balance sheet that balances through a plug.
• Hardcoded values sitting inside a formula column.
• Costs held flat while revenue triples.
• A five-year model against a longer facility.
• Gross margins that do not belong to the sector.
• Another country's currency, tax regime or template headings left in the file.
• No assumptions sheet — or one that nothing else references.

That last one is worth its own test. A reviewer will change one input and watch what happens downstream. If nothing moves, the model is not a model. It is a picture of one.

Before you submit

You do not need the whole list to find out where you stand. Four questions will tell you most of it.

1. Does the model run the full length of the facility? Not five years by habit — as far as the repayment obligation actually runs.
2. Does the money move the way the bank actually moves it? Paid to vendors directly, rather than landing in your account and being spent from there.
3. Is every single month's closing cash positive? Every month, including the ones during construction and the first month of full debt service.
4. Does your revenue reconcile to your capacity? At a utilisation you would be willing to defend out loud.

If any of those four gives you pause, the answer is not to patch that line. It is that the model was built to a different set of questions than the ones it will be read against, and patching rarely survives contact with a reviewer.