Choosing between a bank and a non-bank lender for growth money
Score each lender on five tests before comparing offers. The public sources describe lending conditions in aggregate, so the scores must come from the written terms you collect.

A mid-sized company with a growth plan and a bank meeting in the diary has often made its choice already, without noticing. The bank's offer is the one on the table. The non-bank lender's offer is the one nobody has modelled. The gap matters because the two are priced, documented and policed differently. The Financial Stability Board estimates private credit assets at $1.5–2 trillion, and it says rapid growth in that market has supported financing for mid-sized companies. The question is not which type of lender is better in general. It is which one fits this company's cash needs this year, under conditions that will apply when trading gets harder.
Score each lender on five pillars: cost, speed, covenant strictness, data requirements and continuity risk. Give each pillar a score from one to five, and write one sentence of evidence behind every score. Refuse to score any pillar you have not seen in a written offer. The public sources cannot do this work for you. The supervisory material reviewed for this piece describes lending conditions in aggregate and does not publish the terms of any named lender. The scores are only as good as the documents you collect.
Chapter one: 1. Cost of the money
Compare all-in cost, not the headline rate. A bank quote and a non-bank quote are rarely built the same way. Ask each lender for a schedule that covers interest, arrangement fees, commitment fees on undrawn amounts, legal costs, any charge for early repayment, and any margin tied to a reference rate that can move. Convert every item into one annual figure over the expected life of the facility. A lower rate with a larger fee on headroom you never draw can cost more than the alternative. Do the arithmetic on paper before the meeting, not after it.
The cost pillar also depends on the wider rate environment. The European Central Bank's Bank Lending Survey for Q3 2026 found that loan demand fell across all segments, and that higher interest rates were the main driver cited. Borrowers are already weighing cost against need. A sensible test is to model each offer at a rate higher than today's and see whether the business still services the debt comfortably. In our experience, the offer that only works at today's rates is the one that disappoints first.
Chapter two: 2. Speed to funds
Speed is the pillar most often scored from the first meeting, which is the wrong place to score it. Measure it from the signed term sheet to money in the account, and count every condition precedent. Ask each lender for its own timetable in writing. The timetable should show credit committee dates, legal review, any valuation, identity and sanctions checks, and the paperwork that must be complete before the first drawdown. Then ask what would make it slower. A lender that cannot name its own delay risks has not yet thought through your deal.
Market conditions can change the answer quickly. The ECB reports that euro area banks tightened credit standards for firms in Q3 2026, linked to higher perceived risk and lower risk tolerance, and expects standards to tighten further in Q4 2026. The Federal Reserve's Senior Loan Officer Opinion Survey, which runs quarterly and is timed around FOMC meetings, asks U.S. banks about changes in lending standards and terms. These surveys show the direction of travel for banks. They do not show how long your own lender's committee will take. That is a question to put to the lender directly, and to record in writing.
Chapter three: 3. Covenant strictness
A covenant is a condition the borrower promises to keep, written as a ratio or a test. Strictness depends on how many tests there are, how each one is defined, how often it is measured, and how much room the business has before a breach. Read the definitions section before the headline numbers. Check what counts as debt, what counts as earnings, which adjustments are allowed, and whether the test runs on a single quarter or on a rolling twelve months. Two agreements with the same ratio can behave very differently in a bad quarter.
The public record is silent here. None of the sources reviewed gives covenant terms for any bank or non-bank loan. The ECB's finding that banks tightened standards because of lower risk tolerance points in a direction, but it does not say which terms changed. Tighter standards usually show up in terms as well as in price, in our experience. The practical test is to model the business through a plausible bad year, such as a quarter in which a major customer pays late, and see which covenant breaks first. Then ask what the lender can do about it. A right to renegotiate and a right to take control of the business are very different outcomes, and the agreement should say which one applies.
Chapter four: 4. Data requirements
A lender's data request is a window into its own accounting. According to the IFRS Foundation, the impairment rules in IFRS 9 cover expected credit losses on financial assets and on commitments to extend credit. Classification depends on the lender's business model and on the contractual cash-flow characteristics of the loan. A lender that holds an undrawn facility on its books has to estimate losses on money it has not yet lent. In effect, it will want forward-looking information, not only last year's accounts. Expect requests for monthly management accounts, a three-year plan with downside cases, customer concentration figures and debtor ageing.
Scoring this pillar is a counting exercise. For each lender, list every data item it asks for. Mark whether you already produce it monthly and to a standard an outsider would accept. Then score the gap. A lender that asks for data you do not have is not automatically the wrong partner, but the cost of building that data and the time it takes both belong in the cost and speed scores. Do not let the data request turn into a second negotiation over terms. If the request keeps growing after the term sheet, record each addition in writing.
Chapter five: 5. Continuity risk
Continuity asks what happens to the loan when the lender changes. A bank can be bought, restructured or put into resolution. A non-bank lender can sell the loan, change how it funds itself, or leave a sector. The Financial Stability Board says private credit growth creates potential financial stability risks. Its chair has written that markets could be exposed to a sharp, disorderly adjustment. The FSB's peer review also says resolution for the failure of large systemic banks needs further work, in particular on the implementation of Key Attribute 6. The Bank for International Settlements summarises the Basel Committee's post-crisis reforms on its Basel III page.
The test is practical. Ask whether the agreement allows the lender to sell or assign the loan, and on what notice. Ask what happens on a change of control, a downgrade or a sale of the lender itself. Ask how the lender funds the facility: from its own balance sheet, from investors who can withdraw, or from a bank that may itself be under pressure. Then ask what you would do if the second tranche were delayed by six months. A lender that cannot answer the funding question plainly has earned a low continuity score, whatever its rate. Score it honestly, even when the relationship feels good.
Before you score a lender
Answer each question from the written offer. A no on any question means that pillar cannot yet be scored.
- Does the written offer state an all-in annual cost that includes fees on undrawn headroom?
- Does the lender give a dated timetable from signed term sheet to first drawdown?
- Have you read the covenant definitions, not only the headline ratios?
- Have you modelled a bad quarter against every covenant in the agreement?
- Is every data item the lender requests listed, and marked as produced monthly or not?
- Does the agreement state what happens on a sale of the loan or a change of control?
- Has the lender explained in writing how it funds the facility?
- Have at least two lenders quoted on the same definitions and the same timetable?
Outgrace checklist
Where it breaks
The framework fails in four predictable ways. The first is comparing headline cost across offers that use different definitions of earnings, debt or fees. The second is scoring speed from the first meeting rather than from the signed term sheet. The third is treating a non-bank's flexibility as a sign of looser terms. That flexibility is often paid for somewhere else in the offer, in price, fees or a tighter test that is easy to miss. The fourth is assuming the public record will settle a close call. It will not. The Federal Reserve's survey covers up to 80 large domestic banks and 24 U.S. branches and agencies of foreign banks. The ECB's survey covers around 160 euro area institutions. Neither says anything about the regional bank, the specialist fund or the lender your company is actually talking to.
Weighting is the other place the method can be bent. A business with one large customer should weight covenant strictness and continuity more heavily, because a single lost contract can trigger a breach. A business with a long, contracted revenue base may reasonably weight cost more. Write the weights down before any offer arrives, and keep the note. Otherwise the scoring tends to drift toward whichever lender the founder or finance lead already prefers, and the sheet then becomes a record of that preference rather than a test of it.
Start on Monday with three to five written offers, one per lender, and one sheet with the five pillars down the side. Fill each cell from the document, not from memory, and mark any cell that the document does not answer. Score each pillar, write one sentence of evidence beneath every score, and send the sheet to your finance lead and your board chair before anyone replies to a lender. Ask the lender for its definitions page and its funding explanation first. Those two answers will move the scores more than any headline rate, and they are the answers most likely to be missing from a first proposal.