The term sheet, term by term

All numbers illustrative.

Grouped by what the terms actually control: the economics (what you pay), the structure (how the machine is built), and the controls (when the money stops). Founders read the first group and skim the other two. The other two are where facilities die.

The economics

Commitment amount — the headline is not the number. A "$100M facility" might commit $20M at close, with the rest in uncommitted tranches that unlock only if performance holds, conditions are met, and — read this part twice — the lender consents, in its sole discretion. "Sole discretion" means the upsize is a hope, not a right. I've seen competing sheets with headlines of $50M, $25M, and $15M that all reduced to the same committed dollars. Model your costs, your covenants, and your press-release restraint off the committed amount. The uncommitted tail is real option value — lenders do fund upsizes for performing books, it's their favorite trade — but it's their option, not yours.

While you're here, check what unlocking a tranche requires: outstanding-balance thresholds, seasoning minimums, performance tests, sometimes a fresh fee. And check whether the upsize resets other clocks — no-call periods and exclusivity windows that restart on upsize are a recurring trap (more below).

Advance rate — not a number, a machine. The advance rate is the % of eligible loan balance the lender funds; at 80%, they put up $80 per $100 of loans and you fund the rest. But the single number in the summary box is the least interesting thing about it. What matters is the mechanism — and the market runs at least four:

Whatever the machine, the flip side is your first-loss position — the slice you fund, which absorbs losses before the lender takes a dollar of pain. Think down payment on a mortgage. At a 90% advance rate, a $50M drawn facility needs $5M of your capital riding along; at 70% it needs $15M. This is the structural reason lending startups raise big equity rounds, and you earn no lender-style return on it. The advance-rate machine, run against your growth plan, quietly determines your next equity round. Model it.

Interest rate — spread, benchmark, and the floor. Pricing is usually benchmark + spread: say SOFR + 9%. Three sub-terms carry the real information:

• The floor. A benchmark floor (say 3%) sets a minimum on the benchmark leg, capping your benefit if rates fall. Here's the move to watch for: a lender "matches" your target spread but attaches a floor well above the forward curve — congratulations, you negotiated the visible number and lost the invisible one. The floor is part of the price. Price it as one: compare sheets on all-in yield under your own rate scenario, not on spread.
• Tier direction. Spreads often step down with scale (say +9% on the first $10M, +7.5% beyond $25M). Good — but check whether step-downs apply per-tranche or whole-book, and notice when a sheet applies step-ups whole-book (advance rate) but step-downs per-tranche (pricing). Notice the asymmetry, and ask about it — it's the kind of thing that gets fixed when raised at sheet stage.
• Default interest. A default margin of +2% stacking on during an event of default is standard. But read the formulation: "+2%" is a penalty; "+2% per month while continuing" is an annihilation — four months into a disputed default and your all-in rate has doubled. The difference between flat and escalating default interest is an order of magnitude, hiding in one word.

Mechanics footnote that's worth actual money: interest usually accrues on an actual/360 convention — daily interest computed on a 360-day year but charged for all 365 actual days, which silently uplifts the effective rate by about 1.4%. Sheet says 12%, math says 12.17%. Small, but it compounds with everything else in Part 6.

The fee menu. Different lenders assemble wildly different fee stacks around similar headline pricing, and the stacks are not equivalent:

• Structuring/upfront/origination fee — charged at close on the committed amount (say 1% of $20M = $200k on day one, drawn or not), and often again on each new tranche.
• Per-draw fees — say 0.5–2% of each draw. Cheap-sounding, but on a facility you'll cycle through repeatedly, a per-draw fee is a recurring toll on your own growth; 2% per draw on a fully-deployed facility can quietly exceed a big upfront fee.
• Unused/undrawn fee — say 0.5% annually on committed-but-undrawn capital, from day one until deployed. This is the fee that punishes oversizing (Part 7).
• Agent/admin fees — flat quarterly or annual amounts for facility administration, deployed balance irrelevant.
• Work fees / good-faith deposits — covered in Part 4; they belong in the same spreadsheet.
• Legal and diligence pass-throughs — both sides' lawyers, yours to pay, uncapped unless you cap them.

No single fee is outrageous. The stack is the story — sum every line against your honest draw curve before comparing sheets (Part 6 walks the full model).

Warrants — the equity kicker. Many (not all) credit funds ask for warrants: rights to buy your stock, sized as a % of fully-diluted ownership. On similar deals I've seen the ask run from zero to a few points, which tells you the real range is "negotiable." Things that matter more than the headline percentage:

• Strike. Warrants struck at your last-round or 409A price share upside. Penny warrants — strike of ~$0 — are simply a grant of equity; a "1% penny warrant" costs you the full 1%, not the appreciation on it. Price accordingly.
• Vesting/exercisability. Push for milestone gating: half exercisable at close, half only when the bigger facility actually funds. Pay for capital you receive, not capital you're promised. And check the failure case — if the milestone facility never happens, does the second warrant die or live on anyway?
• The rider clauses. Warrants routinely arrive with investor-grade information rights (audited financials, valuation reports), pro-rata rights, sometimes board observation. And in many deals the warrant documents are defined as loan documents — meaning a breach of the warrant agreement is a default under the facility. Your equity kicker just became a covenant. Read it as one.
• The adjacent asks. Some lenders also want the right to invest directly in your equity rounds. That's neither good nor bad — a debt partner with skin in the equity can be lovely — but it's compensation, and it belongs in your comparison of what each sheet truly costs.

Reserve requirement. Some lenders require a reserve account — say 2–5% of outstanding balance, funded at close or trapped out of excess interest — that tops up payments if performance slips. Better portfolio, smaller reserve; it's a negotiable dial, and it's your cash earning nothing, so dial it.

The structure

The SPV and its cast — covered in Part 2. At the sheet stage, verify: who pays for the backup servicer and verification agent (you, or out of the servicing fee — negotiate this), when each requirement activates (at close vs above a balance threshold), and whether the lender is also the facility agent (fine, but then the "agent fee" is them charging you to administer their own loan; feel free to say so).

Guarantees — the spectrum and the perimeter. "Non-recourse" is the brochure word; the actual recourse in any deal sits somewhere on a spectrum, and the sheet tells you where:

• Bad-acts ("bad boy") guaranty — the floor, present in essentially every deal: the parent (sometimes founders personally) guarantees losses caused by fraud, misappropriation, willful misconduct, voluntary bankruptcy, unauthorized transfers. Survivable and standard — but read the trigger list word by word, because "bad acts" drafted broadly ("any material misrepresentation") starts covering ordinary business mistakes.
• Limited guaranty — parent guarantees a slice, say 10–20%. Two perimeter games hide here. Denominator: 10% of the facility limit vs 10% of outstanding advances vs 10% of the high-watermark of advances are three different numbers — on a $50M limit drawn to $10M, that's $5M vs $1M vs whatever-you-once-owed. Duration: does the guaranty burn off when the SPV structure matures, or live forever? Some deals run full recourse during an initial period and step down later — which means the "bankruptcy-remote" structure you're paying to build doesn't actually protect you until phase two. Know which phase you're in.
• Full corporate guaranty — the parent (sometimes "and all existing and future subsidiaries") unconditionally guarantees everything. This is just recourse lending wearing an SPV costume. Sometimes it's the honest price of being early. But price it — a full guaranty and a limited one are not the same product at the same spread.

Personal exposure. Sheets touch the founders three ways, and you should have a position on each:

1. Personal bad-acts guarantees — founders personally on the hook for the fraud-type triggers. Common in early deals; the negotiation is usually about achieving the same protection at the company level instead. Precedent exists for both; ask.
2. Key-person events — a named founder leaving triggers a default or termination right. Negotiate a cure window: "60–90 days to hire a replacement reasonably acceptable to the lender" is achievable and turns an instant default into a manageable process. No-cure key-person clauses are common in first drafts and frequently softened when challenged. Also check the trigger's breadth — "ceases full-time employment" vs "ceases involvement in a supervisory capacity" are very different lives.
3. Key-person life insurance — policies on the founders with the lender as beneficiary. Standard-ish, cheap-ish, mostly fine; just budget it.

The waterfall. The fixed order in which every collected dollar gets distributed: typically third-party fees (banks, backup servicer, agents) → servicing fee → lender interest and fees → lender principal (to cure any borrowing-base deficiency) → and then residual to you. Set at close, unchangeable after. During a trigger event or default, the waterfall usually flips to "100% to lender until cured or repaid" — the cash sweep. Read the waterfall slowly; it is your actual cash flow, and the placement of the servicing fee (senior, ideally) is what keeps your ops funded in a workout.

Draw mechanics — the growth governor nobody models. Buried in the mechanics section: minimum draw sizes, notice periods (2–6 business days), draw frequency caps ("one draw per week"), monthly draw ceilings, conditions to each draw (no default, no MAC, clean borrowing-base certificate). Run these against your origination curve. A facility that allows, say, $1M of draws per month cannot fund a book originating $2M per month — you'll carry the overflow on your own balance sheet, which is exactly what the facility was for. The facility's plumbing has to match your product's cash rhythm: a 5-day notice period is nothing for 36-month loans and fatal for same-day advances.

The controls — the part founders under-read

Eligibility criteria. The list of requirements a loan must meet for the lender to fund it: borrower attributes (minimum credit score, time in business), loan attributes (max size, max term, rate bounds), documentation standards (guarantees, filings, e-sign compliance), origination standards ("originated under the underwriting guidelines"). This defines which part of your book the facility can actually see — drafted narrow, your $50M facility can only see $30M of your book, and the other $20M sits unfinanced on your balance sheet. Two things at sheet stage: get the list with numbers now (Part 8 explains what happens if you don't), and check the treatment of over-limit loans — "loan above the size cap is ineligible entirely" vs "eligible up to the cap" is a borrowing-base difference worth actual money.

Concentration limits. Caps on portfolio composition — say, no obligor over 3%, no industry over 20%, no channel/partner over 25%, no single geography over 30%, riskier-grade paper capped at X%. Excesses usually just drop out of the borrowing base (fine-ish) rather than triggering default (bad — check which). Two sophistications worth pushing for: a ramp-up holiday — limits suspended until the book hits critical mass, because a 20-loan portfolio violates every concentration test by existing — and limits that reflect your structural reality (if your model concentrates in a few partners by design, a generic 10% partner cap makes the facility unusable; negotiate the number or the definition). Watch also for weighted-average requirements hiding here — minimum weighted-average portfolio yield, maximum weighted-average term. A yield floor is the lender setting a pricing floor on your product: drop your rates below it and your loans stop qualifying. That's a product decision made in a credit document.

Covenants. Parent-level obligations that must hold at all times. The recurring cast:

• Minimum liquidity/cash — often "greater of $X or Y months of operating burn." Sneaky-important: this cash is frozen — count it out of your runway math — and the formulation matters ("6 months of burn" tightens automatically as you scale spend).
• Runway/net-worth tests — same idea, different clothes.
• No additional debt — standard, but check the perimeter: "borrower shall incur no debt" is expected (the SPV is single-purpose); "borrower, parent, and any affiliate shall have no other debt outstanding" locks your entire corporate family out of every other financing on earth — negotiate carve-outs (equipment leases, cards, subordinated notes, a basket).
• Equity-raise requirements — some sheets condition closing, or covenant post-closing, on you raising equity: "minimum $XM of primary capital with at least one institutional investor." Your debt sheet just scheduled your equity round. If the milestone is plausible, fine — but recognize you're signing a fundraising obligation with a default attached, and haircut accordingly.
• SPV separateness, liens, use of proceeds — standard kit.

Performance triggers. Portfolio metrics tested monthly — delinquency ratios, cumulative or vintage loss rates, collection rates, excess spread — with defined consequences. Two rules for reading them:

Consequences matter more than levels. Map the ladder: does a breach pause new funding (recoverable), trigger early amortization with a full cash sweep (painful), cut the advance rate (a margin call), or constitute an event of default (existential)? The graduated ladder — pause → sweep → default — is livable. The compound version — one breach makes draws discretionary and cuts the advance rate 15 points and is an EOD, simultaneously — turns a bad quarter into a death spiral. Same trigger level, utterly different facility.

TBD is a deferred fight. It is remarkable how often every actual number in the triggers section of a term sheet reads "[TBD]" or "to be mutually agreed during documentation." Every one of those is a negotiation you've agreed to have after signing, in exclusivity, with your leverage gone. Part 8 is entirely about this. At sheet stage the rule is: numbers, now, or a written methodology for setting them ("delinquency trigger set at 1.5x trailing-six-month average, floor of X%") — something objective enough to survive the leverage flip.

Events of default. The exhaustive list of conditions that let the lender stop funding, accelerate repayment, and switch on default interest: payment failure, covenant breach, trigger breach, misrepresentation, insolvency, change of control, key-person events, judgments over a threshold, "material adverse change." Three clauses deserve special scrutiny:

• Cross-default breadth. "Default under any other agreement" is one thing; "any circumstance giving any counterparty a right to accelerate, whether or not exercised" means a technical foot-fault in an unrelated contract — one the counterparty happily waived — defaults your credit facility. Narrow it to material agreements, actual acceleration, and a dollar threshold.
• MAC clauses. "Material adverse change" — sometimes drafted as "any material impairment of the prospect of repayment" — is a subjective default that rests entirely on the lender's judgment. You won't delete it. You can narrow it and make it about objective conditions.
• Cure periods. Check which defaults have them. First drafts commonly give payment defaults zero cure, most covenants zero cure, and a short cure for a residual category. Negotiate real cure periods for everything administrative — a late report should never be an event of default at 9am the next morning.

Representations. Claims you certify as true at close and often re-certify with every draw and report — legal compliance, loan validity, data accuracy, no undisclosed liabilities. Banks are strictest here. The practical exposure: every compliance certificate you sign for the life of the facility re-makes these reps, and an inaccurate rep is itself a default. Whoever signs these needs to actually know they're true. Build the internal checklist now.

Reporting requirements. Weekly or monthly borrowing-base certificates with loan tapes; monthly financials on a deadline (20–30 days); quarterly forecasts; annual audited statements (yes, that means you need an auditor); compliance certificates; live read-only bank access; sometimes board-materials rights riding along. Fully covered in Part 10 — at sheet stage, just make sure someone on your team has read the reporting section and said "we can produce this, on this cadence, without hiring three people." Or budget the three people.

Underwriting-guideline lock. Quietly one of the most consequential lines in the document: your credit policy — the box, the scorecard, sometimes the pricing grid — gets attached as an exhibit, and changes require lender consent. From that day, your credit box is jointly owned. Every experiment, every new segment, every pricing move in the financed book routes through a lender approval. Negotiate the scope (does consent cover any change, or only material loosening?), the standard ("not to be unreasonably withheld"), and the timeline (deemed approval after X days). Product velocity is a debt term. Nobody tells you that until it's too late.

Forward flow, capacity rights, and the ROFR family — the terms that mortgage your next raise. Lenders want exclusive rights over your origination flow ("100% of eligible receivables shall be financed through this facility") and over your future financings. The flavors, in escalating order of danger:

• Origination exclusivity — everything you originate (that's eligible) goes through this facility. Near-universal ask early. Livable if the facility can actually absorb your volume — pair it mentally with the draw mechanics and tranche gates, because exclusivity to a lender who can't fund your growth is a cage.
• ROFO (right of first offer) — next raise, they get to bid first. Mild.
• ROFR (right of first refusal) — next raise, they see the competing sheet and can match it. Chills the exact competition that lowers your future cost of capital — knowing the incumbent can match, why would a rival spend committee time on you? — and it hands your incumbent a copy of every competitor's best terms.
• Must-accept ROFR — the sharp end: if they match, you're contractually required to take their money. Scoped over your next $50–100M for multiple years, this is a mortgage on your entire capital-markets future, signed as a side term.

The counterweights to negotiate, in rough order of value: exclusivity dies when funding stops — if the lender declines an upsize or fails to fund, their exclusive rights collapse to whatever keeps the current facility whole (this is the single most important escape hatch in the whole document; demand it by name); scope caps (exclusivity up to $X of outstandings, ROFR over one next facility rather than N years); time decay; and carve-outs for products or channels the facility doesn't finance anyway. If a declined ROFR comes with conditions — e.g., you must maintain "anti-adverse-selection" allocation procedures between lenders — fine, that's reasonable; just make sure the procedures are defined, not "to lender's satisfaction."

Prepayment and call protection. The lender's yield needs a minimum life, so expect no-call periods (no prepayment for 12–24 months) and prepayment fees after (say 3% year two, 1% thereafter — or "months of forgone interest" formulations). Legitimate. The traps are in the clocks and carve-outs:

• Resetting clocks. No-call periods that restart at SPV conversion, at each upsize, or at each amendment. Take three upsizes on a resetting clock and you're perpetually inside the no-call window. Fixed clocks, or shorter resets, are negotiable.
• Acceleration interaction. Some drafts make the prepayment fee payable even when repayment is forced by the lender's own acceleration after default. You default, they accelerate, and you owe a fee for "prepaying." Strike it or cap it.
• Whole-vs-partial. "Prepayable only in whole" blocks the most common real-world move: partially refinancing with a cheaper senior lender while keeping the incumbent.
• Same-lender carve-outs. Fee waived if you refinance into the lender's own next facility — a discount for staying, i.e., a tax for leaving. See it for what it is and weigh it with the ROFR terms above.

Your future cheaper facility is the whole point of performing well. Call protection prices that option. Negotiate it like the option it is.

Deal exclusivity, expiry, and expenses — the binding tail of the sheet — covered in Part 4. Re-read those three clauses right before you sign, once more, slowly.

That's the machine. Now put prices on all of it — because the interest rate you just negotiated is maybe two-thirds of what you'll actually pay.