A credit limit does two jobs at once. It is an exposure ceiling, the maximum loss you are willing to absorb if this customer fails tomorrow, and it is a review trigger, the tripwire that forces a fresh look at the account before exposure grows further. Teams that treat the limit as only a ceiling set it once and forget it; teams that treat it as only a formality raise it reflexively whenever an order bounces. Both habits produce the same outcome: the limit stops carrying information.
The harder problem is that there is no single correct limit for a customer. Each method answers a different question. Anchor to the customer's balance sheet and you answer: how much can they safely owe anyone? Anchor to their purchase volume and you answer: how much will they actually need from us? Anchor to trade data and you answer: how much do peers already trust them with? Good limit-setting usually computes two or three of these and takes a deliberate position between them.
This guide works through the common methods with real arithmetic, shows how security instruments like guarantees and credit insurance change the answer, covers temporary and seasonal limits, and finishes with the operational side: utilization monitoring, when to raise or lower, and how to communicate limits without turning every one into a negotiation.
Why limits exist: ceiling and tripwire
Start with the ceiling function. Your total portfolio can absorb a certain amount of single-account loss before it dents the P&L in a way that matters. If your company earns an 8% operating margin, a $100,000 write-off requires $1.25 million of replacement sales just to get back to even. The limit on any single account is, at bottom, a statement about how much of that replacement-sales burden you are willing to risk on one counterparty.
The tripwire function is less discussed but operationally more valuable. A well-set limit gets approached in the normal course of a growing relationship, and each approach forces the question: has anything changed since we last looked? Fresh bureau pull, current aging behavior, maybe updated financials. A limit set so high it is never touched removes this checkpoint entirely; the account can deteriorate for two years without anyone being forced to look. This is why 'set it generously so sales never complains' is a risk decision disguised as a convenience.
One consequence of the dual role: the right limit is often lower than the maximum you could defend. If a customer's financials could support $200,000 but their realistic requirement is $60,000, a $75,000 limit protects you better and costs the relationship nothing, because the tripwire fires close to real usage instead of at a theoretical maximum.
Method 1: Percentage of tangible net worth
The classic balance-sheet method: set the limit at 5 to 10% of the customer's tangible net worth, that is, total equity minus intangibles like goodwill, capitalized software, and intercompany receivables of doubtful substance. The logic is that no single trade creditor should represent more than a modest slice of the cushion available to absorb losses.
Worked example: a distributor applicant shows total equity of $2.4 million, including $600,000 of goodwill from an acquisition. Tangible net worth is $1.8 million. At 5%, the indicated limit is $90,000; at 10%, $180,000. Where you land in the band depends on quality signals: pick the low end when leverage is high (debt-to-equity above 2.0), profitability is thin or negative, or the bureau file shows slowness; move toward the high end for clean payers with stable earnings.
The method's weakness is staleness and availability. Private-company financials arrive annually if at all, and a net-worth figure from 14 months ago may describe a different company. Use this method as the anchor for larger exposures where you can insist on statements, and pair it with a trade-data cross-check.
Methods 2 and 3: Working capital and trade-reference benchmarks
A close cousin anchors to working capital instead: limit at 10 to 20% of (current assets minus current liabilities). This asks a slightly different question, not 'what cushion exists?' but 'what near-term liquidity exists to pay bills like ours?' Example: current assets $3.1 million, current liabilities $2.2 million, working capital $900,000. At 10%, the indicated limit is $90,000; at 20%, $180,000. Prefer this method for customers whose balance sheets are asset-light but liquid, and be suspicious of it when inventory dominates current assets, because inventory that is not selling pays no invoices.
The trade-reference method skips financials entirely and asks what other suppliers already extend. Collect high-credit figures, the largest balance each reference has carried for the customer, from your references and the bureau's trade lines, then take the median as your starting point. Example: five trade lines report high credits of $18,000, $25,000, $32,000, $45,000, and $110,000. The median is $32,000; that is a defensible starting limit. Use the median, not the mean, because one large supplier (the $110,000 line, likely their primary vendor) drags the average toward an exposure that may reflect a decade-long relationship you do not have.
Agency-suggested limits, the figure a bureau model produces, are essentially an automated version of this: trade-data driven, useful as a sanity check, and best treated as one input rather than the answer. When your own calculation and the agency figure diverge by more than 2x, that gap is information; find out why before choosing.
Method 4: Requirement-based limits
The requirement-based method flips the question from 'what can they support?' to 'what will they actually need?' The formula: expected monthly purchases x expected days to pay / 30, times a safety factor of 1.2 to 1.5 to accommodate order lumpiness and timing overlap.
Worked example: sales projects a customer will buy $40,000 per month on net-30 terms, and customers in this segment typically pay around day 40 (check your own DSO by segment rather than assuming terms equal behavior). Peak natural exposure is $40,000 x 40 / 30 = $53,300. Apply a 1.3 safety factor: $69,300, so set the limit at $70,000. If the same customer paid promptly at 30 days, the calculation gives $40,000 x 30 / 30 x 1.3 = $52,000, a meaningfully lower number, which illustrates a useful property: this method automatically charges slow payers with higher required limits, making their behavior visible.
Requirement-based limits shine for new relationships with real purchase forecasts and for the many customers who will never give you financials. Their weakness is the input: sales forecasts of customer purchases run optimistic. Anchor to a trailing figure once history exists, for example the highest three-month rolling average of actual purchases, and treat the forecast-based number as provisional for the first two quarters.
Method 5: Blended scorecard-driven limits
Mature credit operations stop choosing a single method per account and instead let a scorecard choose the aggressiveness. The scorecard combines weighted inputs, bureau score, financial ratios where available, years in business, trade payment behavior, industry risk, into a score that maps to a risk band, and each band carries a limit rule.
A representative structure: Band A (lowest risk) gets the greater of the requirement-based number or 10% of tangible net worth, capped at your single-account maximum. Band B gets the requirement-based number, capped at 5% of tangible net worth or the trade-median, whichever is lower. Band C gets 50 to 75% of requirement, forcing a review at the first real growth moment. Band D gets no open terms: cash in advance, credit card, or secured only. The bands do the judgment once, centrally and in writing, instead of leaving each analyst to re-derive aggressiveness file by file.
The practical benefits are consistency and auditability: when a limit is questioned later, the answer is 'Band B rule applied to these inputs' rather than a reconstruction of one analyst's reasoning. The maintenance cost is real, the scorecard weights and band cutoffs need annual validation against actual loss experience, but for portfolios above a few hundred active accounts, the consistency gain dominates.
Adjusting for security: guarantees, liens, insurance, and letters of credit
Security instruments do not change the customer's risk; they change your loss given default, and the limit should respond to that, deliberately and by written rule rather than ad hoc generosity.
A personal guarantee from an owner with verifiable assets typically justifies moving one risk band up or extending 25 to 50% above the unsecured calculation, but be honest about collection reality: a guarantee is a lawsuit, not a lockbox, and its value depends on the guarantor's assets staying findable. A standby letter of credit is the opposite extreme: exposure covered by the LC is effectively bank risk, so a $100,000 LC plus a $50,000 unsecured calculation supports a $150,000 limit cleanly. Credit insurance shifts the limit question to the insurer, your limit becomes the insurer's approved coverage plus whatever deductible and uninsured share you are willing to self-fund. In construction and materials supply, statutory security like mechanics lien rights can justify materially higher limits than the balance sheet alone would, provided your notice deadlines are actually being met; lien rights you fail to perfect are worth zero.
Whatever the instrument, record the security against the limit in your system with its expiry. The most common failure mode is not mispricing the security; it is an LC that quietly expired eight months ago while the inflated limit lived on.
Temporary limits, seasonality, and raising or lowering
Not every exposure peak deserves a permanent limit. For one-off large orders, use a temporary limit: a documented increase with an expiry date, typically 60 to 90 days, that reverts automatically. This keeps the standing limit meaningful while accommodating the real order. The discipline that matters is the automatic reversion; a temporary increase that requires someone to remember to lower it is a permanent increase with extra steps.
Seasonal businesses deserve a seasonal limit structure set in advance: a garden-products retailer might carry a $50,000 base limit and a $140,000 limit from February through June, agreed at annual review. Building the season into the limit is better than a scramble of emergency approvals every spring, and it lets you demand something in return, current financials each January, for instance, as the standing price of the seasonal uplift.
For increases, require a trigger plus evidence: utilization above 80% for two consecutive months plus clean payment behavior earns a review, not an automatic raise; the review pulls a fresh bureau file and, above your financial-statement tier, updated statements. For decreases, act on deterioration signals early: payments sliding 15+ days beyond terms, bureau alerts, NSF events, or utilization collapsing (which can signal they have moved their business or their bank pulled their line). Lower limits gradually where you can, and pair the decrease with a conversation; a silent 60% limit cut discovered at order entry converts a risk decision into a relationship crisis.
Monitoring utilization and communicating limits
A limit program is only as good as its monitoring. Track utilization, current balance plus open orders as a percentage of limit, across the portfolio monthly, and flag two tails. Accounts persistently above 80% are review candidates: either the relationship has outgrown the limit (raise it on evidence) or exposure is drifting up on a deteriorating account (lower it or hold orders). Accounts persistently below 20% carry silent risk of a different kind: a $200,000 limit on an account that owes $9,000 is $191,000 of pre-approved exposure that nobody will look at before it fills. Right-size these downward at annual review; unused headroom is not customer goodwill, it is unmonitored risk.
Communicate limits differently to sales and to customers. Sales should see limits, current utilization, and hold status in the tools they already use, so a rep discovers the problem before keying a doomed order, and the escalation path should be published: who to call, what evidence will move the number. Whether to tell customers their limit is a policy choice; many suppliers disclose it on request rather than proactively, which avoids anchoring every negotiation, but always tell a customer when they are approaching it. An order blocked without warning costs more goodwill than any limit number ever will.
Finally, close the loop with outcomes. Once a year, compare write-offs and worst delinquencies against the limits and methods that produced them. If your losses cluster in accounts where the trade-median method set the number, or in Band B of the scorecard, that is your recalibration agenda. Limit methodology is a model, and models earn trust only through validation.
About the author
SGUTTI · Founder, EFILOS
SGUTTI is the founder of EFILOS and the architect of SCREDIT, the trade-credit operating platform. He writes about credit operations, financial statement analysis, and receivables management based on the workflows SCREDIT is built around.
Connect on LinkedIn