Card Limits

A spending limit is a policy setting attached to an instrument. It says how much this card may authorise, over what horizon, and it is evaluated in real time by comparing the requested amount against remaining headroom.

That definition is deliberately narrow, because "limit" is used for at least three things: the ceiling on a card, the credit line extended to a company, and the amount a company is willing to spend on a purpose. Only the first is a card limit. The second is a lending decision; the third is a budget.

Limits are also the control most likely to be set once and never revisited. A ceiling assigned during onboarding reflects an assumption about a role at a moment in time; two years later it is either permanently binding or permanently irrelevant.

Headroom

The amount still available against a limit, after settled spend and pending authorisations are deducted.

Authorisation hold

A reserved amount placed against a card when a merchant obtains approval, before the final amount settles.

Limit template

A reusable set of ceilings attached to a role rather than to a named individual.

Limit creep

The gradual, unreviewed accumulation of headroom as individual increases are granted and never reversed.

Several limit types usually apply to one card at once. They bound different dimensions of risk, and knowing which will bind first is the practical skill.

The tightest applicable ceiling wins. A card can sit far below its monthly limit and still decline because one purchase exceeds the per-transaction cap — intended behaviour, and the most common source of confused support requests. Publishing the ceilings to holders removes most of that confusion at no cost.

Understanding the authorisation sequence explains almost every surprising limit behaviour people encounter. It happens in a fraction of a second and involves no human judgement.

1. The merchant requests authorisation

   An amount, a merchant descriptor and a classification code are sent through the acquiring bank to the network and on to the issuer.

2. The card’s rule set is evaluated

   Status, category permissions, merchant scope and every applicable ceiling are checked against the requested amount and current headroom.

3. A decision returns

   Approve or decline. There is no partial state and no explanation carried back to the terminal beyond a generic response.

4. A hold is placed

   On approval, the amount is reserved against the card. Headroom drops immediately, before any money has moved.

5. The transaction settles

   Hours or days later the merchant submits the final amount, the hold is released and the settled figure is recorded.

Why pending holds cause unexpected declines

The gap between steps four and five is where most confusion lives. Between authorisation and settlement a card carries reserved amounts that reduce headroom even though nothing has been paid. Where a merchant obtains approval for an estimated amount, the reserved figure can be larger than the eventual charge and can persist for days.

Hotels, vehicle hire and fuel routinely work this way, because the final amount is unknown when the card is presented. A card can therefore appear to have room while behaving as though it does not — the single most common explanation for a decline that "should not have happened".

• Expect headroom to reflect pending authorisations, not only settled transactions
• Size travel-related ceilings with reservation behaviour in mind rather than the expected final cost
• Show pending items in any consumption figure you publish, so the number is not optimistic

The same timing gap explains why card consumption and posted expense reports disagree, which surfaces at month end rather than at the terminal. Expense management covers how those differences are reconciled.

These two are conflated constantly, and the distinction has real consequences for who can change what.

The practical consequence: raising a card limit is an internal decision you can make in a minute, while the total capacity of the account is not something you configure at all. Substantial account capacity also says nothing about whether an individual card should carry a high ceiling — that remains a delegation question. Brex Credit Card and Brex Business Credit Card cover credit and charge settlement, and corporate card vs credit card works through why the terms imply different arrangements.

The most consequential structural decision is whether ceilings attach to roles or to individuals. Individual ceilings feel responsive and produce, within a year, a set of numbers nobody can explain and nobody dares change.

Templates make the ceiling a property of the job. Someone joining inherits the template for their role; a decision that field engineers need more headroom is made once and applied everywhere instead of propagating through forty conversations. It also makes the estate legible: a reviewer can ask whether the template is right rather than assessing every card separately.

• Per-transaction and per-period ceilings — the largest single purchase permitted without escalation, and the cumulative spend the job actually involves.
• Category permissions — which classifications the card will authorise, reflecting what the role legitimately needs to buy.
• Escalation route — where an over-ceiling request goes; it should point at the relevant budget owner rather than up a management chain.
• Review cadence — how often the template is reassessed, so ceilings set two years ago do not become permanent.

Available fields vary by platform; the principle — attach to role, not to person — does not. Instrument-side template mechanics are covered on employee cards; the policy questions behind them are covered in employee spending.

The most useful principle available is this: set the smallest limit that avoids weekly exception requests. The asymmetry is the point.

A limit slightly too high costs a bounded amount of unused headroom. A limit slightly too low costs an exception request, an approver’s attention, a delay in someone’s work and — repeatedly — the migration of that spending onto a personal card where you cannot see it. The downside of tightness is larger and less visible than the downside of looseness, which is the opposite of most people’s intuition.

Direction of travel therefore matters. Starting generous and tightening with evidence produces real data, because the spending happens on the card. Starting tight and loosening under pressure produces an exception queue, teaches people the system is an obstacle, and yields no data because the spending never appeared.

Neither column contains an amount, because there is no universal answer. What a limit should be depends on the role, the industry, the spending pattern and the company’s tolerance for a bounded mistake. These observable signals are more reliable than any benchmark, because they come from your own operation.

Some spending is episodic: a conference, an equipment refresh, an annual renewal, a launch. Permanently raising a ceiling for an event that happens twice a year is a common and avoidable mistake, because the elevated ceiling persists for the rest of the year.

If temporary increases are requested constantly for the same reason, that is not an exceptions problem. It is a sizing problem handled one request at a time, and the fix is to change the template rather than keep granting the exception.

Limits only move in one direction unless something deliberately moves them back. Each increase is justified when granted; collectively they accumulate into an estate carrying far more standing authority than anyone chose.

Limit creep is quiet because it produces no symptoms. Nothing declines, nobody complains and no alert fires. It becomes visible only through a scheduled review, and that review is only cheap if each card has a recorded owner and purpose.

• Run the review on a schedule rather than after an incident, since reactive review always happens too late
• Reduce ceilings that have never been approached rather than leaving unowned headroom
• Confirm every temporary increase either expired or was deliberately made permanent
• Treat any card whose owner has left as an immediate closure rather than a review item
• Read the decline log in the same pass, since it is the only honest evidence that limits fit the work

Quarterly suits most organisations. The review is not primarily about fraud — it keeps the program’s picture of itself accurate. Employee cards covers dormant card review from the instrument side.

A hard decline is not a malfunction; in a working program it is the control layer doing exactly what it was configured to do. The problem is that the person experiencing it is usually standing in front of a supplier with no information about why.

Authorisation responses carry almost nothing back to the terminal. The holder sees a refusal and neither they nor the merchant learn whether the cause was a per-transaction ceiling, a category rule or a pending hold. Everything that makes a decline survivable has to be arranged in advance.

1. Publish the rules before they bite

   Tell holders their ceilings, the blocked categories and the escalation route at issuance. A decline against a known rule is comprehensible; an unexplained one feels like a system failure.

2. Notify the holder immediately

   A prompt notification naming the rule that was applied converts a mystery into a fact, and prevents repeated retries that look like fraud attempts.

3. Make escalation fast

   A route that resolves in hours keeps the purchase on the card. A route that takes two days is functionally a block, and people route around blocks onto personal cards.

4. Give holders a fallback that is not their own money

   A team card, a manager-issued virtual card or a rapid temporary increase all preserve visibility. Personal outlay does not.

Tone matters as much as mechanism. Declines treated as compliance incidents make people stop reporting them and start avoiding the card; declines treated as configuration feedback make the program better every quarter. Spending controls covers tuning in detail.

The same limit types exist on both, but they are used almost oppositely, because the two instruments answer different questions.

One asymmetry deserves emphasis. A tight ceiling on a physical card fails loudly but recoverably. A tight ceiling on a virtual card attached to a recurring subscription fails quietly, as a rejected renewal, and the first symptom may be a service interruption days later. Where a vendor increases its charge, an exact-amount ceiling set at signup becomes a scheduled outage. Size recurring credentials with headroom and review them at renewal — see virtual cards and the virtual card guide.

For adjacent reading: budgets covers aggregate ceilings that limits do not replace, employee spending covers the behavioural side, Brex Corporate Card covers the program structure behind per-card configuration, and expense controls covers the transactions that do go through.