The corporate card question everybody researches, and the one the record says matters

Every piece I have written about a corporate credit card startup question was about getting one. Underwriting, limits, personal guarantees, which providers will look at a company with no trading history. That is the question founders ask, so that is the question I answered.

I pulled the federal complaint file this week to check a different number, and the shape of it stopped me. Getting a card is not where the trouble is, and it is not close either.

This is the corrected emphasis, and the whole correction sits in 1 ratio that I could have pulled at any point in the last 2 years and never did, because the question in front of me was always the application and the application is what I kept answering.

The ratio

In the 12 months to July the Bureau logged 95,642 complaints about credit cards. I sorted them by what went wrong, which is a field the Bureau fills in itself rather than one I imposed.

The Bureau’s own label for the biggest one is “Problem with a purchase shown on your statement”, and it carries 27,556 complaints. That is the largest category by a distance. It is a dispute over something already bought, on a card the person already holds, months or years after any approval decision was made.

The label “Getting a credit card” carries 10,990.

So the thing everybody researches before they apply generates fewer than half as many complaints as the thing nobody researches at all. The ratio is about 2.5 to 1 against the question I had been answering.

The rest of the file fills in around those 2. “Fees or interest” holds 8,956. “Closing your account” holds 6,472. “Advertising and marketing, including promotional offers” holds 3,783, and “Trouble using your card” holds 3,464.

Read that list as a life cycle rather than a menu, because the order in which those 6 things happen to a company is not the order in which anybody researches them, and the gap between those 2 orders is where I had been writing. Approval is 1 moment near the start of it. Everything else is the years afterwards, and the years afterwards are where the volume sits.

How I pulled this, in case you want to check it

I ran the query myself rather than quoting somebody’s summary, and it took about 20 minutes.

The Bureau publishes a search API over the complaint database. I asked it for the product “Credit card” across the 12 months to 1 July, took the aggregation buckets rather than the individual rows, and read the counts by issue and by outcome. No key is required and 1 request answers each question.

One practical note if you repeat it. The API returns 403 to a browser user agent and answers a plain curl without complaint, which took me most of an afternoon the first time I met it and is the kind of detail nobody writes down.

I had assumed the top 2 categories would be rejections and credit limits. They are not, and I would not have found that out by reading anything written about startup cards, including my own.

What that means for a finance function of 1 person

Plan for the dispute process before you plan for the application. That is the practical inversion, and it costs nothing to do early.

At a startup the person who applies for the card is usually the person who will later be arguing about a charge on it, and they will be doing it while also closing a month. A dispute is not a 5 minute task. It needs the original invoice, the delivery evidence, the correspondence with the supplier and a written account of what was agreed, and every one of those is easier to assemble in the week it happens than in the month after.

There is a version of this advice that sounds like bureaucracy and is not. Nobody is asking you to build a process. What you are building is the ability to answer one question quickly, months from now, when somebody who has never met your supplier asks you to prove what you bought and when, and the only evidence that will count is the evidence you kept without knowing you would need it.

So build the habit at the start: a folder per supplier, the invoice saved at the moment of purchase rather than at reconciliation, and the card statement checked against it monthly rather than quarterly. None of that is clever or expensive. It is the difference between a dispute you win in an afternoon and 1 you abandon because assembling the evidence costs more than the charge.

What the outcomes actually look like

Now the part that surprised me, and it cuts both ways. I went in expecting the record to make issuers look bad, and what it mostly shows is a process working slowly rather than an industry behaving badly, which is a duller finding and probably a more useful one.

The Bureau records how each one ended. Across all 95,642, the outcome “Closed with monetary relief” appears 13,664 times. That is 14.3 per cent, and it is the single figure in this piece I would carry into a conversation with an issuer, because it says that roughly 1 complaint in 7 ends with the company deciding it is cheaper to pay than to keep arguing. “Closed with non-monetary relief” appears 13,320 times, and “Closed with explanation”, which means the company answered and nothing changed, appears 67,684 times. Another 821 sit as “In progress” and 153 as “Untimely response”.

I had expected the card file to look worse than the bank account file, and it looks better instead. When I ran the same arithmetic on complaints about closing a checking or savings account, money came back 11.8 per cent of the time. Cards run several points higher than that.

My guess is that this is the dispute machinery rather than better behaviour. Card networks have a chargeback process with defined rights and defined timetables, and a bank closing an account has nothing equivalent pointed at it. Where a process exists, some proportion of the time it produces money. That is a guess and I would not defend it hard, but the difference in the numbers is real either way.

The 6,472 nobody plans for

Sit with 1 number from that list before moving on. “Closing your account” carries 6,472 complaints, and a card account closes in 2 directions.

You can close it yourself, which is the ordinary case and generates paperwork rather than complaints. Or the issuer closes it, and that is the case worth planning for, because a corporate card is not just a payment method by the time it has been in use for a year. It is the thing the cloud provider bills, the thing the ad platform charges, and the thing 40 subscriptions renew against.

I went looking for how much notice an issuer has to give and did not find a general answer in 4 sets of terms. It sits in individual agreements, the terms differ, and I am not going to summarise 6 issuers into 1 sentence that would be wrong for most of them.

What I can say is what the failure looks like from the finance seat. The card stops working overnight. The renewals then fail over the following days rather than all at once, so it arrives as a trickle of service warnings rather than as 1 clear event, and somebody spends a week re-entering card details across 40 vendor portals while explaining to a supplier why the account lapsed.

That is not a credit problem. It is an operational problem with a credit trigger, and the mitigation is boring: a second card on a different issuer, dormant, with the 5 subscriptions you cannot afford to lose duplicated onto it. It costs an annual fee and an afternoon.

The category I keep rereading

A short digression, since it changes nothing about your card and you can skip it without losing anything practical.

Advertising and marketing, including promotional offers, accounts for 3,783 complaints. That is a small share of the file and it is the category I find hardest to look away from, because a complaint in it means somebody believed a specific promise and then read the terms.

For a startup card the promises are rewards rates, sign-up bonuses and category multipliers. I have written before that the rebate is marketing and the transaction data is the actual product. The 3,783 do not prove that. They do tell me that a measurable number of people each year discover the gap between the headline rate and the schedule underneath it, and that the discovery happens after the card is already in the wallet.

Anyway, back to the numbers that matter.

What the file cannot tell you

This is a consumer complaint database. There is no business card flag, and I could not find a way to separate a company card from a personal one inside it. The sub-product breakdown puts 84,751 of the 95,642 under “General-purpose credit card or charge card”, a label that covers a company card and a personal one equally, and no finer split exists in the public data.

So everything above is a consumer file being applied to a business question by inference, and the inference is mine. I think it holds across all 95,642 rows, because the dispute mechanics and the statement are the same product regardless of whose name is on the account, but somebody could reasonably argue that a company disputing a supplier invoice behaves nothing like a consumer disputing a subscription.

I also cannot tell you what share of these complaints came from cards issued to companies with under 50 staff, which is the population this site writes for. Nobody publishes it and I would be guessing at a number rather than a direction.

None of this is advice on your treasury and we are not your accountant. A card with a personal guarantee attached, or one issued against a balance you are about to spend down, has consequences this page does not reach and your own agreement does.

What I would do this quarter

Nothing below is difficult and none of it needs a tool. It needs somebody to decide that the statement is a document to be read rather than a total to be approved, and at a small company that decision is usually made once, quietly, by whoever opens it first.

Read the statement line by line once, in full, this month. Not the total, the lines. Most people never do it and it is where every one of those 27,556 disputes began.

Write down the dispute window before you need it. The right to challenge a charge has a deadline attached, that deadline sits in your agreement rather than in general advice, and it is the 1 fact that decides whether a dispute is possible at all.

Keep the card and the operating account at different institutions if you can. If the card issuer and the bank are the same company, a single bad month arrives at both at once, and the 6,472 account closures in this file are a reminder that issuers close things too.

And do not choose a card on the rebate. On 40,000 dollars a month a 1.5 per cent rebate returns about 600 dollars, which is real money and is still the smallest number in any comparison you will run this year against the cost of a dispute you cannot document.

I cannot tell you whether the ratio holds for company cards specifically, and I have said why above. I also could not establish how many of the 27,556 statement disputes were resolved in the customer’s favour, because the outcome field describes how the company closed the complaint rather than who was right.

What I keep thinking about is that 2.5 to 1. Every guide, including mine, is written for the week you apply, and the file says the trouble arrives in the years you hold it. I wrote about the application because that is when people ask, and answering the question people ask is not the same as answering the question that matters, which is an uncomfortable thing to notice about your own back catalogue.