The Cost of Waiting: Latency as a Price Nobody Quotes | HL Hunt
The Cost of Waiting: Latency as a Price Nobody Quotes
Funds clear in a few days. Disputes resolve in weeks. Refunds post when they post. None of these delays appear in any price, and they are discussed as operational facts rather than as terms. But waiting has a cost, it is charged in full to whoever is waiting, and its size varies by roughly two orders of magnitude depending on who that is. Four days on $1,200 costs a household with savings 53 cents and a household without a buffer about $70, which is 532% annualized for the privilege of waiting on money it already owns.
What you'll learn
The core thesis
Every financial process has a queue. Funds clear in one to five days. A disputed charge resolves in weeks. A security deposit comes back within a statutory window. An underwriting decision takes minutes or takes a fortnight. A refund posts when it posts.
None of these delays appear in any price. They are described as operational facts, the way weather is an operational fact, and they are negotiated over about as often.
Our position is that latency is a price, that it is charged in full to the party waiting, and that its real cost varies by roughly two orders of magnitude depending on who that party is. A five-day hold on $1,200 costs a household with idle cash exactly nothing. It costs a household that will overdraft twice about $70. Same product, same disclosure, same delay, and a cost ratio that is effectively infinite.
This is not a claim that delays are imposed maliciously. Most exist for defensible reasons: settlement finality, fraud windows, investigation time, batch processing. The claim is narrower and harder to argue with. The benefit of the delay accrues to the institution, the cost accrues to the customer, and because the cost is never quantified it never enters the decision about how long the delay should be.
What four days actually costs
Take $1,200 held four days longer than it needed to be, and price it against what the household would otherwise do:
| Household position | What the gap is filled with | Cost of the four days |
|---|---|---|
| Idle cash available | Nothing | $0.00 |
| Savings available | Forgone interest | $0.53 |
| Revolving a card at 24% | Card balance | $3.16 |
| No buffer | Two overdraft fees at $35 | $69.96 |
The bottom row is the one worth converting into a rate, because a fee is a price and a price over a period is an interest rate.
Per our overdraft analysis, this is the mechanism by which timing failures get priced as though they were credit failures. The household did not borrow anything. It waited for its own money and was charged for the wait by a third party.
The same calculation runs through every other queue in the system. A $340 charge disputed and not provisionally credited leaves the household short for up to 45 days; bridging that at the rates available to someone in that position costs roughly $168, or half the amount in question. Per our contest threshold analysis, that is frequently more than the dispute is worth pursuing, which is one of the reasons small disputes go uncontested.
Why the cost is regressive by construction
The cost of waiting is the amount, multiplied by the days, multiplied by the waiting party's marginal cost of funds. The first two terms are set by the institution and are identical for everyone. The third term is inversely related to wealth.
That produces a property worth stating plainly: a uniform delay is a steeply regressive charge. Applying the same five-day hold to every customer is not neutral treatment. It is a charge of zero to some customers and several hundred percent annualized to others, levied by a rule that looks even-handed on its face.
Three consequences follow.
- Faster access is sold as a premium feature to the people for whom it is worth the most, which is a rational product decision and also a way of charging the most desperate customers for the removal of a cost the institution created.
- The households that most need speed have the least ability to buy it. Per our buffer analysis, the absence of a buffer is what makes the delay expensive and also what makes the expedite fee unaffordable.
- Nobody totals it. Per our correlation analysis, the cost surfaces as overdraft fees, late fees, and emergency credit, each attributed to its own cause, none attributed to the wait that produced it.
A falsifiable claim
If latency costs are real and regressive, then shortening a hold should reduce fee incidence among low-balance customers substantially and among high-balance customers not at all. Any institution that has changed a funds availability policy can test this directly: compare overdraft and late fee incidence before and after, split by average balance. If the reduction is uniform across balance bands, the mechanism proposed here is wrong. If it is concentrated in the bottom band, the delay was functioning as a charge on exactly the customers least able to absorb it, and the case for treating latency as a priced term rather than an operational fact is established empirically rather than argumentatively.
The asymmetry in every direction
Set the common queues side by side and a pattern appears that is difficult to explain by operational necessity alone.
| Flow | Typical speed | Who benefits from the timing |
|---|---|---|
| Money leaving the customer | Immediate | The institution |
| Money arriving to the customer | One to five days | The institution |
| A charge posting | Immediate hold | The institution |
| A refund posting | Three to ten days | The institution |
| A payment counted as late | On the date | The institution |
| A payment counted as received | On processing | The institution |
Every row points the same direction, and that consistency is the finding. Individually each has a technical explanation. Collectively, a set of independent operational constraints that all happen to favour the same party is not a set of independent operational constraints.
Per our authorization analysis, the hold-versus-settlement asymmetry is genuinely structural: an authorization is designed to be instant and a settlement is designed to be final, and finality takes longer. That defence is real for card rails. It does not explain why a refund takes longer than a charge on the same rails, and it does not explain the payment-timing rows at all.
The case for the delays
The strongest arguments on the other side are not about convenience, and dismissing them makes the analysis worse.
- Fraud windows are real. Instant irreversible transfer is the environment in which per our push payment fraud analysis the victim has no recourse at all. Some delay is the only remaining point of intervention, and faster rails have demonstrably increased certain fraud losses.
- Settlement risk is real. Making funds available before settlement means advancing credit, and an institution doing that at scale needs capital and pricing for it. Availability is a lending decision wearing operational clothes.
- Investigation takes time. A dispute resolved in an hour is a dispute not investigated, and per our chargeback analysis the party on the other side has interests too.
- Reversal is worse than delay. Funds released early and clawed back later can leave a household worse off than a straightforward wait.
The first point is the serious one, and it means the goal is not zero latency. It is that the length of each delay should be set with the customer's cost included in the calculation, rather than set by whatever the operational constraint permits and then described as fixed. The current arrangement optimizes a cost that is measured against a cost that is not.
Where provisional credit changes the analysis entirely
The dispute case is the cleanest illustration, because a remedy already exists. Provisional credit during an investigation moves the cost of the waiting period from the customer to the institution while the question is being settled, without shortening the investigation at all. That it is standard in some contexts and absent in others is a policy choice, not an operational limit, and it is the single clearest place where who waits is decided by rule rather than by necessity.
What an institution can do about it
For a lender or a processor, treating latency as a priced term rather than a constant produces a short list of concrete moves.
- Measure it. Compute the fee incidence caused by your own hold periods, split by customer balance band. Almost nobody has this number and it is derivable from data already held.
- Differentiate on risk, not on willingness to pay. A hold length set by the risk the specific transaction presents is defensible. A hold length uniform across customers and shortened for a fee is a charge for the removal of a self-imposed cost.
- Make refunds as fast as charges. Where the rails allow it, the asymmetry is a choice, and it is among the most visible ones a customer experiences.
- Provisional credit as the default in disputes, per our compliance analysis, with the burden of the waiting period falling on the party that can absorb it.
- Count payments as received when sent, not when processed, which costs almost nothing and removes a class of late fees with no credit content, per our timing analysis.
- Publish the timings. A stated maximum resolution time is a commitment. An unstated one is an unbounded call on the customer's cash.
Item five is the cheapest item on the list and the most overlooked. A payment initiated on the due date and posted two days later is recorded as late, generating a fee and often a reporting consequence, for a customer who did the thing they were asked to do on the day they were asked to do it. Per our recency analysis, that record then prices their credit for years.
Scenarios and what we are watching
| Scenario | What we would expect |
|---|---|
| Instant rails become the default | Overdraft incidence falls among low-balance customers specifically, and certain fraud losses rise, and both effects are real |
| Expedite fees proliferate | Latency is confirmed as a priced product, sold hardest to those for whom the underlying delay costs the most |
| Provisional credit becomes standard across dispute types | The largest single improvement available, and the one that requires no technical change at all |
| Nothing changes | The cost continues to be recorded as overdraft revenue, late fee revenue, and short-term credit demand, with the wait that caused it recorded nowhere |
The second row is the one to watch most carefully, because it is the market's own admission that the thesis here is correct. An institution that charges for faster access has priced the delay. It has simply priced it as a service it is providing rather than as a cost it was imposing, and the difference between those two descriptions is the entire argument.
Frequently asked questions
Between nothing and about $70 for four days on $1,200, depending entirely on the customer. The overdraft case annualizes to roughly 532%.
Because the cost is the amount times the days times the customer's cost of funds, and only that last term varies. The same rule charges nothing to some and hundreds of percent to others.
Fraud windows and settlement risk are real justifications. They do not explain why refunds are slower than charges on the same rails, or why payments are late on the due date but received on processing.
Provisional credit as the default in disputes. It shortens no investigation and requires no technical change; it only decides who bears the wait.
Key takeaways
- Latency is a price: the amount, times the days, times the waiting party’s marginal cost of funds.
- Only the third term varies, so a uniform hold is a steeply regressive charge that appears in no disclosure.
- Four days on $1,200 costs 53 cents with savings and about $70 without a buffer, which annualizes to roughly 532%.
- A $340 dispute without provisional credit can cost about $168 to bridge over 45 days, which is why small disputes go uncontested.
- Money leaves instantly and arrives slowly, charges hold instantly and refunds post slowly, payments are late on the date and received on processing: every asymmetry points one way.
- Fraud windows and settlement risk are genuine justifications for some delay, but they do not explain the refund or payment-timing asymmetries.
- Provisional credit is the largest available improvement and requires no technical change, because it only decides who bears the wait.
- Counting payments as received when sent rather than when processed removes a class of late fees with no credit content, at almost no cost.
This analysis is educational and does not constitute financial or legal advice. All figures are stylized illustrations using assumed amounts, fees, hold periods, and funding costs, and do not describe any particular institution or customer. Funds availability requirements, dispute investigation timelines, provisional credit obligations, and payment crediting rules vary by product, by payment rail, and by jurisdiction, and change over time.