That message — "an error has occurred and we are looking into it. please refresh and try again. if the problem persists please contact customer support" — is a server-side apology, not a diagnosis. It means the service failed to complete your request and is telling you, in polite boilerplate, that the fault is probably on its end. Your job is to rule out the three things you control (connection, browser state, account state), retry once cleanly, and then decide whether to contact support. If a single hard refresh clears it, the error was transient. If it survives a private window and a different network, stop retrying and start reporting.

The wording matters. "We are looking into it" implies a known or suspected incident on the provider's side. "Please refresh and try again" is the catch-all fix for stale sessions and half-loaded pages. "If the problem persists" gives you the timeline: one clean retry, then escalate. This pattern appears across banking apps, e-commerce checkouts, cloud dashboards, government portals, and game launchers. The message is generated by the site's error-handling layer, not by your device, and it almost never tells you the actual error code — which is why knowing what to check yourself matters.

One boundary up front: this article helps you distinguish a problem you can fix locally from one you can only report. It doesn't replace advice from the specific service's support team, and for anything involving money, identity, or legal deadlines, you should follow that provider's official escalation process. For urgent account security issues (suspicious charges, unauthorized access), contact the provider's fraud line directly rather than submitting a web form.

What the Refresh-and-Try-Again Message Actually Signals

The phrase is a generic error page, usually served with an HTTP status code in the 5xx family or a 4xx that the site has dressed up to sound friendlier. A 500 Internal Server Error means the server itself broke while handling your request. A 503 Service Unavailable usually means the service is overloaded or in maintenance, and many providers pair it with a "try again" instruction plus a Retry-After hint. Cloudflare's documentation for its error pages distinguishes these clearly, and the distinction changes what you should do: 5xx means wait, 4xx means check your input, your session, or your permissions.

The "we are looking into it" half is a status claim, and you can test it. Providers that run public status pages often flag incidents in real time — Amazon Web Services maintains its Service Health Dashboard, Microsoft publishes Azure status, and most major SaaS products run a status subdomain. If an incident is posted, your local troubleshooting is wasted effort; wait it out. If no incident appears while the error is hitting you repeatedly, the problem may be scoped to your account, your region, or a single shard of the service.

Stale sessions are the single most common local cause. Session tokens expire, cookies get corrupted by an interrupted write, or your browser holds an old version of a form that the server has since rotated. That's why "refresh and try again" appears in the message at all. A plain refresh can resubmit the same bad state; a hard refresh, which bypasses the cache, or a private window, which starts with no cookies, is the cleaner test.

How long to give it depends on the failure type. A transient 5xx at a major provider is usually resolved within minutes, though the widely reported outages of the past decade show that "minutes" can stretch to hours when the failure is in a core dependency like DNS or authentication. Build your own timeline: one clean retry after a hard refresh, one attempt in a private window, one attempt on a different network. If all four fail inside 15 minutes, it's not you.

Three Retry Conditions: Connection, Browser State, Account State

Before you touch anything, classify the failure. There are only three places a request can die on your side, and each has a different fix. Connection problems show up as pages that never load at all, or load without styling. Browser-state problems show up as errors that follow you across tabs but vanish in a private window. Account-state problems show up as errors that appear only after you sign in, or only on one specific action like checkout or password reset.

Symptom Likely layer Fastest test Fix if confirmed
Other sites load fine; this one errors instantly Server-side Check the provider's status page Wait, then retry
Error vanishes in a private window Browser state Private/incognito window, same action Clear cookies for that site only
Error appears only after sign-in Account state Sign out, sign in, retry the action Reset session, contact support
Error follows you to another device on the same Wi-Fi Network or server Switch to mobile data Restart router, retry
Error only on one form or one button Application bug Try the same action in another browser Report exact steps to support

The table is a decision tool, not a checklist. Run the test that matches your symptom first; working down the list in order wastes the most useful signal, which is whether the error is reproducible in a clean environment. A private window is the cheapest isolation test available because it strips cookies, extensions, and cached files in one step.

Connection problems deserve a specific note. A Wi-Fi network that loads ordinary pages can still break a single service if that service depends on a port, DNS record, or content filter your network handles oddly — a common pattern on corporate networks and in some countries' ISP-level filtering. Switching to mobile data is a one-minute test that rules out the entire local network. If the error disappears on mobile data, your router, DNS settings, or network filter is implicated.

Account-state failures are the ones that survive everything and get misread as server outages. A permissions change, an expired subscription, an unverified email, or a locked account can all surface as a generic "error occurred" page because the site doesn't want to leak which check failed. If the error appears only when signed in and only on one action, suspect permissions or account status before suspecting the server.

A Four-Test Sequence That Separates Local From Server Faults

Run these in order, and stop as soon as one changes the result. Each test takes under a minute. The sequence is designed so that every test after the first isolates one variable, which means a pass or fail is informative rather than ambiguous.

1: Hard refresh the page. On Windows and Linux, Ctrl+F5 or Ctrl+Shift+R. On macOS Chrome and Firefox, Cmd+Shift+R. On Safari, hold Shift while clicking reload. A hard refresh bypasses the cached copy and re-fetches the page and its assets from the server. 2: Repeat the exact same action in a private window. This strips cookies, cached storage, and most extensions. If the error disappears here, your normal browser profile is the problem. 3: Switch networks. Turn off Wi-Fi and retry on mobile data, or use a phone hotspot. If the error disappears, your local network or DNS resolver is involved. 4: Try a second device on the same network, signed into the same account. This separates device faults from account faults: if a different device on the same network fails identically, the account or the server is implicated.

A failed test is more useful than a passed one. Failing test 2 tells you to clear site data, not to keep refreshing. Failing test 3 tells you to check your router's DNS or any content filter, not to reinstall the browser. Failing test 4 while test 3 passes points squarely at your original device's browser profile.

Keep a small log while you run the sequence: timestamp, test, result, and the exact text of any error code shown. Support teams route tickets faster when the first message contains a time window and a reproduction path. If the page shows a reference number or incident ID in small text near the message, copy it — those IDs often map directly to server logs on the provider's side.

When a test passes, don't assume the problem is gone permanently. Intermittent failures that clear on a network switch and return an hour later are usually load-balancer or regional-routing issues on the provider's side. Note the times they recur; a pattern of failures at a consistent hour points to scheduled jobs, peak traffic, or maintenance windows.

What to Check Before You Contact Support

The message tells you to contact support if the problem persists, and it will ask you for information you should prepare in advance. The more precise your first message, the fewer round trips. Most support systems also attach your account automatically, so focus on the failure, not your identity.

Gather these six items before opening a ticket: the exact page or action that triggers the error; the time and time zone of the last three failures; the error code or reference number if one is displayed; the result of the private-window test; the network or country you were on; and any recent change to your account such as a password update or billing change. These six items cover roughly the fields support agents are trained to ask for, which is why they cut resolution time.

Check the provider's status page and its official help center first. Many providers publish incident notices that explain the error and give an estimated resolution window. If an incident is open and matches your symptom, a ticket adds noise; subscribe to the status update instead. If no incident is posted, state that in your ticket — it tells the agent the issue may be scoped to your account.

Write the first message like a bug report. One sentence on what you were doing, one on what happened, one on what you tried, and one on what you need. Include screenshots if the form accepts them, and crop them to the error text plus the page context. Avoid describing your emotional state; agents triage on reproducibility, not frustration.

If the error involves money — a failed payment, a duplicated charge, a checkout that showed success but produced no order — do not rely on the web form alone. Use the provider's phone line or in-app chat and ask for a written confirmation of the ticket number. For a duplicated charge, also notify your card issuer's dispute process if the provider hasn't reversed it within the timeframe stated in your card agreement.

The Failure Case That Defeats Every Retry: The Success-But-Failed Transaction

Here's the scenario that wastes hours: you click Pay or Submit, the error message appears, the page says the action failed — and the action actually went through. The payment was taken, the order was placed, the password was changed, and now your second attempt creates a duplicate. This happens when the server completes the write but the response to your browser is lost to a timeout, a proxy, or a crashed response handler. The error page is truthful about the response and wrong about the outcome.

The practical rule: before retrying any action that changes state — payments, transfers, password resets, bookings, form submissions that create records — check whether the first attempt landed. Look for a confirmation email, a new row in your account history, an updated balance, or an order number. If the first attempt succeeded, do not retry. Contact support and describe the duplicate risk, quoting both the error and the confirmation.

This failure mode is common enough that many payment systems use idempotency keys — a unique token per attempt that lets the server recognize a repeat of the same operation and return the original result instead of processing it twice. Stripe's API documentation describes this mechanism openly. The problem is that not every site implements it across every form, and the generic error page gives you no way to tell. Assume no protection until you've verified the first attempt failed.

Set a personal rule now: any time the error message follows a click that could have moved money or created a record, verify before retrying. The 60 seconds it takes to check your email or account history is cheaper than reversing a duplicate charge, which can take days and, on some card networks, involves a formal dispute process. For password-reset errors specifically, check whether the reset link arrived before requesting another; request three resets in five minutes and some providers invalidate all of them.

In a corporate or shared-account setting, the same failure can appear as an error for one user while the action succeeds for the account. If your team uses a shared billing or admin portal, ask a colleague whether the change shows on their side before assuming the operation failed. This one habit resolves a surprising share of "the system won't let me do anything" tickets.

When a Status Page Confirms the Error Isn't Yours

Public status pages are the fastest way to end your troubleshooting. If the provider has posted an incident that matches your symptom, every local test is a distraction. Status pages are updated by humans or by automated monitoring, and their granularity varies: some list individual services, regions, and features, while others report only a single overall state.

Read the status page critically. A green overall indicator with a known issue listed for a sub-feature means the failure is real but scoped. A yellow or red indicator with a resolution estimate gives you a timeline — check back rather than refreshing. Historical incident reports, which many providers publish after major failures, are worth a glance if the same error has hit you twice in a month; they often name the root cause (a bad configuration push, a certificate expiry, a dependency outage) and the preventive change.

The catch is that not every provider runs a public status page, and some update theirs late. Absence of an incident is not proof the problem is yours. Treat the status page as one input: if it's green and your four-test sequence points to the server, say so in your ticket and include the status page's silence as evidence. Support teams can see internal monitoring that you can't, and a well-framed report helps them escalate.

Regional differences matter here. A service can be healthy globally and degraded in one region because of a local data center, a regional DNS resolver, or a peering issue between your ISP and the provider. Status pages that break down by region, such as those run by major cloud platforms, are more useful than single-indicator pages for this reason. If your status page doesn't break down by region, note your country in the ticket.

Region, Version, and Browser Differences That Change the Fix

The same error message can come from different code paths on different platforms, which is why a fix that works on desktop Chrome may not touch the problem in the mobile app. Web front ends, native mobile apps, and desktop clients each handle errors differently: the web app might show the message after a failed API call, while the mobile app might show it when its cached token fails to refresh. Name your platform in any support ticket.

Browser version matters more than most people expect. Sites routinely drop support for older browser versions, and an unsupported browser can trigger errors on features that depend on modern JavaScript or security headers. If you're on a browser version more than a couple of years old, update it before running the four-test sequence; otherwise you're testing the wrong variable. The same logic applies to mobile operating systems, where outdated OS versions can break secure connections used by banking and commerce apps.

Geographic restrictions are a real but frequently misdiagnosed cause. Some services block access from certain countries or regions for licensing or compliance reasons, and the block often appears as a generic error rather than a clear message. If the error started after you moved, changed VPN, or traveled, test without the VPN and note your location. VPN users should also consider that some services deliberately reject known VPN ranges; the error will persist across every test in the sequence and look like a server fault.

App-specific caches deserve their own step. Mobile apps cache data more aggressively than browsers, and a corrupted app cache can reproduce the error across restarts. On iOS app updates are handled through the App Store; on Android, through Google Play. Clearing the app's cache (not its data, which signs you out) is the standard fix and is available in Android's app settings; iOS requires deleting and reinstalling the app to get the same effect. Both approaches preserve your account because the account lives on the server.

Reporting the Error Without Making It Worse

A good support ticket is short, specific, and free of guesses. State the action, the timestamp, the error text, the results of your tests, and the outcome you need. Don't speculate about the cause; support tools can correlate your account with server logs, and confident wrong guesses slow triage.

Attach evidence that matches the failure. A screenshot of the error page with the browser's address bar visible is more useful than a cropped image of the text alone, because it shows the exact URL and page state. For mobile apps, a screen recording of the steps that trigger the error gives support more than any written description. If the failure involves a payment, include the last four digits of the card and the amount — never the full card number.

Set a follow-up expectation in the ticket itself. Ask for a ticket reference, an estimated response time, and what evidence would help. Requesting a written response rather than a phone callback keeps a record you can forward if the issue escalates or if you need to dispute a charge. If the first reply is a template that doesn't address your tests, reply once with a one-line summary of the tests and ask for escalation; repeated template replies are a signal to switch channels to chat or phone.

What not to do: don't send multiple duplicate tickets, which fragment the record; don't paste your password or full card details in the message, because support never needs them; and don't delete the app or clear all data as a first step if you're mid-transaction, since that can destroy the local evidence. Preserve the state until you've verified whether the action succeeded.

FAQ

Does "we are looking into it" mean the problem is on the website's side?

Usually yes — the phrase is a server-side status statement, and the site is admitting it failed to complete your request. But the same template is sometimes served for account-specific failures the site won't explain, such as permission errors or locked accounts. Test the server theory by trying the same action in a private window and on a different network. If the error survives both, treat it as server-side and report it.

Can I safely retry a payment or transfer after this error?

Only after you've confirmed the first attempt didn't complete. Check your account history, email confirmations, and balance before clicking again — a lost response can hide a successful charge, and a retry can duplicate it. If you find a confirmation, don't retry; contact support with the confirmation details. If no confirmation exists after 15 minutes, retry once and keep the timestamp.

Why does the error disappear when I use incognito mode?

Incognito strips cookies, cached files, and extensions, so if the error vanishes there, something in your normal browser profile is corrupted — most often a stale session cookie or an extension interfering with the page. Clear site data for that one domain and try again in a normal window. Don't clear all browsing data; clearing the single site's cookies is targeted and keeps you signed in elsewhere.

Is a VPN causing this error?

It can. Some services reject traffic from known VPN ranges or flag a sudden country change as suspicious, and they show a generic error rather than a clear block message. Turn the VPN off, retry once, and compare. If the error clears without the VPN, the service is filtering VPN traffic, and no amount of refreshing will fix it while the VPN is on.

How long should I wait before contacting support?

Fifteen minutes, assuming you've completed the four tests. That window is long enough for a transient server fault to clear and short enough to escalate before a time-sensitive task is missed. For anything with a deadline — a bill, a booking, a filing — contact support immediately after the first failed retry and state the deadline in the ticket.

What does the reference number on the error page do?

It maps to a specific server-side log entry, which is why support asks for it. Copy the number exactly, including any leading letters, and include it in your first message. Without it, support has to search by time and account, which adds a step. With it, an agent can often see the failure in seconds.

Why does this error show up in the mobile app but not the website?

Mobile apps cache tokens and data more aggressively than browsers, so a stale token can fail inside the app while the website's fresh session works. Log out and back in inside the app first; if that fails, clear the app's cache on Android or reinstall on iOS. Name your app version and OS version in any ticket, because the failure may be version-specific.

Should I keep trying if the status page shows no incident?

Try the four-test sequence once more, then report. A green status page doesn't rule out a fault scoped to your region or account. Include the result of each test and the timestamp of the status page check in your ticket, and note that no incident was posted. That combination — reproducible failure plus a quiet status page — is exactly the signal support needs to escalate.

This article is general educational information and doesn't replace guidance from the specific service's support team, your bank, or your employer's IT department. For issues involving payments, identity, or legal deadlines, follow the provider's official escalation process and, where money has moved incorrectly, your card issuer's dispute procedure. If you suspect fraud or unauthorized account access, contact the provider's fraud line and your local consumer protection authority directly.