Why My VPN Crawled on Hotel Wi-Fi but Flew on Mobile Data

The presentation upload reached 18 percent and stopped. I had eleven minutes before a client call from my hotel room in Shanghai, and the VPN was moving data at less than half a megabit per second. I blamed the Wi-Fi signal, moved closer to the door and restarted the upload. It climbed to 7 percent, slowed and froze again. Then I connected the laptop to my phone’s roaming hotspot without changing the VPN server. The same file began moving almost twenty times faster.

The hotel Wi-Fi did not look broken. Local websites opened quickly, the lobby television streamed without trouble, and a test to a nearby server showed more than 200 Mbps.

The slowdown appeared only after the VPN connected.

That made the mobile-data result difficult to dismiss. My phone had a weaker signal than the hotel router, yet the VPN behaved far better through it. The problem was not simply that Wi-Fi was slow and mobile data was fast. The two connections were taking different routes.

This matters to a growing number of visitors. China recorded more than 17.8 million visa-free foreign entries in the first half of 2026, up 30.6 percent from a year earlier. (Gov) More short-stay tourists and business travelers are arriving with the same expectation: connect to hotel Wi-Fi, turn on a VPN and continue using the services they depend on at home.

My requirement was narrower. I needed to upload a 160 MB deck and join one video call without burning through the rest of my roaming allowance.

I first treated it as a weak Wi-Fi problem

I disconnected the VPN and tested the hotel connection again.

Fast.

I turned the VPN back on.

Slow.

I forgot the Wi-Fi network, rejoined it and completed the hotel login page again. Then I restarted the laptop, moved into the corridor and eventually carried it down to the lobby.

The signal improved. The VPN did not.

That ruled out the most reassuring explanation. I could no longer blame the distance from the access point or a badly placed router in my room.

The large VPN provider I was using had been a sensible choice. It had years of public history, extensive support documentation and enough server locations to suggest that one of them must work well.

I chose a nearby endpoint. The upload accelerated for a few seconds, then settled into the same crawl. I tried another country, followed by the provider’s automatic option.

Each connection began with a short burst and then slowed.

On my phone’s hotspot, those same routes performed much better.

A brief public traveler discussion described the same split: VPN connections worked over roaming data but failed or slowed on hotel Wi-Fi. (Reddit) The useful point was simple. Changing the network underneath the VPN could matter more than changing its server.

That was exactly what my own test was showing.

The phone was taking a different road

International roaming often sends traffic back through the traveler’s home carrier before it reaches the wider internet. Hotel Wi-Fi usually enters the local internet directly. (IETF)

Both appear on the laptop as ordinary internet connections, but they can pass through very different routing and traffic-management systems.

The fast hotspot therefore did not prove that mobile data had more bandwidth. It showed that the roaming route avoided whatever was slowing the VPN over the hotel’s local connection.

Research into China’s filtering systems has shown that encrypted traffic can be identified and disrupted by its connection patterns, not only by the website someone is trying to reach. (Usenix) A conventional VPN tunnel may connect successfully and still become painfully slow once sustained traffic begins.

I could not observe the hotel operator’s or network provider’s internal filtering rules. I could see the outcome: the raw hotel Wi-Fi was fast, the established VPN was fast over roaming, and the two together were almost unusable.

Once I understood that, rotating through more server locations felt less useful. The obstacle was not a lack of choices inside the app. It was the route between the app and the open internet.

The hotspot rescued the call, not the evening

With seven minutes left, I stopped troubleshooting and used the phone hotspot.

The deck uploaded in a little over two minutes. I joined the call on time, shared the slides and began presenting.

Then my phone displayed a temperature warning.

It was charging, maintaining an international roaming connection and running a hotspot while sitting beside the laptop’s warm exhaust. The battery percentage was still falling despite the cable.

Halfway through the call, the mobile signal weakened. My camera froze, the audio broke up and the meeting app warned that the connection was unstable. I switched off video and finished with the phone balanced against the window.

The hotspot had saved the deadline, but it had not made the hotel Wi-Fi useful. It had simply let me avoid it.

That distinction became important as soon as the call ended. I still needed to download a recorded product demonstration, review several cloud documents and send a revised deck. Doing all of that over roaming would consume an expensive allowance, drain the phone and keep it tied to the laptop for the rest of the evening.

The mobile result had nevertheless taught me something valuable. I had started by asking which VPN server was fastest. Now I knew that server choice was secondary.

The question was which service could make the network I actually wanted to use—the hotel Wi-Fi—work properly.

The smaller app matched the situation

I opened OnlydogVPN and selected the preset intended for a restrictive network.

The app did not send me back into another round of country, server and protocol guesses. I started the connection and returned the laptop to the hotel Wi-Fi.

My first test was the recorded demonstration, a file large enough to expose the same slowdown quickly.

It began downloading at a steady rate.

I waited for the opening burst to collapse. It did not.

Next I uploaded the revised deck. The progress bar moved without the long flat pauses I had seen earlier. While it continued, I opened the client’s cloud workspace, made two edits and sent the updated link.

The hotel Wi-Fi had not changed. The VPN route through it had.

The smaller app combines traffic obfuscation with an HTTP/3-based transport. The obfuscation makes the tunnel less recognisable as conventional VPN traffic, while the transport keeps data moving more cleanly when the connection becomes uneven. (IETF)

That was all the technical detail I needed. The hotel already had plenty of bandwidth. The service’s job was to stop that bandwidth from becoming useless once the VPN switched on.

In this test, it did.

The proof was putting the phone away

I disconnected the hotspot, placed the phone on the charger and continued working over hotel Wi-Fi.

The product demonstration finished downloading. The revised deck reached the client. A twenty-minute video call remained stable enough that I could leave the camera on.

The most noticeable improvement was not a spectacular benchmark result. It was the disappearance of the routine that had consumed the previous hour:

Change server. Test. Wait. Reconnect. Tether the phone. Watch the battery.

Once the urgent work was finished, I moved to the hotel café. The laptop joined another access point, the call audio paused briefly and then continued. I did not reopen the VPN app.

That smaller moment gave me another reason to keep the service installed. Hotel networks are rarely one clean, permanent connection. Devices move between access points, captive portals expire, and performance changes as more guests return in the evening. A useful travel VPN has to handle that instability without turning every fluctuation into another troubleshooting session.

The limitation did not decide this comparison

The smaller service has fewer locations than the established provider, a shorter public history and fewer independent ratings. Those are real limitations.

They would matter more if I needed an endpoint in one particular country or placed the greatest weight on years of public scrutiny.

Neither explained why the VPN was slow on Wi-Fi but fast on mobile data.

The established provider offered more servers, but the hotel network handled its connection poorly. The hotspot found a better route, but it cost roaming data, battery life and the normal use of my phone. The smaller app made the hotel’s already-fast Wi-Fi useful for the work I still needed to finish.

I had assumed that the better connection would be the one with the larger speed-test number. The split between Wi-Fi and mobile data showed why that comparison was incomplete.

When a VPN crawls on Wi-Fi but flies on mobile data, raw bandwidth is not the deciding factor. What matters is whether the VPN can turn the network in front of you into a route you can actually use.