Degradation since 31/12/24 when 2 Openreach vans visited neighbour

[#1]

Interesting - the bad drops seem almost hourly, on the hour, which is suspicious, although your max appears to be steadily falling. As I said, I'm not a telecoms engineer, but something has suddenly started to affect your connection.

I'd wait and see what the situation is when you next speak to @Windings .

[#2]

Understand your reluctance.  All I can suggest at the moment is as I said above.

[#3]

It could be worth trying a couple of other speed test benchmarks just to see if they give the same results, e.g.

• CloudFlare
• ThinkBroadband 
• Ookla 

[#4]

---

@jab1 wrote:

Interesting - the bad drops seem almost hourly, on the hour, which is suspicious,

---

Don't forget that the graph points are only every 15 minutes, so don't give a continuous picture of the download speed.

[#5]

Noted.

[#6]

Thank you for highlighting this @corringham 

The speed tests occur every 15 mins, offset to the following times

xx:08

xx:23

xx:38

xx:53

The graph attempts to show change between these times, however it's artificial. Useful for tracking trends over long periods, but not necessarily accurate at a single point in time except those above.

[#7]

Have been doing some more investigation.

By default, Ookla Speedtest-cli selects the server based on certain factors which I don't yet understand (ip location or some other factor?)

It seems that within minutes of  the Openreach engineers arriving, the speed tests started selecting different servers and experiencing poor results.

I've found that hardwiring Speedtest-cli to a specific server (Iomart in this example for now), gets the expected full speed.

No idea what would have caused this, need to do more investigation. I've added the server id to the dataset - the coloured dots/lines on the chart.

The change on 31/12 is quite visible, along with the improvement since hardwiring to a specific server id yesterday,

I'll see if I can try and identify the server id(s) that's causing the poor speed tests. Maybe something changed in terms of routing, which is causing Ookla to provide a different set of servers.

Still not sure whether the Openreach visit had anything to do with it, just seems remarkably convenient timing.