Key takeaways, 2026
- Latency is round-trip time, not speed. Under 20 ms is excellent, above 100 ms hurts VoIP, payments and checkout conversion.
- FTTP fibre delivers 5 to 15 ms in 2026, against 15 to 25 ms for FTTC and 30 to 50 ms for 4G.
- Every 100 ms of added latency on an e-commerce checkout cuts conversion by 7% to 10% (Akamai, Aberdeen Group).
- Fix it without rewiring: FTTP upgrade, CDN, UK cloud region, SD-WAN, router QoS. A leased line is only worth it with an SLA.
10 ms
FTTP fibre
Median RTT to a UK datacentre.
7%
Conversion lost
Per +100 ms on checkout.
150 ms
VoIP red line
One-way call quality drops sharply.
78%
UK FTTP coverage
Openreach footprint, May 2026.
What latency actually is, and why it is not "speed"
Latency is the round-trip delay, in milliseconds, between your device firing off a request and the response arriving back. It is governed by physics (the speed of light through fibre), by the number of network hops, and by the queuing behaviour of every router along the path. Bandwidth, by contrast, is how many packets per second can flow once the connection is established. A 1 Gbps line with 120 ms of latency will feel slow, while a 100 Mbps line with 8 ms of latency will feel snappy.
In practice, latency is the metric that decides whether a Teams call breaks up, a Stripe payment goes through on the first tap, a Salesforce dashboard loads in one second or eight, and whether a customer abandons their basket. It is also the metric your broadband router never shows you, which is why most UK businesses have no idea what theirs is.
Latency benchmarks, UK 2026
The table below shows the typical round-trip time we observe from a UK business premises to a London or Manchester datacentre in 2026, by access technology. Numbers vary with distance to the local exchange, congestion and the specific peering of your ISP, but the bands are stable across our 2026 measurements.
| Access technology | Typical RTT | Best case | Suitable for |
|---|---|---|---|
| FTTP (full fibre) | 5 to 15 ms | 3 ms | Everything : VoIP, trading, e-commerce, video |
| 5G standalone | 5 to 20 ms | 4 ms | Fixed wireless, pop-up sites, failover |
| FTTC (fibre to the cabinet) | 15 to 25 ms | 12 ms | General office, single VoIP line, light cloud |
| Starlink Business | 25 to 50 ms | 22 ms | Rural sites, construction, remote depots |
| 4G LTE | 30 to 50 ms | 25 ms | Temporary site, vehicle, mobile workers |
| ADSL (legacy copper) | 40 to 80 ms | 35 ms | Migration target : replace this |
Source: Selectra UK measurements, January to May 2026, n=2,400 SME sites.
What latency actually costs you in pounds
The financial impact of latency is hardest to see and easiest to ignore, but it is well-quantified. The most-cited research, from Akamai's State of Online Retail Performance and the older Aberdeen Group studies, finds that every 100 ms of added latency on a checkout cuts conversion by 7% to 10%. Google's own data on Core Web Vitals, now a confirmed 2024+ ranking signal, finds that pages above a 2.5-second Largest Contentful Paint lose roughly 24% of impatient users.
Worked example
£500,000/month e-commerce shop, 4G failover left on for a week
- ▸Base latency on FTTP: 10 ms. Conversion rate: 2.4%.
- ▸Router failed over to 4G modem: 110 ms (+100 ms).
- ▸Expected conversion impact: 7% to 10% drop, so 2.16% to 2.23%.
- ▸Revenue impact: £35,000 to £50,000 lost in a single month.
- ▸Annualised: a quiet seven-figure cost from one mis-configured backup connection.
Calculation: revenue × conversion delta. Akamai elasticity of 7% to 10% per +100 ms.
Where latency hurts the most
Not every workload is latency-sensitive. A nightly cloud backup will run fine over a 60 ms connection. But the workloads below are the ones that quietly bleed money on a copper or congested link.
Real-time and voice
- ▸VoIP and SIP trunking: jitter and one-way delay above 150 ms collapses call quality (Teams, Zoom, 3CX, RingCentral)
- ▸Card terminals: Worldpay and Stripe Terminal time out after about 8 seconds, common on ADSL during peak
- ▸Video meetings: Teams, Meet and Zoom downshift to audio-only above 200 ms
- ▸Trading and finance: sub-20 ms is mandatory ; sub-1 ms for proximity trading
Web and cloud apps
- ▸E-commerce checkout: every +100 ms cuts conversion 7% to 10%
- ▸Microsoft 365 and Google Workspace: feel sluggish above 100 ms RTT to nearest region
- ▸Salesforce, HubSpot, Xero: dashboard render times double from 20 ms to 80 ms latency
- ▸Remote desktop and VDI: mouse and keystroke lag becomes painful above 50 ms
How to test the latency on your line
You do not need any paid tools. Three free tests give you a complete picture in under five minutes.
Ping
Run ping cloudflare.com from a command prompt. Average under 20 ms means good ; above 50 ms means investigate. Run it three times across the day.
Measures: baseline RTT and jitter
Traceroute
Run tracert cloudflare.com on Windows or traceroute on macOS. A big jump at one hop tells you exactly where the delay lives.
Measures: per-hop delay
Web speed test
speed.cloudflare.com or speedtest.net report latency, jitter and packet loss together.
Measures: RTT, jitter, throughput
Five ways to reduce latency in 2026
In order of cost and disruption, here is what to do once you know your line is too slow. Pick the lowest item that solves your problem ; you very rarely need a leased line if FTTP is available.
| Fix | Latency improvement | Typical cost | Effort |
|---|---|---|---|
| Router QoS | Saves 10 to 30 ms under load | Free | 30 minutes |
| CDN (Cloudflare, Fastly) | Cuts web RTT by 40 to 80 ms | Free to £200/month | 2 hours |
| Move to UK cloud region | Saves 60 to 90 ms vs. US | Migration project | 1 to 4 weeks |
| Switch FTTC → FTTP | Saves 10 to 15 ms | £25 to £80/month | 2 to 4 weeks install |
| SD-WAN | Routes around congestion | £40 to £150/site/month | 1 to 2 weeks |
| Dedicated leased line | Guaranteed sub-10 ms + SLA | £200 to £700/month | 4 to 12 weeks install |
CDNs, edge compute and IXP peering
For any business with a customer-facing website, app or API, a content delivery network is the cheapest single intervention. A CDN replicates your static assets, and increasingly your dynamic responses, to dozens of points of presence around the UK and Europe. A user in Glasgow hitting your London-hosted shop suddenly fetches the page from a Glasgow server 4 ms away instead of a London server 25 ms away. Cloudflare's free tier covers most brochure sites ; the Pro and Business tiers (£18 and £200 a month) add advanced caching and image optimisation.
Edge compute, popularised by Cloudflare Workers, Fastly Compute@Edge and AWS Lambda@Edge, pushes the application logic itself to the nearest point of presence. For a Shopify or Next.js storefront this can collapse first-byte latency from 200 ms to 30 ms. The major UK Internet Exchange Points (LINX, LONAP, IXManchester, IX Scotland) are the other piece of the puzzle: an ISP that peers at multiple IXPs gives you shorter paths to most UK destinations.
Leased line versus FTTP: which to buy
For most SMEs in 2026 the choice is simple: buy FTTP unless you can articulate a specific reason for a leased line. FTTP is now available to 78% of UK premises (Openreach + altnets like CityFibre, Hyperoptic and Community Fibre). Pricing for a 1 Gbps business FTTP line ranges from £40 to £85 a month with a 24-month term, and gives you 5 to 15 ms latency without any additional engineering.
A leased line costs five to ten times more, but adds three things FTTP cannot: a contractual SLA (typically 99.95% to 99.99% uptime with a 4-to-6-hour fix), symmetric upload and download (your customers' uploads to your servers are as fast as their downloads), and a dedicated bearer with no contention. Buy one when downtime would directly cost you more than £2,000 per hour, or when your business relies on heavy uploads (broadcast, surveillance, large file transfer).
Frequently asked questions
Latency is the round-trip time, measured in milliseconds, that a packet of data takes to travel from your device to a server and back. It is not the same as bandwidth: a 1 Gbps line can still feel slow if every request is delayed by 80 ms. Latency is what makes a Microsoft Teams call lag, a card payment time out, or a Shopify checkout feel sluggish even on a "fast" connection.
For a UK business connecting to UK or Western European servers, anything under 20 ms is excellent, 20 to 40 ms is good, 40 to 80 ms is acceptable for general office use, and above 100 ms is a problem for real-time apps. FTTP fibre routinely delivers 5 to 15 ms; FTTC sits at 15 to 25 ms; 5G standalone is in the same band as FTTP at 5 to 20 ms; 4G typically returns 30 to 50 ms; Starlink Business, widely deployed in rural UK since 2024, lands at 25 to 50 ms.
The most cited research, from Akamai and Aberdeen Group, shows that every 100 ms of extra latency on an e-commerce checkout cuts conversions by 7% to 10%. For a £500,000 a month online shop, moving from FTTP (10 ms) to a poor 4G connection (110 ms) translates into a £35,000 to £50,000 monthly revenue loss. On a 200-seat contact centre, latency-driven re-dials and customer drop-offs typically waste 4 to 8 agent-hours per day.
The quickest free test is to open a command prompt and run ping bbc.co.uk or ping cloudflare.com ; you want the average to sit below 20 ms on fibre. For a richer picture run tracert cloudflare.com (Windows) or traceroute (macOS/Linux) to see hop-by-hop delays. Web tools like Speedtest by Ookla, fast.com and Cloudflare Speed Test all report latency and jitter alongside throughput. Run the test three times across a working day to capture peak-hour variation.
Five no-rewire options work in 2026: switch to FTTP if available (Openreach now reaches 78% of UK premises), put a CDN like Cloudflare or Fastly in front of any web property, move critical workloads to a UK or Irish cloud region rather than a US one, use SD-WAN to route traffic over the lowest-latency path in real time, and enable QoS on the router to prioritise voice and video over background backups.
For a single-site SME, no, an FTTP fibre line gives the same latency at a tenth of the cost. A dedicated leased line (also called Ethernet First Mile or EFM) is worth it when you need guaranteed latency in an SLA, symmetric upload and download, and a 4-hour fix time. Typical 2026 pricing is £350 to £700 per month for 1 Gbps in central London, falling to £200 to £400 in regional cities. Trading floors, broadcast studios and 24/7 contact centres are the usual buyers.