What Actually Happens During a UK VoIP Contact Centre Deployment (Not the Brochure Version)

From my experience delivering UK contact centre deployments since 2019, the gap between the sales pitch and the operational reality is enormous. Let me walk through what actually happens day-by-day on a 30-seat UK contact centre deployment, so you know what to expect if you are buying one.

The customer profile

Real customer, anonymised. 30-agent B2C contact centre in Reading, inbound sales and support, mixed UK and Ireland, operating 08:00-22:00 Mon-Sat. Migrating off a legacy on-premise PBX (Mitel 3300) to cloud contact centre. 6-week project. Here is how it actually went.

Week 1, discovery and pain-point mapping

The sales pitch says 'discovery takes 2 days.' The reality is we spent 5 working days shadowing 8 different agents across different shifts to understand what they actually do vs what the process document says they do. Major findings:

• The process document said agents transfer to a 'specialist team' for complex cases. Actually, there is no specialist team. Agents DM each other on a Slack channel called #help-me which nobody in management knew about.
• Agents used 4 different CRMs simultaneously — main CRM for accounts, spreadsheet for escalations, PDF knowledge base, and a 15-year-old access database nobody could turn off.
• Average handle time was 7 minutes per the report, 14 minutes in reality (the 7 excluded hold and wrap time because the old PBX did not report those).

This week is tedious but if you skip it, the deployment fails. The number of 'surprise' requirements that only surface when you shadow agents is always more than the customer expects.

Week 2, call flow design

The fun bit on paper. Draw the IVR, the skill-based routing, the overflow rules. Sign off with the customer.

The reality is 3 rounds of iteration. Customer initially wanted 14 IVR menu options at the main line. We talked them down to 4. Then the sales director over-ruled us and demanded 6. Then we ran a mock and the CEO could not navigate her own IVR. Back to 4. This happened over 4 meetings.

The rule nobody tells you, IVR options should be limited to 3-4 at the top level. Every additional option reduces call completion rate by measurable percentages. But customers always want more because they think it 'routes better'. It does not.

Week 3, parallel running setup

We set up the new cloud contact centre in parallel with the old PBX. Agents log into both. Calls route to the new system via a secondary number during test. Old system stays live on the real number.

This is where pain lives. The cloud contact centre requires agents to have stable WiFi AND a decent USB headset. We discovered:

• 8 of 30 agents had home WiFi that dropped every 30-90 minutes. Not suitable for cloud voice. Had to ship them MiFi devices.
• 11 of 30 were using sub-£15 headsets with awful microphones. Customer complaints skyrocketed during test. Had to provision proper USB headsets (Jabra Evolve 40s, about £60 each).
• 3 of 30 had work laptops with outdated audio drivers that caused one-way audio. 2 of those laptops were sub-spec 2017 models. Budget £1500 for emergency laptop refresh.

Total unplanned hardware spend, £4,200. Not in the original project budget. This is normal. Budget 15-20% contingency on contact centre deployments for exactly this.

Week 4, agent training

The brochure says 'intuitive interface, minimal training needed'. In practice, 30 agents × 2 hours of real training = 60 agent-hours, plus 4 hours of supervisor training for the contact centre manager and 2 floor leads. That is a full working week of capacity lost.

We did it in 4 cohorts of 8 agents over 2 days. Structure:

• 30 min intro and log-in walkthrough
• 45 min hands-on call taking, transferring, parking, conferencing
• 30 min dashboard overview
• 15 min Q&A

Do NOT try to train more than 10 people at once. Questions come thick and the slower learners get lost. We made that mistake on project #4 and paid for it in re-training weeks later.

Week 5, soft go-live

First 3 days, 20% of inbound traffic routed to new system, 80% stayed on old. Monitor everything. First 2 days we had 6 significant issues:

1. Skill-based routing sent complex tickets to new agents. Fixed by adjusting priority weights.
2. Wrap-up codes were different between old and new systems, confusing reporting. Had to build a translation table.
3. Call recording was storing in the wrong region (Ireland instead of UK). GDPR concern. Fixed by switching data residency setting.
4. Dashboards showed different numbers than legacy because calculation methods differ. Had to document 'old metric X = new metric Y + new metric Z' for every KPI.
5. One supervisor could not log in. Turned out to be a typo in his email. 20 minutes to fix, 2 hours of him being blocked.
6. Saturday shift had only 4 agents on new system, all got overwhelmed simultaneously. Had to route overflow back to old system mid-shift.

This is week 5. Expect issues. If your deployment partner says 'go-live should be smooth,' they have not done enough of these.

Week 6, full cut-over and close

Final week, 100% traffic moved to new system. Old PBX stays alive for 2 more weeks as emergency fallback, then decommissioned. Agents had built up enough pattern recognition for the new interface that calls felt normal. Average handle time dropped from 14 min to 10 min, primarily because the new screen-pop showed caller history immediately instead of requiring 3 clicks.

Key metrics 30 days post-launch:

• Call abandonment dropped from 12% to 7%
• First-call resolution up from 58% to 73%
• CSAT up from 3.8 to 4.3
• Agent satisfaction (internal NPS) actually dropped week 1-2 then recovered by week 4

Total project cost, £38,000 in setup + hardware, then £24 per agent per month recurring. Payback vs the Mitel system's maintenance contract was about 5 months.

The patterns I have seen across 14 UK contact centre deployments since 2022

• 100% have at least one 'surprise CRM' that nobody in management knew about
• 70% underestimate agent hardware quality
• 60% want too many IVR options at launch
• 90% fail to train supervisors separately from agents
• 100% require more training hours than the vendor quotes

If you are evaluating contact centre platforms, the honest criteria are:

1. Skill-based routing that actually works (test with a realistic ruleset during trial)
2. Screen-pop with CRM integration (test with YOUR CRM, not a demo one)
3. Real-time supervisor dashboards (watch them during a busy hour)
4. Recording retention that matches your compliance needs (ICO for UK, PCI if card data)
5. WFM integration if you have a workforce management tool

The UK providers that do contact centre well in 2026 are a small group. At DialPhone we are one option for smaller UK contact centres (5-50 seats). For 100+ seat deployments, Genesys Cloud, Five9, or NICE CXone are probably better fits. Do not let us pitch you into something that does not suit your scale.