Features
What to test in a small business VoIP trial before signing: call quality, sample sizes, exit terms
What to test in a small business VoIP trial: sample sizes, call quality thresholds, failure rules, exit terms, and porting and E911 checks before signing.
What to take away
- Set pass and fail numbers before the first test call, not after the first invoice.
- Run at least 200 outbound and 100 inbound calls per site, including home offices and mobile endpoints.
- Log packet loss, jitter, one way latency and mean opinion score for each call, then count the failures.
- Demand a 30 day exit clause, a written porting window and an E911 test in the contract.
- Score the vendor on evidence, and keep the written requirement that drives a real trial.
Decide pass and fail before the trial starts
A trial without thresholds is a demo. Agree on the numbers in writing with the provider before day one, so 30 days produce evidence instead of opinions.
Keep the small business VoIP checklist short enough that one person can run it. Your plan needs four parts: a call quality floor, sample sizes per site, the features you truly use, and the conditions that let you walk away. A vendor that resists writing these down has answered your question already.
Name the features before the trial starts: auto attendant and IVR menus, ring or hunt groups, call forwarding, voicemail to email, call park and pickup, conferencing, the desktop softphone and mobile app, call recording, and any CRM or help desk integration. Have the people who will use each one test it, and write down what happens when it breaks.
Sample sizes that make a VoIP pilot test plan meaningful
Small samples hide problems. Ten calls will not expose a jitter spike that appears only at 2 p.m. on a weekday. Name every site and user group in the VoIP trial checklist, then set a floor for each.
VoIP trial sample sizes
- 200outbound calls
- 50outbound to mobile
- 100inbound calls
- 10peak-hour calls on three days
Four sites means 800 outbound calls across the trial, 200 at each site. Divide them across quiet Mondays and busy Fridays so the network shows its worst behavior.
VoIP call quality testing: the numbers to log
VoIP sends calls as data packets, so most audio problems start as network problems. Ask the provider for its call detail records (CDRs) and quality reports, and run your own probes inside the local network with tools such as iperf3, Wireshark or PingPlotter, which log loss, jitter and latency by the hour.
VoIP call quality thresholds
Pass
- Packet loss
- under 0.5%
- Jitter
- under 20 ms
- One way latency
- under 100 ms
- Mean opinion score
- 4.2 or higher
Warning
- Packet loss
- 0.5-1%
- Jitter
- 20-30 ms
- One way latency
- 100-150 ms
- Mean opinion score
- 4.0-4.2
Fail
- Packet loss
- above 1%
- Jitter
- above 30 ms
- One way latency
- above 150 ms
- Mean opinion score
- below 4.0
VoIP call quality testing
| Metric | Pass | Warning | Fail |
|---|---|---|---|
| Packet loss | under 0.5% | 0.5% to 1% | above 1% |
| Jitter | under 20 ms | 20 to 30 ms | above 30 ms |
| One way latency | under 100 ms | 100 to 150 ms | above 150 ms |
| Mean opinion score | 4.2 or higher | 4.0 to 4.2 | below 4.0 |
Review the log weekly. One bad call is noise. A cluster of bad calls at the same hour points to a circuit, a codec or a firewall rule you can fix before signature.
A 30 day trial schedule
30 day VoIP trial schedule
- Days 1-3Confirm endpoints, circuits and thresholds
- Days 4-7Install phones and verify E911 data
- Days 8-14Run first 100 outbound calls
- Days 15-21Test inbound routing and voicemail
- Days 22-26Stress test conference and failover
- Days 27-30Score results and submit port request
Days 1 to 5: install and baseline. Confirm dial tone at every site, walk the feature list with the people who will use it, and pull one baseline quality report before the live traffic starts.
Days 6 to 15: outbound and inbound calls. Begin the 200 outbound test calls at each site, spread across mornings and afternoons, and confirm that inbound calls land on the right desk.
Days 16 to 25: busy hours and failover. Run calls at the busy hour, drop the primary circuit to test failover, and check the emergency address behind every handset. Log each call that breaches the fail column.
Days 26 to 30: score and decide. Compare the log against the thresholds, request any missing call detail records, and only if the score passes submit the port request; otherwise renegotiate or walk away.
Failure thresholds and the exit clause
Set a count based rule. If more than 2 percent of test calls breach the fail column, the trial fails and you owe nothing. Write that sentence into the agreement.
VoIP contract evaluation is where much of the value sits. Ask for a 30 day exit window with no early termination fee, service credits for downtime, and a maximum porting window. Simple ports must complete within one business day under the number portability rules, so a vendor quoting three weeks is describing its own process.
Example: a twelve seat insurance office
A 12 seat agency ran this plan over four weeks. It logged 240 outbound calls, 60 of them to mobiles. Nine calls breached the jitter threshold, all on the backup LTE circuit between 1 p.m. and 3 p.m.
The agency kept the primary fiber provider, wrote a failover requirement into the agreement, and refused a term contract until the exit clause matched the trial.
Number portability and emergency service checks
Porting is where trials collapse after signature. Before you sign, collect the losing carrier's account number, the service address and the authorized contact, and confirm who submits the request.
Test E911 as well. Call the non-emergency line at your public safety answering point (PSAP) and check the address that appears, since the FCC's 911 rules at 47 CFR Part 9 require an interconnected VoIP provider to route emergency calls with location data.
Choosing the system and the contract together
If you are still comparing providers, treat choosing VoIP systems as a written trial with scored requirements before you shortlist, because a shortlist without a test plan proves nothing.
