Bandwidth Planning by Employee Count

Per-employee assumptions

These are peak figures, not steady-state. Average use is typically 10-30% of peak.

The concurrency factor

If you have 25 employees each capable of peaking at 25 Mbps, naive sizing says 625 Mbps. In practice, not all 25 hit peak at the same moment — even in the busiest hour, perhaps 30-50% are at peak simultaneously. Practical sizing applies a concurrency factor:

required_capacity = employee_count × peak_per_employee × concurrency_factor

Common concurrency factors:

• 0.2-0.3 for typical office knowledge work (most people not on a call at the same moment).
• 0.5-0.6 for sales floors or contact centers where everyone is actively on the network most of the time.
• 0.8-1.0 for meeting-heavy days, all-hands events, or training sessions where everyone is on video at once.

Worked examples

Upload vs download

Many ISPs sell asymmetric plans (download faster than upload) because residential workloads are mostly downstream. Office workloads are more symmetric:

• Video calls send your video upstream and receive others' downstream at roughly the same bitrate.
• Cloud backups are mostly upstream.
• Hosted services (mail server, VPN) receive incoming requests upstream and respond downstream.

An office with 500/50 Mbps may saturate the 50 Mbps upload long before the 500 Mbps download. For multi-employee offices, symmetric or near-symmetric plans are usually correct.

Adding headroom for events

Sizing for typical busy hour leaves you congested during atypical events:

• All-hands video meetings.
• Coordinated software releases or large updates.
• Year-end / quarter-end financial close uploads.
• Onsite events (conferences, trainings) where guest Wi-Fi load spikes.

Add 30-50% headroom above the typical busy-hour calculation. For predictable events (an all-hands every other Friday), QoS prioritization is more useful than just buying more capacity.

QoS and bandwidth shaping

Once usage approaches capacity, QoS prevents the worst symptoms. Priority order is typically:

1. VoIP and real-time voice/video.
2. Interactive traffic (SaaS, screen sharing, SSH).
3. Bulk file transfers and downloads.
4. Background tasks (system updates, cloud backups).

Under contention, QoS slows the bottom of the list to keep the top responsive. It is no substitute for actual capacity, but it preserves the most painful workloads.

How to verify your current sizing is right

1. Pull utilization data from the firewall or router over a week. Most modern devices graph this.
2. Look at the busy-hour peak across the week. Compare to plan capacity.
3. Sustained periods above 80-90% during business hours = undersized.
4. Peaks at 60-70% with occasional excursions = correctly sized.
5. Never above 40% even during all-hands = oversized; you can downgrade.

When to upgrade vs when to fix QoS

Symptoms point you in different directions:

Frequently Asked Questions

How much internet bandwidth does a business need per employee?

Rough starting points: 5-10 Mbps per knowledge worker for mostly web and email, 15-25 Mbps for heavy video conferencing, 50+ Mbps for media production or large file transfers. These are concurrent peaks per employee; not every employee uses peak bandwidth at the same time, which is why oversubscription works.

What is oversubscription?

Selling or providing less capacity than would be needed if every user was at peak simultaneously, relying on the statistical fact that they aren't. ISPs do this with consumer service; offices do it implicitly when sizing their connection. A reasonable office oversubscription factor is 3-5x: a 25-person office can be fine on 100-200 Mbps even if peak per-employee is 25 Mbps.

Does upload bandwidth matter for office work?

Increasingly yes. Video calls are symmetric — uploading your video stream takes about the same bandwidth as receiving theirs. Cloud backups, large file uploads, screen sharing, and any hosted service contributes to upload. Asymmetric plans with 10x more download than upload often run out of upload first.

How do I know if my office needs more bandwidth?

Symptoms: choppy video calls, slow uploads to cloud services, complaints clustered at midday or right before all-hands meetings, speed tests that show much less than your plan during business hours. Confirm by measuring utilization at the router or firewall over a week; sustained periods above 80% of capacity indicate undersized or oversubscribed service.

Should I buy bandwidth for peak or average?

For something between the two. Sizing for absolute peak wastes money; sizing for average creates daily congestion. A reasonable target is the 95th percentile of busy-hour traffic with 30-50% headroom. That handles normal peaks comfortably and only saturates during unusual events.