Most businesses sign up for a cloud backup service, install the agent, press “start” and assume they are protected by the end of the day. In reality, cloud backup initial seeding — the very first full copy of your data travelling from your office to the provider’s data centre — can take days or even weeks on a typical UK business broadband line. Until that first copy finishes, your shiny new backup protects only part of your data, and the files that are not yet uploaded are exactly as exposed as they were before you signed the contract.
This guide shows you, step by step, how to work out how long your first backup upload will really take, how much upload bandwidth cloud backup needs once it settles into daily incrementals, when to ship your data on an encrypted offline seeding drive instead, and how to throttle backup traffic so Teams calls and card terminals keep working while gigabytes leave the building. Every figure is worked through so you can plug in your own numbers, and every recommendation is anchored to UK broadband reality — FTTC, FTTP, leased lines and 4G/5G failover.
What is cloud backup initial seeding?
Cloud backup works in two phases. The first is the seed: a complete copy of every protected server, NAS share, virtual machine or endpoint is compressed, deduplicated, encrypted and sent to the provider. The second is the incremental phase: from then on the agent only sends blocks that have changed since the last job, which is typically a small fraction of the total. Incrementals are cheap and quick; the seed is the expensive, slow part, and it is the moment most backup projects quietly stall.
There are two ways to seed. Online seeding pushes the full data set over your internet connection. Offline seeding (also called physical data shipping, disk seeding or sneakernet) writes the first full backup to an encrypted drive on site, couriers it to the provider’s UK data centre, and the provider imports it directly into your storage. After the import, your agent “maps” to the seeded copy and carries on with incrementals over the wire. Most mainstream platforms support one or both: Veeam Cloud Connect allows backup copy jobs to be mapped to a seeded repository, Acronis offers Physical Data Shipping, and Microsoft’s Azure Backup MARS agent supports offline backup using Azure Data Box.
The choice between the two is not a matter of taste. It is arithmetic: data volume, upload speed, how much of that upload you can spare during the working day, and how long you are prepared to run with a partial backup. The rest of this guide is that arithmetic, laid out so you can make the call with confidence. If you are still deciding on the overall architecture, read our guide to the 3-2-1 backup rule for ransomware-proof cloud backup first — seeding is the step that turns that design into a real off-site copy.
Ask your provider for their seeding options before you sign. Some include one free seed drive per contract, some charge £150–£400 per shipment, and a few do not offer offline seeding at all — which can rule them out if you hold several terabytes behind a 20 Mbps upload.
Seeding readiness scoring — where most UK businesses sit today
Before you calculate anything, it is worth being honest about the starting position. When we review backup projects for small and mid-sized UK organisations, the same weak spots appear again and again. None of them are exotic — they are simply the things nobody checks until the first backup has been “running” for a fortnight and the progress bar still reads 40%.
Score your own environment against the three cards below. Anything marked high risk is a likely cause of a stalled seed or an unworkable incremental window.
The pattern is consistent: organisations usually know roughly how much broadband they pay for, but not how much upload is genuinely available after VoIP, Microsoft 365 sync and video calls have taken their share, and very few know the real size of the data set once forgotten archives, PST files and duplicated project folders are counted. Those two unknowns account for most seeding estimates that turn out to be wrong by a factor of two or three.
Cloud backup initial seeding by the numbers
A few headline figures frame the problem. They are based on typical UK business broadband products in 2026 and on the data volumes we see in organisations of 10–150 staff.
The first figure is the one that surprises people. One terabyte sounds modest — it is a single external drive from a high-street shop — yet it is eight trillion bits. Divide by 20 million bits per second and you get 400,000 seconds, or about 4.6 days of uninterrupted, flat-out uploading. Nobody gets flat-out uploading for 4.6 days on a line that the whole office also uses, which is why the realistic figure is closer to a week or more.
Online seeding vs offline seeding drive — how they compare
Both approaches end in the same place: a complete, encrypted copy of your data in a UK data centre, with daily incrementals running over the internet. The difference is how long you live with a partial backup, how much disruption you cause on the way, and what you pay.
Online seeding
Full backup sent over your broadband
Offline seeding drive
Encrypted disk couriered to the data centre
The important thing to notice is that offline seeding time barely changes as data grows: writing 10 TB to a local USB 3 or eSATA drive takes hours, not days, and the courier takes the same time whether the drive is full or half empty. Online seeding time scales linearly with data. That is why there is a crossover point, which we calculate in the next section, beyond which the drive always wins.
There is also a middle path many businesses overlook: hybrid seeding. Upload the small, critical, fast-changing data online straight away — finance, line-of-business databases, active project folders — and ship the large, slow-changing archive on a drive. You get meaningful protection within a day or two and avoid a fortnight of saturated upload.
How long will my first cloud backup take? The upload time formula
You can estimate first backup upload time with one line of arithmetic. Take the data size in gigabytes, multiply by 8,000 to convert to megabits, apply the expected reduction from compression and deduplication, then divide by the upload speed you can genuinely sustain:
Upload hours = (Data in GB × 8,000 × reduction factor) ÷ (upload Mbps × efficiency) ÷ 3,600
Use a reduction factor of around 0.6–0.7 for ordinary office data (documents, spreadsheets, databases, email exports compress and deduplicate well), and 0.95–1.0 for already-compressed content such as video, JPEG photo libraries, CAD exports and ZIP archives. Use an efficiency of 0.6–0.8: protocol overhead, encryption, TLS handshakes, small-file processing and contention all take their share, and you will almost certainly throttle backup traffic during the working day.
The chart below shows the time to upload 1 TB of raw data (no reduction applied, 75% efficiency) across the connection types most UK businesses actually have. Bar length is proportional to the hours involved.
Two things stand out. First, the gap between asymmetric and symmetric connections is enormous: an FTTP 1000/115 line downloads nine times faster than it uploads, and it is the upload figure that governs backup. Second, the 1 Gbps figures assume the provider’s ingest and your backup appliance can keep up. In practice a single backup stream often tops out at a few hundred megabits; multiple parallel streams, a modern appliance and a provider with proper ingest capacity are needed to get close to line rate.
Now work an example. A 35-person accountancy practice has 1.8 TB on its file server and a 600 GB SQL-based practice management database. Assume a 0.65 reduction factor overall, an FTTC 80/20 line, and 70% efficiency at night but only 25% of upload allowed during office hours. That works out at roughly 1,560 GB after reduction, or about 12.5 million megabits. That gives roughly 5 Mbps for eleven office hours and 14 Mbps overnight and at weekends — a blended average of about 10 Mbps — so the seed takes in the region of 14–15 days. That is two weeks of partial protection — long enough to justify shipping a drive.
Measure, do not assume. Run three upload speed tests from a wired PC at different times of day — ideally to a UK test server — and use the lowest figure. Many FTTC lines sync well below 20 Mbps upload, and line length from the cabinet matters more than the product name on the bill.
Why most of the seed is data you rarely touch
A useful way to think about seeding is to separate hot data — files and databases that change daily — from cold data that has not been modified in the past year. On a typical small-business file server, cold data dominates: closed client folders, old financial years, scanned paper archives, photo libraries, superseded drawings and former employees’ home drives. Cold data is the bulk of the seed and the least urgent part of it.
This matters for three reasons. First, cold data is the ideal candidate for an offline seeding drive: it will not have changed by the time the drive reaches the data centre, so the first incremental after import is small. Second, cold data is often the easiest to reduce before seeding — duplicated copies of the same project, ISO images, installers and personal media can be archived or deleted under your retention policy rather than backed up for years. Third, hot data is where your recovery point objective really bites, so it should be seeded first. Our guide to backup retention policies covers how long each class of data should be kept, which in turn tells you what genuinely needs protecting.
A quick way to measure the split on a Windows file server is to run a storage report by last-modified date — File Server Resource Manager, TreeSize or a short PowerShell script will all do it. Tag anything untouched for more than 12 months as cold, everything else as hot, and you have the basis for a seeding plan.
How to seed a cloud backup — step by step
The timeline below is the sequence we follow for a typical seeding project at a 20–100 person organisation with one or two on-premises servers and a NAS. It assumes a hybrid approach: hot data online, cold data on a drive. Adapt the durations to your own arithmetic.
The ordering matters. Throttling is configured before any upload starts, so the first users to notice the backup are not your customers on a dropped call. The local backup carries on until the cloud copy has been tested, so there is never a day without at least one restorable copy. And the test restore is a formal step, not an afterthought — our backup restore testing guide explains how to build it into a routine.
Should you ship a seed drive? A decision gauge
To turn the arithmetic into a decision, score your environment from 0 to 100 on five questions, 20 points each. Award the full 20 where the statement is true, 10 where it is partly true and 0 where it is false:
- The protected data set is larger than 1 TB after clean-up.
- Sustained upload available for backup is under 50 Mbps.
- Your estimated online seed time is longer than seven days.
- The office depends on real-time services (VoIP, video, card payments, remote desktop) during the working day.
- Your provider offers encrypted offline seeding at a reasonable price.
A score of 60 or more points firmly towards offline or hybrid seeding. Between 30 and 60, hybrid seeding is usually sensible. Under 30, an online seed with sensible throttling will normally finish within a weekend or two. The worked accountancy example from earlier scores 70 — full marks on data size, upload speed, seed time and reliance on VoIP, but nothing on price because its existing provider charged for every drive:
The score is not a substitute for judgement. A business with only 800 GB but a 10 Mbps upload on a rural FTTC line may still be better off shipping a drive, and a business with 5 TB on a symmetrical gigabit leased line will usually seed online over a single weekend. Use the gauge to start the conversation, then check the result against the formula.
Seeding costs — what each option really costs
Offline seeding is rarely expensive, but it is not free, and the costs that catch people out are the indirect ones: staff time, a broadband upgrade bought in a hurry, or a fortnight of lost productivity while the office internet crawls. The table below sets out typical UK costs in 2026 for a business protecting 2–5 TB. Figures exclude VAT and vary by provider.
| Seeding option | Typical direct cost | Hidden or indirect cost | Typical time to protected |
|---|---|---|---|
| Online seed, unthrottled | £0 | Degraded calls and cloud apps for days; complaints; jobs paused and restarted | 1–3 weeks on FTTC |
| Online seed, throttled to nights and weekends | £0 | Longer exposure window; more monitoring effort | 2–4 weeks on FTTC |
| Provider seed drive (one shipment) | £150–£400 | Two to four hours of engineer time to write, verify and package the drive | 4–8 working days |
| Hybrid: hot data online, cold data on a drive | £150–£400 | Planning time to classify data | 1–2 days for critical data; 1 week for all |
| Temporary broadband upgrade or second line | £300–£1,500 install plus rental | Lead time of weeks for FTTP or a leased line; minimum contract terms | Depends on install date |
In most cases the cheapest route to full protection is the hybrid option. A permanent connectivity upgrade can still make sense — if your upload is struggling with a backup, it is probably also struggling with Teams, SharePoint sync and VoIP — but it should be justified on its own merits rather than bought in a rush just to seed. For a broader view of what cloud backup costs once you are past the seed, see our UK cloud backup pricing guide.
Upload bandwidth for cloud backup once seeding is done
After the seed, the question changes from “how long will the first upload take?” to “does each night’s incremental fit inside the night?”. Use the same formula with the daily change instead of the full data set. Taking the accountancy example: 2.4 TB with a 2% daily change rate is about 48 GB of changed data; after a 0.65 reduction factor that is roughly 31 GB, or 250,000 megabits. At 14 Mbps of sustained overnight upload, the incremental takes about five hours — comfortably inside a 19:00–07:00 window. At a 5% change rate it would take over twelve hours and start spilling into the working day, which is the point at which a faster upload or a smarter backup design becomes necessary.
Change rate is driven by a few specific culprits. Large Outlook PST files change on every open, so a 20 GB PST can upload in full each night if the agent does not track changes at block level. SQL Server and other databases without proper transaction log handling generate large daily deltas. Defragmentation jobs and some antivirus scans touch blocks without changing data. Video editing and CAD projects produce large new files every day. Identify these before go-live and either handle them with block-level change tracking, schedule them separately or exclude them where another copy already exists — for instance, mailboxes already in Microsoft 365 should be protected with a Microsoft 365 backup rather than via PST files on a server.
Backup bandwidth throttling — benchmarks that keep the office working
Throttling is what makes an online seed tolerable. Most business backup agents — Veeam, Acronis, Datto, N-able Cove, Azure Backup’s MARS agent and others — let you set a different bandwidth cap for working hours and out-of-hours, and many can follow a weekly schedule. On top of that, a firewall or SD-WAN appliance such as Cisco Meraki MX can classify backup traffic and give voice and video strict priority. Use both: the agent cap stops the backup flooding the line, and the traffic-shaping rule protects calls when anything else, such as a OneDrive resync, competes for upload.
The benchmarks below are the starting values we use. They are expressed as a share of measured upload capacity and should be tuned after the first two or three days of real traffic.
Recommended upload allocation for backup traffic
Two of these rows deserve a comment. Backup traffic should never run over a mobile failover link: 4G and 5G business data bundles are usually capped, and a seed can exhaust a month’s allowance in hours while also starving the failover of the capacity it exists to provide. Configure the firewall to block or heavily restrict backup destinations whenever the primary line is down. And remember that restores run in the opposite direction — download is usually faster on asymmetric lines, but a large restore can still saturate the connection, so plan the allocation for that too. Our guide to static IP addresses on business broadband is useful here, as many providers allow you to lock backup ingest to known source addresses.
Common cloud backup seeding mistakes to avoid
Almost every stalled or failed seed we see comes down to one of a short list of mistakes. Each is easy to avoid with an hour of planning.
- Trusting the headline speed. “80 Mbps broadband” on an FTTC line usually means a 20 Mbps upload at best, and often less. Seed estimates based on the download figure are wrong by a factor of four or more.
- Seeding everything at once in no particular order. If the agent works through shares alphabetically, the finance database might be the last thing protected. Seed in order of business impact.
- Switching off the old backup on day one. Until the cloud copy is complete and tested, the old backup is the only full restore point you have. Keep it running in parallel.
- Leaving throttling at the default. Many agents ship with no cap at all. The first sign of a problem is a queue of complaints about choppy Teams calls and dropped VoIP.
- Posting the drive and the passphrase together. An encrypted seed drive is only safe if the key travels separately. Never write it on the drive, the packaging or the courier note.
- Restarting jobs after interruptions. Some configurations restart a failed full backup from zero. Check that your agent resumes where it left off before trusting a multi-day upload to it.
- Backing up junk. Installers, ISO images, duplicated project copies and personal media can add a third to the seed. Clean up under your retention policy before you start.
- Forgetting the restore direction. Seeding solves the upload problem, but a full restore of several terabytes over the internet may take days. Ask about reverse shipping, where the provider sends your data back on a drive.
A seed drive contains a complete copy of your business data. Treat it as a high-risk asset under UK GDPR Article 32: hardware or AES-256 software encryption, a tracked and signed-for courier, a recorded serial number and written confirmation from the provider when it has been imported and securely wiped. If a drive goes missing and was not encrypted, you may have a reportable personal data breach to the ICO within 72 hours.
Cloud backup seeding checklist — the 12-point essentials
Use this list before you press “start” on a new cloud backup, or when migrating from one provider to another. It condenses the steps above into a single page you can work through with your IT team or managed service provider.
- Inventory every server, VM, NAS share, cloud service and endpoint that will be protected, with its used capacity.
- Measure real upload speed three times at different times of day from a wired connection, and record the lowest result.
- Check the broadband contract for fair-use limits, traffic management or data caps.
- Clean up duplicates, installers and personal media in line with your retention policy.
- Classify data as hot or cold by last-modified date and agree a seeding order with each data owner.
- Calculate the online seed time with the formula in this guide, using a realistic reduction factor and efficiency.
- Decide on online, offline or hybrid seeding, using the decision gauge as a starting point.
- Configure agent throttling schedules and a firewall or SD-WAN rule that prioritises voice and video.
- Block backup traffic over any 4G/5G failover link.
- If shipping a drive, encrypt it, store the key separately, record the serial number and use a tracked courier.
- Keep the existing local or legacy backup running until the cloud copy is complete and tested.
- Perform and document a test restore from the cloud copy, then check that nightly incrementals finish inside their window.
The NCSC’s Small Business Guide recommends keeping backups separate from your main systems and testing that you can restore from them. Seeding is the step that creates that separate copy — but it only counts once the restore test has passed. Record the date and result of the test alongside your other security evidence.
Real-world example — seeding a 4 TB architecture practice on FTTC
A Bristol-based architecture practice with 28 staff came to us after its previous attempt at cloud backup had been running for three weeks and was still only 55% complete. The practice had around 4.2 TB on a NAS, much of it CAD models and rendered images that barely compress, and a single FTTC line with a measured upload of 16 Mbps. The backup agent was uncapped, so every afternoon the office’s VoIP calls broke up and staff paused the job — which, on that configuration, sent it back several hundred gigabytes each time.
Running the numbers made the problem plain: with a reduction factor close to 1.0 and an efficiency of around 70%, a full online seed of 4.2 TB would need over 34 days of continuous upload. Instead, the practice cleaned out 900 GB of duplicated project exports and old render caches, leaving 3.3 TB. The 260 GB of live projects and the accounts folder were seeded online over a weekend with a 90% cap. The remaining 3 TB of closed projects was written to an encrypted seed drive in an afternoon, couriered the next morning and imported by the provider within three working days. Throttling was set to 20% in office hours with a Meraki traffic-shaping rule prioritising voice. The first incremental after import was 41 GB and finished overnight.
“We had assumed the backup was nearly done because it had been running for weeks. Seeing the arithmetic on one page made it obvious it never would be on our line. A week later everything was in the cloud, tested, and nobody had noticed it happening.” — Practice manager, Bristol architecture firm
The practice later upgraded to FTTP for reasons unrelated to backup, but the design did not depend on it. Nightly incrementals had been fitting comfortably inside the window on the original FTTC line from the first week.
At a glance — cloud backup initial seeding key facts
| Topic | Key fact |
|---|---|
| What seeding is | The first full backup copy sent to the cloud; everything after it is incremental |
| Upload time formula | (GB × 8,000 × reduction) ÷ (Mbps × efficiency) ÷ 3,600 = hours |
| 1 TB on FTTC 80/20 | About 4.6 days theoretical; 6–7 days realistic |
| 1 TB on FTTP 1000/115 | About 26 hours at 75% efficiency |
| 1 TB on a 1 Gbps leased line | About 3 hours if the appliance and provider can keep up |
| Reduction factor | 0.6–0.7 for office data; 0.95–1.0 for video, images and CAD |
| Typical daily change rate | 1–5% of protected data |
| Offline seed drive cost | Typically £150–£400 per shipment, sometimes included |
| Offline seed end-to-end time | About 4–10 working days, largely independent of data size |
| When to ship a drive | Over ~1 TB with under ~50 Mbps upload, or an online seed longer than a week |
| Office-hours throttle | 20–35% of measured upload; 90% overnight and at weekends |
| 4G/5G failover | Never route backup traffic over it |
| Seed drive security | Encrypted, key sent separately, tracked courier, wipe confirmation |
| Old backup | Keep running until the cloud copy is complete and tested |
| Sign-off | A documented test restore from the cloud copy |
How Cloudswitched handles cloud backup seeding
Cloudswitched designs, seeds and manages cloud backup for UK organisations, from single-office firms on FTTC to multi-site businesses with leased lines. We measure your real upload capacity, size and classify your data, choose between online, offline and hybrid seeding, configure throttling and traffic shaping so the office keeps working, and sign the project off with a documented test restore. Once seeding is complete, we monitor nightly jobs and storage growth so the backup keeps pace with the business.
Get your first cloud backup finished properly
Talk to our team about a seeding plan sized to your data and your broadband, with a tested restore at the end.
Talk to us about Cloud BackupFrequently Asked Questions
What is cloud backup initial seeding?
Cloud backup initial seeding is the process of creating the very first full copy of your data in the cloud backup provider’s storage. Every server, share, virtual machine or device in scope is compressed, deduplicated, encrypted and sent to the provider — either over your internet connection (online seeding) or on an encrypted drive that is couriered to the data centre (offline seeding). After the seed is in place, the backup agent only sends changed blocks each night, which is far quicker. Until the seed is complete, only part of your data is protected off site, so it is worth planning the seed rather than leaving it to run in the background.
How long does a first cloud backup take to upload?
It depends on three things: how much data you have, how fast your upload really is, and how much of that upload you allow the backup to use. As a rule of thumb, 1 TB takes about 4.6 days at a flat-out 20 Mbps, which is the top upload of most FTTC lines, and 6–7 days in practice once overheads and throttling are included. On an FTTP 1000/115 line it takes about a day, and on a symmetrical gigabit connection a few hours. Use the formula in this guide with your own measured upload speed to get a realistic figure.
How much upload bandwidth does cloud backup need?
For the seed, as much as you can spare — the faster the upload, the shorter your exposure window. For day-to-day incrementals, the test is whether each night’s changed data uploads comfortably inside the overnight window. A business with 2 TB and a 2% daily change rate needs to send roughly 25–30 GB a night after compression, which takes about five hours at 14 Mbps of sustained upload. If your incrementals regularly run into the working day, look at your change rate culprits first, then at a faster upload.
What is an offline seeding drive?
An offline seeding drive is an encrypted hard disk or SSD, usually supplied by the backup provider, onto which the first full backup is written locally at your office. The drive is then sent by tracked courier to the provider’s data centre, where the data is imported into your cloud storage. Your backup job is then pointed at the imported copy and continues with incrementals over the internet. Veeam Cloud Connect, Acronis Physical Data Shipping and Azure Backup’s offline backup with Azure Data Box are well-known examples of this approach.
When is it worth shipping a seed drive instead of uploading?
Shipping usually makes sense when you have more than about 1 TB to protect and less than about 50 Mbps of upload, or when your calculated online seed time is longer than a week. It also makes sense when the office relies heavily on VoIP, video calls or card payments during the day and cannot afford a saturated upload. A hybrid approach — critical, fast-changing data online and the bulk archive on a drive — often gives the best balance of speed, cost and disruption.
Will cloud backup slow down our internet?
It can, particularly during the seed, if it is left uncapped. The fix is to set a bandwidth schedule in the backup agent — typically 20–35% of measured upload during office hours and up to 90% overnight and at weekends — and to add a traffic-shaping rule on your firewall or SD-WAN appliance that gives voice and video priority over backup traffic. With both in place, most users never notice the backup running.
Is it safe to post our data on a seed drive?
Yes, provided it is handled properly. The drive should be encrypted with AES-256, either by the backup software or in hardware, with the passphrase stored in a password manager and never sent with the drive. Use a tracked, signed-for courier, record the drive’s serial number, and get written confirmation from the provider when the data has been imported and the drive securely wiped. These steps support your obligations under UK GDPR Article 32 to protect personal data with appropriate technical and organisational measures.
How does compression and deduplication affect seeding time?
Compression and deduplication can reduce the volume sent during seeding by roughly a third for typical office data such as documents, spreadsheets, databases and email exports. They make little difference to content that is already compressed, such as video, JPEG photo libraries, CAD exports and ZIP archives. When estimating your seed time, use a reduction factor of around 0.6–0.7 for office data and close to 1.0 for media-heavy data. Design and engineering businesses should usually assume almost no reduction.
Can we use the backup while the first upload is still running?
You can usually restore anything that has already been uploaded, but anything not yet seeded is not protected off site. That is why the seeding order matters: protect the line-of-business database, finance data and active projects first. It is also why you should keep your existing local or legacy backup running in parallel until the cloud seed is complete and a test restore has passed.
Should backup traffic use our 4G or 5G failover line?
No. Business 4G and 5G failover services usually come with limited data allowances, and a backup job can consume a month’s allowance in a few hours while also using the capacity that the failover exists to provide for calls and cloud applications. Configure your firewall so backup destinations are blocked or tightly restricted whenever the primary line is down, and let the backup resume when the main connection returns.
What happens if the seed upload is interrupted?
Most modern business backup agents resume an interrupted full backup from the last completed block or file, but some configurations restart from the beginning, which can turn a two-week seed into a never-ending one. Check how your agent behaves before relying on a long online seed, avoid pausing and restarting jobs manually, and monitor progress daily against your estimate so you spot a stalled job quickly.
Do we need to seed again if we change backup provider?
Usually, yes. Backup formats and encryption are generally provider-specific, so moving to a new platform means creating a new seed in the new provider’s storage. Plan the overlap carefully: keep the old service running until the new seed is complete and tested, and check the old contract’s retention terms so historical restore points remain available for as long as your policy requires.
Related reading
These recent Cloudswitched guides cover the connectivity, server and governance decisions that sit around a cloud backup project.
Ready to plan your cloud backup seed?
Cloudswitched sizes, seeds, tests and monitors cloud backup for UK businesses, so your first full copy is finished, verified and kept up to date.
Explore Cloud Backup








