The Livestream That Couldn't Go Down

TL;DR: A European consumer brand's Hangzhou team runs its Tmall, JD and Douyin storefronts. For ten months of the year an IT problem is an inconvenience. On 11.11, a workstation that will not wake up during a two-hour livestream window has a cost finance can calculate to the yuan. Permanent local IT headcount is the wrong shape of fix for peak-shaped risk — pre-positioned capacity is.

Why foreign consumer brands run China e-commerce out of Hangzhou

Hangzhou is where China's platform economy concentrates, and for a foreign consumer brand building a China e-commerce operation, that concentration is the whole reason to be there. The talent pool understands platform mechanics — storefront operations, campaign planning, traffic buying, content production, livestream hosting — in a way that is genuinely difficult to hire for elsewhere. Agencies, production studios, MCN partners and freelance hosts are all locally dense. Somebody who has run three 11.11 campaigns is findable in Hangzhou in a way they are not in most other cities.

This is also why Hangzhou tends to be the second or third China location a foreign brand establishes rather than the first. The corporate entity, the legal function and often the regional leadership sit in Shanghai. The factory or the distribution relationships may be in Guangdong or the Yangtze delta. Hangzhou gets the e-commerce team specifically, because that is where the people who can do this work are.

That sequencing has a consequence that matters more than it appears to. The Hangzhou team is usually the furthest from the company's IT function, both organisationally and in terms of what that function understands about how the team works. Global IT policy was written for offices where people use a laptop, a browser and a collaboration suite. It was not written for a team running capture hardware, editing large video files locally, and operating platform back-ends and livestream tooling that global IT has never seen.

Brocent covers five priority cities in China Mainland — Beijing, Shanghai, Guangzhou, Shenzhen and Hangzhou — and Hangzhou's role in that footprint is specifically platform, software, e-commerce and growth-stage businesses. That is not an accident of coverage; it reflects what companies in Hangzhou actually ask for, which differs materially from what a Shanghai regional headquarters or a Guangzhou manufacturing site asks for.

The scenario: 45 people, and two weeks that carry the year

Consider an illustrative composite — not a named client, but a pattern assembled from the shape of engagements that recur in this segment. A European consumer brand runs its China e-commerce operation from a Hangzhou office of about 45 people, split three ways.

Storefront operations, roughly fifteen people, run the Tmall, JD and Douyin storefronts day to day: listings, pricing, inventory sync with the warehouse, promotional mechanics, platform back-office work. They live in browsers and platform consoles, and their tooling is largely whatever the platforms provide.

Content and creative, roughly eighteen people, produce everything the storefronts and campaigns consume — product photography, short-form video, banners, livestream assets. They work on high-specification workstations with very large local project files. Video projects in particular are worked on locally because that is where the performance is, and they are archived — when they are archived — by somebody remembering to do it.

Customer service, roughly twelve people, handle pre-sale and post-sale enquiries through the platforms' own service tools, at volumes that swing by an order of magnitude between a normal Tuesday and a campaign night.

The campaign calendar dominates everything. 618 in June and 11.11 in November are not busy periods in the ordinary sense; they are the periods that determine whether the year's numbers are met. Around each of them sit pre-sale windows, warm-up phases and livestream schedules that run late into the evening. A livestream session with a booked host and paid traffic behind it has a start time that cannot move.

The team has no local IT headcount. Support comes from a mixture of remote help from the Shanghai office, a break-fix vendor engaged by the office manager two years ago, and one person in the creative team who is good with computers and has become the de facto first line — which is the same single-point-of-knowledge pattern that shows up in every under-supported office, with the added problem that this person's actual job is producing the campaign assets.

The real problems this produces

• Creative work living on local disks with no backup. This is the largest exposure and the least visible one. A workstation holding three weeks of unfinished campaign video is, in practical terms, an unbacked-up production system. The company has a backup policy; it covers file servers and cloud storage, and it does not describe what happens to a 400 GB project folder that only exists on one machine because that is where the editing performance is.
• Livestream hardware that nobody owns between campaigns. Capture cards, cameras, lighting, audio interfaces, the dedicated streaming machine. Between campaigns this equipment sits in a room. Nobody powers it on, nobody patches the machine, nobody checks that the capture card still enumerates after the last driver update. The first time anyone finds out is during setup for the next campaign — or, worse, during it.
• Capacity sized for a normal week. The office network, the workstations, and the internet path were specified for ordinary operations. On a campaign night the same infrastructure carries a livestream upload, a customer-service team at ten times normal volume, creative uploading assets under deadline, and everyone else working normally. Nothing here is a fault; it is a sizing decision that was correct in March and is not correct on 11 November.
• Global security policy that did not anticipate local platform tooling. Platform back-ends, livestream software, and the collaboration tools Chinese partner agencies actually use were not part of the global standard-build assumptions. What follows is predictable: exception requests that take weeks, or — much more commonly — informal workarounds. Personal devices used for platform access, files moved through consumer file-transfer services, and a shadow toolchain that IT does not know exists and therefore cannot protect.
• Cross-border access to head-office systems that is fine until it is not. The Hangzhou team needs the group's ERP, its asset library, and its collaboration platform. Access works, most days. It degrades at unhelpful moments, and when it degrades there is no local person who can distinguish between a platform problem, a network path problem and a policy problem — so the ticket goes to a queue in another timezone and the campaign continues without the asset.
• Nobody is accountable for readiness. The most consequential gap is the simplest. No one's job description includes "confirm the Hangzhou office is ready for 11.11." Individuals do what they can, close to the date, in the time they have left over from their real work.

Brocent's perspective: peak-shaped risk needs pre-positioned capacity, not permanent capacity

The instinctive fix for an under-supported office is to hire an IT person. For this office, it is the wrong shape of answer, and it is worth being precise about why rather than just asserting it.

A single local IT hire is under-used for ten months of the year — which sounds like an efficiency argument, and is not the main objection. The main objection is that a single hire is still unavailable at 11pm on the one night that counts. One person cannot cover a livestream window that runs past 10pm, be present for a 7am pre-sale cutover, and also be the person who prepared the hardware three weeks earlier. Peak-shaped risk is not solved by adding an average-shaped resource; it is solved by putting capacity where the peak is.

The alternative most companies fall back into — ad-hoc break-fix — fails in the opposite way. It is cheap and adequate in March. It is nearly worthless at 9pm on a campaign night, because a break-fix vendor called during a livestream is starting from zero: no knowledge of the environment, no spare parts staged, no idea which machine is the streaming machine, and a response commitment that was written for business hours.

What actually works is preparation plus presence, in that order:

Pre-position the capacity before the window opens. Readiness checks weeks ahead, not days: workstations and the streaming machine powered up, patched and verified; capture hardware tested end to end with the actual software that will be used; the network path checked under something approximating load; creative assets confirmed to be backed up somewhere other than the machine they were made on; spares staged on site for the components most likely to fail.

Put a resident engineer inside the campaign window. Not a permanent hire — a scheduled on-site engineer for the days that matter, who was involved in the readiness check and therefore already knows the environment. Brocent's full-time onsite IT support service is built to be scheduled this way: designated full-time or part-time engineers at defined levels, with background checks, onboarding into the client environment and a documented in-take survey before the first day. The engineer who staffs your 11.11 should not be meeting your setup for the first time on 11.11.

Define out-of-hours escalation for the window itself. A livestream that runs to 11pm needs a support commitment that also runs to 11pm, with a named path beyond the on-site engineer. Brocent's 24×7 multi-lingual help desk runs around the clock from decentralised centres in China, Hong Kong and Malaysia, answering 90% of calls within 40 seconds, and escalates to on-site dispatch when remote resolution is not possible — which is the correct structure for a campaign night, where the first question is always whether this can be fixed from where you are.

Fix the backup problem outside the campaign entirely. Creative assets on local disks are a year-round exposure that merely becomes catastrophic during a campaign. Cloud managed backup with 24×7 job monitoring and regular recovery drills addresses the actual risk, and the drills matter more here than almost anywhere, because large media files are where "the backup completed" and "the file is usable" diverge most often.

What preparation actually looks like

A campaign readiness cycle for an office of this shape is not elaborate. It is simply owned, and it happens on a schedule rather than in the last week.

• Four to six weeks out: inventory and assessment. What hardware exists, what condition it is in, what is out of warranty, what the network path looks like, where the creative files actually live. This is also when the gap between the global standard build and the local toolchain gets documented honestly rather than discovered during an incident.
• Three to four weeks out: remediation and staging. Patch and verify the machines that matter, replace what is failing, order and stage spares locally, and confirm backups of creative work are running and restorable. Ordering a replacement capture card three weeks out is procurement; ordering one on 10 November is a crisis.
• One to two weeks out: full rehearsal. Run the livestream setup end to end with the real software, the real host machine and the real network. Most campaign-night failures are discoverable here, and almost none are discoverable by inspection alone.
• During the window: scheduled on-site presence plus out-of-hours escalation. The engineer is there for the hours that matter, the help desk covers the rest, and the escalation path is written down before it is needed.
• After: a short review that feeds the next cycle. What broke, what nearly broke, what the spares list should look like next time. 618 and 11.11 are six months apart, which is exactly long enough to forget.

Underneath all of this sits ordinary managed IT — monitoring, patching, help desk, backup — running year-round for a 45-person office at a scale appropriate to 45 people. The campaign preparation is an overlay on that, not a substitute for it.

Three ways to cover a Hangzhou e-commerce team, compared honestly

• Hire local IT headcount. Someone is always there, knows the team, and can act immediately. Against that: under-used for most of the year, still a single person, still unavailable at 11pm, still on leave sometimes, and — the part that gets underestimated — still without the spare parts, vendor relationships or escalation depth that a campaign night actually needs. For a 45-person office this is an expensive way to buy presence and a poor way to buy resilience.
• Ad-hoc break-fix. Cheap, flexible, no commitment, entirely adequate for a broken laptop in March. Against that: no knowledge of your environment, no staged spares, no preparation, and a response window written for business hours. On the night that carries the year, you are calling a stranger and explaining your setup from scratch.
• Pre-positioned managed coverage (the Brocent model). A standing managed arrangement year-round for monitoring, patching, help desk and backup, plus a scheduled resident engineer booked into the campaign windows, plus readiness checks and staged spares weeks in advance, plus a defined out-of-hours escalation path for the livestream hours themselves. Against that: it requires planning the calendar in advance rather than reacting, and it is a commercial commitment rather than a per-incident cost — which is precisely the trade you are making when the failure mode you care about is concentrated into a handful of nights.

Where to go from here

The useful diagnostic is a single question asked in a quiet month: if the streaming machine fails to boot ninety minutes before a booked livestream, what happens? Not who would you call — what actually happens, in what order, and how long does each step take. Most Hangzhou teams discover in answering it that they have no staged spare, no tested alternative machine, and no support commitment that covers the hour in question.

Brocent has operated in China Mainland since 2007 and covers Beijing, Shanghai, Guangzhou, Shenzhen and Hangzhou as priority cities, with the Hangzhou footprint aimed specifically at platform, software and e-commerce businesses. Managed IT plans include 24/7 monitoring, help desk, patch management and backup at every tier; on-site engineers can be scheduled by level and duration; and campaign-window coverage is scoped against your actual calendar. You can see the plan structure on managed IT support, or get in touch to work through what your next campaign window needs.

Frequently asked questions

Do we need full-time IT staff in Hangzhou?

For an office of roughly 35 to 60 people with a campaign-shaped calendar, usually not — and the reason is not primarily cost. A single full-time hire gives you presence during business hours but still leaves you with one person, on leave sometimes, unavailable late in the evening, and without staged spares or escalation depth. A standing managed arrangement for the year-round work, combined with a scheduled on-site engineer through the campaign windows, puts the capacity where the risk actually is. If the office grows past roughly a hundred people, or acquires a permanent late-shift operation rather than seasonal peaks, the calculation changes.

What happens during 618 and 11.11?

Under a pre-positioned arrangement, most of the work happens before them. A readiness check four to six weeks out, remediation and spares staging three to four weeks out, a full end-to-end rehearsal of the livestream setup one to two weeks out. During the window itself, a scheduled on-site engineer covers the hours that matter — including evening livestream sessions — with 24×7 help desk coverage and a defined escalation path behind them. The objective is that the campaign night contains no surprises that could have been found in October.

Where should our creative files be backed up?

Somewhere other than the workstation they were created on, with a restore that has actually been tested. That is the whole requirement, and it is more often unmet than met in creative teams, because large project files are precisely the ones that are inconvenient to back up. Brocent's cloud managed backup covers standard file, image and video formats with 24×7 job monitoring and regular recovery drills; the drills are the part that matters most for media, because large-file backups are where a completed job and a usable restore diverge most frequently. The design question — local plus cloud, or cloud only — depends on how much of the workflow needs to survive an office-level failure.

Can our HQ security policy work with local platform tools?

In our experience, usually yes, but only if the gap is documented rather than left implicit. The failure pattern we see is not a policy that forbids something outright; it is a policy written without knowledge of the local toolchain, an exception process that takes longer than the campaign calendar allows, and staff who quietly work around both. Documenting what the team actually needs — platform back-ends, livestream software, the collaboration tools partner agencies use — and putting those in front of the security function as a defined list is what turns this from a shadow-IT problem into a policy decision. What is or is not permissible under your group policy, and any question about cross-border data movement or platform regulation, is a matter for your own legal and compliance functions; we can describe the operational shape and support what they decide.

How quickly can an engineer be on site in Hangzhou?

Hangzhou is one of Brocent's five priority China cities, so dispatch is a standing arrangement rather than an ad-hoc search for a local contractor. The honest answer, though, is that for a campaign night you should not be relying on response time at all. The distinction that matters is between dispatch — which is what you need for an unexpected fault in an ordinary week — and scheduled presence, which is what you need for a livestream window with a booked host and paid traffic behind it. If the answer to your campaign risk is a response-time commitment, the campaign risk has not been addressed.

Is Hangzhou covered differently from Shanghai?

The service model is the same; what typically differs is the mix. Shanghai engagements are more often regional headquarters work — corporate systems, executive support, regional coordination. Hangzhou engagements skew toward platform, software and e-commerce businesses, which in practice means more creative workstation and media-file work, more campaign-shaped scheduling, and more infrastructure sized around peaks rather than averages. Both sit within the same five-city China coverage and the same managed IT plan structure.

How is campaign-window coverage priced?

It has two parts. The year-round managed arrangement — monitoring, help desk, patching, backup — is priced per user per month against the plan tiers, with China priced natively rather than converted from a regional average. Campaign-window on-site engineering is scoped separately by engineer level and by the number of days and hours you actually need, since a two-week 11.11 window and a three-day 618 pre-sale are different commitments. Contract duration also affects the rate — longer commitments attract lower monthly rates, with standard terms of 6, 12 and 24 months. Current plan structure and inclusions are on the managed IT support page; a scoped figure needs a look at your campaign calendar.