Annual Video Content: Projects, Retainer or In-House Team?

Video Production Budget Decisions 10/10 · ONCE Guide to Video Production Budget Decisions

When a team needs video throughout the year, “cost per video” is often the least useful number. One hero film, six cutdowns and twelve formatted files may come from one production cycle. Three exports with different aspect ratios are not automatically three independent creative assignments. The procurement model should be compared on annual capacity required to deliver accepted work, not on the number of files exported.

The conditional answer in this worked scenario is straightforward. At 180 production-equivalent days a year, project-by-project buying costs CNY 516,000, a reserved-day or monthly service costs CNY 435,000, and a two-person internal team with specialist outsourcing costs CNY 429,000. The internal route appears only CNY 6,000 cheaper than the monthly service, while converting CNY 330,000 into a fixed commitment and adding hiring, single-point skill and peak-capacity risk. With demand not yet proven stable and an annual production cash ceiling of CNY 450,000, a quarterly reviewed monthly service is the better fit. In-house can become the better choice only when demand remains above the modelled crossover, the work mix is stable and management capacity already exists.

Every CNY amount and utilization rate below is a hypothetical modelling assumption, not a market rate, industry average, supplier quote or ONCE price list. Replace each line with actual salaries, quotes, taxes and workload.

Define what the year must actually deliver

Assume an established consumer-electronics brand serves Chinese and international ecommerce and social channels. Over the next twelve months it expects four product moments, ongoing product education and performance-creative testing. Work must meet a consistent brand finish, show real product functions, supply core 16:9 and 9:16 versions, include necessary subtitles, preserve usage records and remain searchable in an asset archive. Normal jobs need a five-business-day response, while three launch windows create parallel demand.

The brand owns product facts, samples, claims, the annual calendar, channel list, available assets and final approval. The delivery team owns scheduling, capture, post-production, versions, rights records and archiving. The relevant owner and contract must still confirm regulatory, cast, music and third-party-material scope. Waiting caused by an unstable sample or unapproved claim should not be disguised as low production productivity.

For a reproducible comparison, this article calls one specialist-role day that produces accepted work a “production-equivalent day.” It is not an external day rate, and it does not claim that a director day, camera day and edit day are the same kind of labour. It is only a common capacity unit for the same annual scope.

If real work cannot be converted reliably, do not force it into 180. Rebuild the workload from the last two quarters of tickets, shoot days, post hours and rework causes. File count can remain a delivery checklist, but it is not a workload measure.

What current public evidence can and cannot establish

Wistia's 2026 State of Video Report, published on 22 April 2026, combines a survey of more than 900 professionals with data from over 13 million videos and 79 million viewing hours on its platform. It reports that both in-house work and outsourcing grew, that about a quarter of companies also outsource to freelancers or production agencies, and that 76% of respondents make at least one video a month. This sample shows that blended teams are common; it does not prove an optimal procurement model for a specific company, and its dollar budget distribution is not a Chinese production rate card.

Employee cost cannot be copied from a salary listing either. The US Bureau of Labor Statistics June 2026 ECEC release, published on 9 September 2026, reports that wages and salaries were 70% of private-industry employer compensation and benefits were 30%. This is an aggregate US sample, not an input for Chinese video salaries. It only illustrates why a buyer should include its own benefits, paid time, recruitment and training, equipment, software, space and management time.

AICP's March 2026 bidding guidance emphasizes clear questions, assumptions and scope during bidding. It is US commercial-production guidance, not a regional price table. Adobe's current US teams page lists Creative Cloud Pro for teams at USD 99.99 per license per month, annual and billed monthly. A public software subscription is only one tool line; it is not the total cost of an internal team or a production project. LinkedIn's current video-ad specifications accept landscape, square and vertical formats, reinforcing why a yearly plan needs genuine version management rather than counting arbitrary crops as new ideas.

Price three procurement models to the same scope

First calculate the affordable ceiling. Assume the brand has CNY 520,000 of annual content cash, protects CNY 55,000 for media testing, research and channel activation, and keeps CNY 15,000 as an operating contingency:

Annual production ceiling = 520,000 − 55,000 − 15,000 = CNY 450,000.

All unit values below are calculation assumptions. The model treats its listed amounts as tax-inclusive; real enquiries must confirm tax, invoices, cast and music rights, travel, overtime, cancellation, rollover of unused capacity and out-of-scope rates.

Lean option: buy projects and preserve flexibility

Assume repeated annual bidding, contracting, onboarding and handover costs CNY 30,000. The 180 equivalent days break down as follows:

Project buying is not a low-quality option. It buys the ability to stop when demand falls and change specialists when the work changes. Its weakness is repeated explanation of the brand, sample, version and approval system. If all 180 days will occur, duplicated procurement and fragmented unit costs push the model above the CNY 450,000 ceiling.

Standard option: reserve days or buy a monthly service

Assume a CNY 210,000 annual minimum covers ongoing planning, production management, reserved capacity, asset management and collaboration. Accepted work then draws a blended CNY 1,250 per equivalent day:

CNY 210,000 + 180 × CNY 1,250 = CNY 435,000.

That is CNY 81,000 below the project model and inside the ceiling. It works only if the minimum really covers the required roles, normal response, agreed revisions and asset administration. Rollover, cancellation, peak capacity, cast, locations and complex motion also need written rules. If a “monthly service” simply re-quotes every job, it has not purchased the continuity assumed here.

Enhanced option: build a small team and outsource specialist work

Assume the internal core is one producer-shooter and one editor-content operator. Fixed cost is more than salary:

Then assume external directors, cinematographers, art, talent, sound, complex motion, locations and peak help form an annual variable reserve equal to CNY 550 per equivalent day:

CNY 330,000 + 180 × CNY 550 = CNY 429,000.

The model is CNY 6,000 below the monthly service, but that difference does not justify hiring by itself. Two people cannot cover every specialist function simultaneously. Turnover, sick leave, equipment failure and three overlapping launches still require external help. If a manager spends four hours a week on scheduling and approval, record opportunity value first. It becomes cash savings only when overtime, temporary procurement or a real third-party expense is reduced.

The crossover matters more than three headline totals

Let D be annual production-equivalent days. The three cost functions are:

• Project buying: CNY 30,000 + CNY 2,700 × D
• Monthly service: CNY 210,000 + CNY 1,250 × D
• In-house plus outsourcing: CNY 330,000 + CNY 550 × D

They produce three modelled intersections:

• Projects versus monthly service: about 124.1 days; from the 125th whole equivalent day, monthly service is cheaper in this model.
• Projects versus in-house: about 139.5 days; from day 140, in-house is cheaper.
• Monthly service versus in-house: about 171.4 days; from day 172, in-house is cheaper.

An intersection is not a purchasing instruction. Suppose the two employees have 440 paid workdays in total. Meetings, leave, training, brand approvals, asset management, maintenance and idle time consume 180 days, leaving about 260 direct-delivery days. Utilization is 69% at 180 days and 88% at 230. Near 90%, one overlapping launch can create delay; below roughly 65%, fixed capacity may sit idle. A real team must replace 260 with its own calendar.

“In-house is cheapest” does not make the high-demand case affordable. CNY 456,500 still exceeds the ceiling, while 88% utilization leaves little buffer. Add cash, remove tasks, defer moments or budget peak support. Keep the zero-lift case as well: the videos may generate no measurable incremental orders. Procurement still needs to pass the pre-approved product-proof, content, version, rights and response criteria. Brand content should not be forced into a fictional short-term ROAS.

Add fixed commitment only after the evidence exists

Moving from projects to a monthly service requires at least three conditions: approved twelve-month demand reaches roughly 125 equivalent days; the product and channel calendar is sufficiently stable; and rollover, cancellation, peaks and specialist rules are clear. If two consecutive quarterly forecasts fall below 125 days, reduce the minimum or do not renew instead of buying idle capacity for the sake of continuity.

The mathematical monthly-to-in-house crossover is about 172 days, but the operating gate should also require two quarters of sustained demand, a task mix suited to the two core roles, internal scheduling and asset management, backup for critical skills, and hiring cash that does not displace distribution or specialist work. If direct utilization stays below about 65%, pause headcount growth. If it rises above about 85% while launches overlap, buy peak help or change the calendar instead of hiding a capacity problem in overtime. The 65% and 85% marks are management assumptions in this model, not industry standards.

The counterexamples are clear. A brand with only two large productions a year, each needing different directors, casts and locations, should not build a fixed team merely because it has “ongoing content.” An organization that needs daily product training, founder updates and rapid sales demonstrations may benefit from internal response and accumulated knowledge even if its individual productions are modest.

Nine completed studies form a conditional budget map

The first nine studies address different stages and assignments. They are not one rate card; each model requires the buyer's actual operating inputs, quotes and responsibility split.

A fill-in annual procurement review

Before buying, record: the twelve-month product and channel calendar; equivalent days for strategy, capture, post, specialist work and QC; quality and response standards; current staff and equipment; salaries, benefits, recruitment, software, storage, space and management cost; actual external quotes; rights and version scope; rollover, cancellation and unused-capacity rules; peak months; available cash; and protected media, research and operating funds.

Then calculate:

• Annual production ceiling = available cash − protected media and channel costs − measurement − operating contingency
• Total model cost = annual fixed commitment + unit variable cost × actual equivalent days + peak support + confirmed rework and rights cost
• Internal direct utilization = actual equivalent days ÷ direct-delivery capacity
• Procurement crossover = D where two total-cost functions are equal, rounded up to a purchasable whole task unit
• Stop condition = every feasible route exceeds the ceiling and reuse, self-production, reduced scope, phasing or calendar changes cannot make cost and affordability overlap

To review a real annual plan, share with ONCE the product, target market, use, existing assets, sample status, twelve-month task calendar, necessary response time, acceptable budget, and only the salary or supplier assumptions required for the calculation. Customer lists, private contracts and unnecessary sensitive financial detail are not required. We can first turn file counts back into accepted tasks, then replace the three cost functions with actual quotes.

Source note: Wistia's 2026 State of Video Report was published on 22 April 2026; the US Bureau of Labor Statistics Employer Costs for Employee Compensation — June 2026 release was published on 9 September 2026; AICP published updated bidding guidance in March 2026; Adobe teams pricing, Adobe generative-AI terms and LinkedIn video-ad specifications were read on 9 October 2026. Overseas samples, public software prices and industry guidance do not establish Chinese production rates. Every CNY salary, day, rate, capacity, utilization and total in this article is an independent modelling assumption, not an ONCE price list, market average or performance promise.