You choose capability mix and risk ownership—not merely cheap vs expensive
“All in-house” sounds controllable; “all agency” sounds easy—both can fail at month twelve. Agency vs in-house is really: which capabilities must sit on payroll, which peaks are cheaper outside, how decisions and IP land, who can take over after failure. At KSX Studio in Shanghai, brand, manufacturing, and startup clients rarely pick a pure pole. More often: co-create discovery and v1 with a studio, internalize core domain logic over time, and run SEO/brand-site peaks as project or retainer. Below: decision dimensions, cost truth, and transition paths so one RFP does not freeze your org design.
Map capabilities: core, leverage, commodity
Split work three ways: core (differentiation and data moat—pricing engines, supply algorithms, model strategy); leverage (high impact but standardizable—design systems, marketing sites, analytics instrumentation); commodity (mature patterns—basic site refresh, email templates). Core leans in-house or deeply embedded long-term partners; leverage fits senior agencies; commodity can be project-bid. Failure modes: outsourcing core as commodity with no knowledge transfer, or staffing commodity work full-time until headcount bloats. Redraw the map yearly.
Pace and peaks: staff to peak or to baseline
Campaign sites, brand upgrades, and pre-fundraise sprints create eng/design peaks. Hire to peak and you idle after; hire only to baseline and peaks break quality. Agencies and flexible contractors fit peaks; in-house fits continuously evolving product kernels. Write a 12-month peak calendar (redesign, locales, app release, compliance) before setting outsourcing ratio. Startups undercount post-launch ops and content baseline—that layer often belongs inside or on a small retainer.
Hiring reality: time-to-seat and senior density
In Shanghai and remote markets, seniors who can own a Next.js site, design system, and baseline SEO often take months to seat; design leads and growth engineers are scarce too. One agency value is an already-teamed squad now. Agencies do not replace your product decision-maker and domain experts—without an internal PO, outsourced work ships and then idles. If you must launch in 90 days with empty seats, agency or hybrid wins; if you will iterate one product for 18 months, investing in hire-and-grow usually wins.
Control: code, accounts, decision rights
Whether agency or in-house, source repos, cloud, domains, stores, analytics, and ad accounts should be under your auditable ownership; the agency gets granted access. Decision rights: business priority and go/no-go stay yours; implementation can be delegated. Contracts cover IP, subcontracting, offboarding, data deletion. Equity-for-build partnerships need the same account and repo clarity. Fuzzy control makes year-two switching costs explode.
Hybrid models: the usual sober choice
Model A: agency ships v1 + training; you hire one full-stack/front-end to maintain. Model B: in-house design+product; eng peaks outsourced. Model C: in-house core product; brand site/SEO/campaign pages on retainer. Model D: embedded joint squad on your cadence. For each, write success metrics and exit conditions (e.g. “knowledge transfer complete six weeks after internal hire”). KSX often sees A/C: discovery workshop, project launch, then optional retainer.
Red flags: when not to outsource / not to force in-house
Avoid pure agency when: nobody owns product decisions; compliance demands code residency you cannot audit abroad; requirements change daily under a frozen fixed-scope contract. Avoid forced in-house when: a 90-day existential deadline with zero seats; a one-off mega-migration staffing a permanent team; leadership cannot tech-referee but insists on hiring “the strongest architect.” On red flags, return to the capability map and peak calendar—not emotional “we must have our own team / we must outsource the blame.”
90-day action: from frame to contract or JD
Weeks 1–2: capability map, peak calendar, success metrics. Weeks 3–4: if agency-leaning, write scope and Definition of Done and shortlist studios (see Shanghai selection checklist); if in-house-leaning, write JD and interview loops and start hiring, with transitional contractors for the gap. Weeks 5–8: sign or offer; establish account ownership and weekly status cadence. Weeks 9–12: first shippable increment and a review on hybrid ratio. When you need an external partner, a discovery workshop is a low-regret first buy: align the problem before buying implementation.
Checklist
- 1Core/leverage/commodity capability map done (with this year’s peaks)
- 212/24-month cost compare includes liaison time and risk premium
- 3Ownership inventory for repos, cloud, domains, analytics complete
- 4Internal product decision-maker named with allocated time
- 5Hybrid success metrics and exit conditions on a one-pager
Key takeaways
- Agency vs in-house is capability and peak configuration—not a moral stance
- Without an internal product owner, any agency becomes an expensive feature factory
- Fuzzy account and code ownership punishes you with year-two switching cost
FAQ
- Should a startup outsource first or hire engineer #1?
- If founders can decide product and must validate in ~3 months, sprint v1 with agency/contractors while hiring the long-term core—usually safer than waiting on offers. If the product is deep tech moat and funding allows, prioritize a senior FTE and outsource only non-core.
- We already have an agency—how do we transition in-house?
- Reclaim accounts and docs first, then hire with overlap (4–8 weeks dual-track). Transfer via checklist: architecture, release, monitoring, known debt. Contract for handover days. Avoid parallel big features during cutover; stabilize the release path first.
- Should SEO and brand sites be long-term in-house?
- Content ops and business keyword strategy belong inside; technical SEO, redesign peaks, and complex multilingual builds often fit agency or retainer. Keep metrics and CMS editorial control yours; buy systems and campaign capacity outside.