The hybrid marketing model: what to keep in-house, what to outsource
How to split marketing work in a hybrid model: what stays in-house, what goes to a specialist agency, what to automate, plus key-person risk and exit terms.
TL;DR
Agency or in-house is no longer the useful question; which work goes where is. Work that needs company context and accountability stays in-house, work that needs deep but part-time expertise goes to a specialist, and repetitive, rule-based work gets automated. Accounts, code and data stay in the company's name throughout, and the exit clause goes into the contract on day one.
A hybrid marketing model is an operating model in which a company splits its marketing by task instead of handing all of it to one side: it keeps decisions and context in-house, runs work that needs deep expertise with a specialist team, and leaves repetitive work to automation. The decision is made task by task.
This post closes our series: agency or in-house marketing covers the growth curve and the real cost of an in-house team covers cost.
The question changed: which work goes where
Among large advertisers, in-house agencies are now common. In 2023, 82% of ANA members had one, up from 42% in 2008 (ANA, 2023). Yet 92% of respondents also work with external agencies, and for them an average of 61% of all work is done in-house.
Companies that brought work in kept their outside partners and redistributed the work. Over the past three years, 65% moved some established business from external agencies in-house, and the top challenges they report are managing workflow as projects multiply, scaling efficiently and prioritising projects (ANA, 2023). Bringing work in turns the bottleneck into a capacity problem.
ANA members are mostly large, US-centric advertisers, so for a Turkish brand with a five-person team the ratios show a direction, not a target. Nor is the direction new. In 2018, 42% of global marketers were taking a hybrid in-house agency approach, and Forrester recommended a blended model (Forrester, 2018).
What AI changes and what it does not
AI mostly changes the cost of production: draft copy, image variants, a first summary of the data. In Gartner's 2025 survey, 22% of marketing leaders say generative AI has let them reduce their reliance on external agencies for creativity and strategy building, while 39% plan to cut agency budgets and 39% seek to reduce labour spending (Gartner, 2025). Respondents were 402 CMOs and marketing leaders in North America, the UK and Europe, surveyed in February and March 2025, most at companies with revenue above 1 billion USD: large-company benchmarks, not SME norms. Both budget lines are under pressure; we read that as companies splitting the work.
What AI does not do on its own is prepare the team. In a survey of 402 senior marketing leaders in North America and Europe, run from August to October 2025, 65% of CMOs say AI advances will dramatically change their role within two years, yet only 32% say significant changes to the CMO profile and skill set are needed (Gartner, 2026). That is expectation data, not a measured outcome, but it shows a gap between expectation and preparation.
Nor does AI repair a broken data layer. Only 52% of senior marketing leaders say they can prove marketing's value and receive credit for it, and three of the biggest barriers are talent-related: soft skills, analytical talent to generate insight, and talent to integrate and analyse data (Gartner, 2024). Put AI on top of tracking that counts the wrong events and you get the wrong decision, only faster.
Work that stays in-house
The first test: does the task need company context that is expensive to transfer, and accountability that cannot be handed over?
- Strategy and pricing. Portfolio priority, market choice, where margin must be protected. A specialist prepares options; the company decides.
- Budget and targets. Total ad budget, acceptable customer acquisition cost (CAC) and payback period. These depend on the company's cash position.
- Final sign-off on brand voice. A specialist can build the creative system; someone inside decides what sounds like the brand.
- Customer conversations. Complaints, returns, questions that need a real answer.
- Ownership. Ad accounts, analytics, the domain, store admin and the database are opened in the company's name.
For daily, continuous, context-heavy work, an in-house team is usually both better and cheaper. If community management or product content fills a full-time week every week, outsourcing it rarely makes sense.
Work that benefits from a specialist
The second test: does the task need deep expertise, but not enough continuous volume to fill a full-time role?
Take server-side measurement. In our setups it takes a few weeks of intensive engineering: an event plan, deduplication between the Pixel and the Conversions API, reconciliation with the store's own order count. After that, a few hours of auditing a month is usually enough. A full-time measurement engineer for it means twelve months of salary for work whose load sits in the first few months. More on this in measurement, conversion and retention.
Ad account structure, experiment design, site speed engineering and the creative system follow the same pattern. Media is split: 54% of in-house agencies already handle some media planning or buying, while those who considered bringing media in-house and have not done so called it too complex in qualitative discussions (ANA, 2023).
A specialist who runs many accounts usually spots a failure pattern earlier, having seen it elsewhere. And one hire cannot cover six disciplines: brand, web, ads, social media, SEO and automation. So the company pays for depth only as far as it uses it.
Work that should be automated
The third test: does the task repeat, can it be written as rules, and is its result measurable? If all three are yes, a person's week is the expensive place for it.
In our own operation, four robots do this work: content, ads, communication and operations. They schedule 60 to 90 posts a month across seven platforms, apply budget and creative rules nightly, send order and shipping emails, and compile the Monday report. Strategy, creative direction and customer conversations stay with people; details in the marketing automation stack for a five-person team.
The automation post's rule applies here too: fix the process before you automate it.
The comparison: task by task
The table shows a typical split for a mid-sized e-commerce brand. The rows shift in every company; what matters is that every row has an owner.
| Task | Keep in-house | Specialist team | Automate |
|---|---|---|---|
| Strategy, pricing, product priority | Decision, accountability | Options from data | Nothing |
| Budget and target CAC | Ceiling and target | Target from margin | Spend alerts |
| Brand voice and creative direction | Final sign-off | Creative system, campaign idea | Resizing, variants |
| Customer conversations | All of it | Nothing | Routing, context |
| Measurement infrastructure | Account ownership | Setup, deduplication, audits | Daily consistency checks |
| Ad accounts and budget rules | Budget approval | Structure, rules, test design | Nightly rule execution |
| Site speed and conversion tests | Priority order | Engineering, test design | Monitoring, alerts |
| Social media publishing | Approval, community | Content system | Scheduling, publishing |
| Lifecycle emails | Consent policy | Flow design | Sending, consent checks |
| Reporting | Reading, deciding | Interpretation | Collection, distribution |
Key-person risk, lock-in and the exit clause
The most neglected question in a hybrid model is where the knowledge lives, and the risk is not only on the agency side: a lone in-house specialist who built the ad account carries it too.
When the person who knows the account leaves, the replacement changes a structure they do not understand. Changed campaigns can re-enter the platform's learning period. Our first learning period after launch runs 30 to 45 days, and a restart can cost part of that time again. Meanwhile the bidding algorithm works with less signal, so spend usually becomes less efficient and customer acquisition cost rises. If tracking was never documented, a broken event goes unnoticed and the algorithm optimises on incomplete data. CAC that took months to bring down can climb back in weeks.
Lock-in usually comes from convenience: the account sits in the agency's Business Manager, the Pixel is attached to the agency's account, the rules live only in the agency's tool. When the contract ends, the company loses its history and an account that had already learned. Our clients pay ad budgets directly to the platforms, so the billing relationship stays with the company and no margin hides in media spend.
Knowledge held by one person is a single point of failure
If only one person knows the ad rules or the tracking setup, or holds admin access, their departure resets the account. Open accounts in the company's name, keep two admins per critical account and write rules down with names and dates.
The exit clause is agreed when the contract is signed. At minimum it states which assets (code, content, accounts, documentation) are handed over and in what form, how many days the handover takes, and that data can be exported as flat files. In our contracts the client takes the code, content, accounts and documentation on exit.
The operating rhythm of a hybrid team
A written division of work still falls apart without a rhythm. Ours has three layers.
Weekly report, Monday 08:00. Spend, revenue, CAC, conversion, and what changed last week and why. It is read before any meeting, so meetings are spent on decisions.
Monthly strategy meeting. Priorities, budget, creative direction. The decision belongs to the in-house owner; the specialist brings options and risks.
Quarterly review. The division of work itself: what moves in-house, what gets automated, which expertise is no longer needed.
The fourth element is direct access. We sit design, engineering and ad management at the same table with no account-manager layer, because every relay loses detail and adds delay. Proximity to the data matters too: among marketing leaders who meet regularly with analytics leaders, 62% say they can prove marketing's value, against 30% among those who meet infrequently (Gartner, 2024). That survey covered 378 senior marketing leaders in April and May 2024.
See how the team is set up on the about page.
Checklist before you move to a hybrid marketing model
- Every marketing task is listed and marked in-house, specialist or automated.
- Strategy, budget, creative direction and customer conversations each have a named in-house owner.
- Ad, analytics, domain, store and social accounts are in the company's name; specialists work with role-based access.
- At least two people hold admin access to every critical account.
- Ad rules, the tracking plan and automation scopes are written down with names and dates.
- Code sits in the company's repository, content in the company's library.
- The exit clause lists the assets, their format and the handover timeline.
- The Monday report gets read, monthly decisions get recorded, the split is reviewed quarterly.
Where to start and what it costs
For most companies the first step is the foundation; without measurement, a single database, automation and account ownership, debating who does what is premature. The Launch Program builds that foundation in eight weeks from contract to launch for $9,900, paid 40/30/30 across three milestones.
After launch, Monthly A is a flat $1,290 a month, and Monthly B is a lower fixed fee ($990) plus a share of net online revenue. Both run on a 12-month term. Details and add-ons are on the programs page; scope is on the marketing automation and performance ads service pages.
If your inventory shows mostly continuous, context-heavy, full-time work, hire: an in-house team is the better choice there. If the picture is mixed, a hybrid model is usually the lower-risk path. The brief takes about ten minutes; a written summary follows by email, and the contract and payment link arrive within one business day. The process page shows each step.
Also available in Türkçe.
