Outsource the engineering. Never outsource the judgement.
The strategy layer is cheap to buy and expensive to get wrong. The technical layer is the reverse, which is why so much of it gets built in house by people doing it for the first time. The parts of SEO worth handing to a specialist are the ones with no opinion in them.
Keep in house
Four things that should never leave your agency
- Keyword and market judgement
- The thing your client is paying you for. Outsource it and you are a reseller with a markup.
- Content and editorial voice
- Cheap to buy and expensive to fix. The failure mode is generic rather than wrong, which is harder to spot in review.
- Client relationship and reporting
- Every conversation you do not have is context you do not get. That context is what makes the renewal.
- Link acquisition
- Whoever does this carries the risk for it. Owning the risk and owning the decision have to sit together.
Worth buying
Four things a specialist does faster and better
- Data acquisition
- Finding, pulling and reconciling public sources is a specialist skill with no strategic content. Nobody renews because your agency personally wrote a scraper.
- Schema design
- Get it wrong and every page type built on it inherits the mistake. It is the highest-leverage hour in the project and it is pure engineering.
- Page generation at scale
- Thousands of URLs off one dataset. Either you have built this before or the estimate is a guess.
- Geo and search infrastructure
- Proximity queries that run in the database rather than in application code. Boring, well understood, and easy to get wrong slowly.
None of those four is something a client would be upset to learn a specialist did. All four are things an agency without a data engineer will either avoid or do slowly and badly.
The arithmetic
Against hiring for it
A developer who can do this work in the US costs six figures, takes three months to find and needs enough of this work to stay busy. Most agencies at eleven to fifty people have one or two projects a year that genuinely need it.
A build here is $5,000 to $22,000 and produces the artefact rather than the capacity. Where the pipeline is continuous rather than occasional, capacity is $8,000 a month, which is the point at which hiring starts to make sense instead.
I will tell you when you have crossed that line. An agency running this constantly is better off hiring, and pretending otherwise would cost you more than it earns me.
FAQ
Questions agencies ask
Is outsourcing SEO a bad idea?
Outsourcing judgement is. Outsourcing engineering is not. The split that works is that anything a client would be upset to learn was done by a stranger stays with you, and anything that is a solved technical problem goes to whoever solves it fastest.
How is this different from outsourcing SEO to India?
Different purchase entirely. Offshore SEO is labour arbitrage, sold by the hour or the package, and the reason it is cheap is that it is generic. This is one specialist delivering one artefact for a fixed price. If price per hour is the deciding factor, I am the wrong call and I will say so early.
What if you disappear halfway through?
Nothing is billed in advance and no phase runs longer than two weeks, so your exposure at any moment is a fortnight of work you have already reviewed on staging. The repo and database sit in your client's accounts throughout, not in mine.
How do we control quality on work we cannot evaluate?
By making it checkable rather than trusting it. Record counts are verifiable on staging against the public source. Generated pages are inspectable before publish. Reconciliation scripts run as a dry run and report what they would change before anything writes.
Do we have to explain a subcontractor to our client?
No. Nothing carries my name, and I am not on calls or in documents. Most agencies present it as internal capacity, which it functionally is for the duration.
What is the smallest sensible engagement?
A static build at $5,000. Below that, outsourcing costs more in coordination than it saves in time and you are better off doing it in house or not at all.
How much of our time does this take?
A scope call, a schema decision in week one, and a review at each phase. Four or five hours across a six week build. If a vendor needs more of your week than that, you have hired a team rather than bought a deliverable.
Next step
Bring the project you keep deferring
Thirty minutes. The one that needs ten thousand pages and keeps getting pushed because nobody can scope it. I will tell you whether the data supports it and what it costs, or that it does not, at no charge either way.