What I Learned Bringing Google Ads In-House

Engineers don't usually write about ad spend. But one of the more valuable things I did wasn't shipping code — it was taking our paid search off an external agency and running it myself, saving roughly £48k a year and learning a lot about where engineering meets the P&L.

---

The Decision

We were paying an agency around £4,000 a month to manage our Google Ads. On paper that's a normal arrangement — you hand off a specialised, fiddly channel to people who do it all day. In practice, every time I asked a pointed question about why a campaign was structured a certain way, or what a particular spend was actually returning, the answers were vaguer than I was comfortable with.

The case for bringing it in-house came down to four things:

• Cost — £4k a month is £48k a year, and a meaningful chunk of that was the management fee rather than ad spend that reached a customer.
• Control — I wanted to be able to change a campaign on a Tuesday afternoon because I'd noticed something, not file a request and wait.
• Speed — the feedback loop with an intermediary is always slower than the one with yourself.
• Data ownership — and this was the big one. I didn't just want the ads run; I wanted the data they generated, in our systems, tied to our outcomes.

The Transition

Owning a channel is different from understanding it in the abstract. There was a real learning curve: the structure of campaigns and ad groups, match types, the bidding strategies, the quiet ways Google's defaults spend your money for you if you let them. I'd built plenty of software; I hadn't been the person personally accountable for where several thousand pounds a month went.

The part that came naturally was the instinct to measure properly before changing anything. As an engineer you don't optimise a system you can't observe, and paid search is no different. So the first real work wasn't tweaking bids — it was instrumenting the channel: making sure conversions were tracked honestly, that the numbers I was looking at reflected real outcomes and not vanity clicks, and that I could trust a change had actually caused the effect I was seeing rather than coinciding with it.

Where Engineering Met Marketing

This is where the two halves of the job stopped being separate. The advantage of an engineer running the channel isn't that I could write better ad copy — I couldn't. It's that I could build the tooling and the data plumbing to make better decisions, and tie campaigns back to real business outcomes rather than the platform's in-house metrics.

Google will happily tell you a campaign drove a lot of conversions. Whether those conversions were profitable, whether they'd have happened anyway, whether they were the customers we actually wanted — that's a question you answer with your own data, joined up across systems Google can't see into. Being the person who could do that join, rather than describing it to an agency and hoping, was the whole point.

Reflection

What this taught me is why wearing both the commercial and technical hat makes me a better engineer, not a distracted one.

Once you've personally owned a P&L line — felt the difference between spend that returns and spend that evaporates — you stop treating "the business impact" as someone else's concern that arrives as a ticket. You start weighing technical choices against it by default. Is this optimisation worth the engineering time given what it actually moves? Is this feature solving a real commercial problem or a hypothetical one? Is the cheaper, slightly-less-elegant approach the right call because of where the money is?

Few engineers can credibly say they've owned both sides. Having done it, I default to asking what a technical decision is for in business terms — because I've sat on the side of the table where that question gets answered in pounds.