Marketing engineering · 7 min read

Marketing engineering: the trend of 2027–2028

Ten years ago a good performance marketer was someone who could out-bid and out-structure the competition. Today Google and Meta do the bidding and most of the structuring. What they cannot do is know your margin, your returns, or which customer is worth keeping. Whoever wires that knowledge into the machine wins the next two years.

The job moved, the titles didn't

Look at what a modern ad account actually asks of you: pick a conversion, set a target, upload a feed, hand over creatives. The manual levers that used to separate experts from beginners — keyword lists, bid modifiers, placement exclusions — are mostly gone or ignored by the algorithm. Smart Bidding, Performance Max and Advantage+ took them.

What's left is the input. And the input, for most businesses, is wrong. GA4 reports revenue, not profit. A purchase event fires on an order that will be cancelled tomorrow. A returning customer and a first-time buyer look identical to the platform. The algorithm is optimising brilliantly — toward the wrong number.

Marketing engineering is the discipline of making sure the number the algorithm optimises for is the number the business actually cares about. It sits between marketing, analytics and backend development, and right now almost nobody owns it.

What a marketing engineer actually builds

Measurement that survives the browser. Server-side tagging, first-party identifiers, consent-aware pipelines. Client-side pixels now miss a large share of events; the fix is infrastructure, not another tag.
A profit layer. Order data joined with cost of goods, shipping, payment fees, returns and cancellations — usually in a warehouse or even a plain Postgres replica, surfaced through Metabase or similar. Revenue is a vanity metric; contribution margin per order is what you can scale on.
The feedback loop. Offline and enhanced conversions, Conversions API, value adjustments: sending the corrected numbers back into Google and Meta so the algorithm learns on profit, not on gross.
Behavioural context. Microsoft Clarity, Umami, session data — the "why" behind the funnel numbers, without a cookie banner where possible.
Unit economics reporting. Dashboards that a founder reads on Monday morning and can act on: CAC vs. first-order margin, payback period, cohort retention by acquisition channel.

Why 2027–2028 specifically

Automation has finished eating the easy work. By now every competitor runs the same automated campaign types. The remaining edge is data quality, and it compounds: a better conversion signal improves bidding, which improves data, which improves bidding.
Privacy regulation is no longer a future risk. GDPR enforcement, the DMA, consent mode v2, Safari and Firefox tracking limits — the client-side pixel era is over. Businesses that have not moved measurement server-side are flying on partial data and don't know it.
Margins are under pressure. Rising CPCs and CPMs mean the old "ROAS 4 is fine" rule hides losses. When half the catalogue has a 12 % margin, you cannot run ads without knowing which half.
AI made the tooling cheap. Writing SQL, building a GTM server container, wiring an API — tasks that needed a developer sprint now take a marketer with an LLM an afternoon. The bottleneck shifted from implementation to knowing what to implement.

What this means for teams

The role that will be hired for in 2027 is not "PPC specialist" and not "data analyst". It is someone who can open the ad account, open the database and open the tag manager in the same hour — and understands why the three disagree. Agencies that only touch the ad account will be competing on price. In-house teams that treat analytics as an IT ticket will keep optimising toward GA4 revenue.

If you run an e-commerce business, the practical question for the next planning cycle is simple: what number do our ad platforms optimise for today, and how far is it from our actual profit? If nobody can answer that in one sentence, that gap is the job.

Where to start

Audit which conversion each campaign is optimising to, and what it is worth after returns and cancellations.
Get order-level margin into a queryable place. A read replica and Metabase is enough.
Move the purchase event server-side and start passing real order value — or margin — back to the platforms.
Replace the weekly "ROAS by campaign" report with "contribution margin by campaign". Then move budget accordingly.

None of this is exotic. It is plumbing. But in a market where everyone has the same automated campaign types, plumbing is the product.

Oleksandr Korovin — performance marketing & marketing engineering. I help e-commerce and service businesses measure profit, not just revenue, and scale ads on that data. Get in touch →