Research

Notes on search, advertising and how marketing gets done: research, working ideas and lessons from building the systems around it.

Showing 19 pieces

Search & acquisition

11 pieces

Notes from running paid search: platform changes, bidding and measurement.

The work in Search & acquisition

Approved is not serving

Platform update

Google is extending impression caps based on account trust signals to every Ads product, and because nothing gets disapproved it looks like weak performance rather than a policy state.

Campaign-level lift is not account-level lift

Point of view

Digital Traction reads Google’s AI Max figure as a campaign-level result, the level at which reallocated conversions look identical to new ones.

Updated

Local Services Ads are moving into Google Ads

Platform update

The standalone LSA dashboard is being retired and campaigns migrate to a pay-per-lead Performance Max type, keeping the lead model but removing manual bidding, vertical-level targets and your historical reports.

Updated

Where AI traffic actually lands in GA4

Guide

GA4’s AI Assistant channel covers ChatGPT and its peers but explicitly excludes AI Overviews and AI Mode, which sit in Organic Search, so there is no single AI number to report.

Feeding real outcomes back to the platform

Guide

Form fills teach automated bidding the wrong lesson; importing qualified leads and revenue from your CRM teaches it what a real opportunity looks like.

Updated

Advertising in AI

6 pieces

Tracking what is live, in testing or only announced as advertising moves into AI answers.

The work in Advertising in AI

What answer-engine advertising actually is today

Point of view

It now spans two different models, existing ad systems extending into AI search and purpose-built conversational ad platforms, and the honest thing is to keep them apart.

Updated

Marketing systems

2 pieces

Lessons from building content systems: what an agent should know before it drafts, and what not to build.

The work in Marketing systems

Query fan-out does not mean you need more pages

Point of view

The model generates related queries around yours before answering, and the common response of building a page per variation is the one thing Google explicitly warns against.