Dom Carter

Playbooks

The long versions.

Methods I’ve built and run, written up in full. Each one is a few thousand words, starts with the short version, and ends with the questions people actually ask.

  1. Marketing16 min read

    Black Friday 2026, decoded: what the UK's 2025 data tells you

    Black Friday 2026 is Friday 27 November. What UK Black Friday 2025 statistics from 27 sources say, and 10 moves for retailers and brands this year.
    • Online Black Friday week 2025 was flat in the UK, down 1.2% on IMRG's index. What grew was the value of each order, not the number of people buying.
    • The discount was already in the price. Between 83% and 90% of it was live before the week started, and Which? found no deal at its cheapest of the year on the day.
    • Shoppers don't believe the number. 68% doubt the deals are real. The retailers who did well sold proof, like Currys' lowest-ever price guarantee, or held full price, like Next.
  2. AI7 min read

    Half of agencies say they use AI agents. Almost none can prove it.

    Half of US agencies say they run AI agents. Only 3 of the 20 I studied can show a client-verified result. Part 1 of 3 on what agencies are actually doing.
    • Half of US agencies say they use agentic AI for marketing execution. Of the 20 cases I studied, 3 carry a named client result, and all 3 were co-published with the vendor that sold the tool.
    • Agents that actually act cluster in 3 jobs: media setup and reporting, agent-to-agent programmatic buying, and creative generation inside a governed loop. Everything else agencies call an agent is advisory.
    • Every credible case puts a human at the approval point before spend or send. The design is universal and now has brand names: human as judge, our team approves, agents execute.
  3. AI7 min read

    What the agencies that can prove it actually built

    The independent agencies with real results share a stack, a design pattern and one governance rule. Part 2 of 3: the 10 builds worth copying.
    • The small-agency stack has converged without anyone coordinating it: Claude as the reasoning layer, n8n or Zapier for triggers and delivery, Supermetrics or platform APIs for data, Slack or Teams as the interface.
    • The best-evidenced result in the market is Butler/Till's agent-run connected TV campaign: 5.5x supply-chain cost efficiency and setup time down 87%, co-reported with the vendor and the client named.
    • The most replicable result is client reporting: a 25-person agency went from 10 hours to 20 minutes per cycle by connecting Claude to live ad accounts, and 2 other builds landed in the same range.
  4. AI6 min read

    The two agents I built, and the gaps nobody has filled

    An audit engine that runs 120 checks on Google and Meta accounts, a Slack agent that reads release notes, and 5 agents no agency has published. Part 3 of 3.
    • I built a self-serve audit engine: upload Google Ads and Meta exports, get an A to F health score, an executive summary and a 90-day plan. 74 Google checks and 46 Meta checks run as code before the model writes a word.
    • I built a Slack agent that reads Google Ads and Meta release notes on a schedule, classifies each change by relevance, and posts a plain-English digest with an action-needed flag. Nobody reads release notes until something breaks.
    • Neither has a measured result yet. The metrics to capture are listed below, because an unmeasured agent is an announcement, and part 1 was about how little those are worth.
  5. Marketing10 min read

    Reaching the unreachable

    Ad platforms find people who were going to buy anyway. The 4-step method we used to fund and measure the customers they ignore, without missing a target.
    • Most paid media reaches 3 groups: your customers, people already in market, and nobody else. The third group is where growth lives and the platforms won’t show them your ads.
    • Fund it by stripping spend that claims credit for sales that were happening anyway: brand search when you already rank first, and retargeting that has over-indexed.
    • Block invalid clicks. On one account, 14.68% of paid clicks in 30 days came from invalid sources. Blocking them returned that share of the budget without a single extra pound.
  6. Marketing7 min read

    Paid media as a creative test lab

    How a luxury cycling brand kept its loyal performance audience and won a younger one, using paid campaigns to test creative and judging on resonance.
    • The brief: keep a fiercely loyal, performance-obsessed customer base and attract a younger, more progressive one, without the first group noticing a change they would hate.
    • The old model was a top-down content calendar and one brand voice. The new one was weekly creative drops tested through paid media, with the metric set to resonance rather than sales.
    • Cheapest cost per thousand impressions is the honest signal: the platform charges less for creative people react to. Use that to find the message before you spend on reach.
  7. AI6 min read

    Where AI earns its place in ecommerce content

    Most product pages on a luxury site get no video, no model and no campaign shot. How to tier the catalogue and point AI at the pages production never reaches.
    • Audit the catalogue by content tier. On most large sites a few hero products get video, model shots and campaign imagery, the middle gets product and model stills, and the long tail gets product-only shots.
    • The long tail is where the catalogue’s revenue sits in aggregate and where production never goes, because a shoot per product doesn’t add up.
    • AI earns its place there: model-on-body from a product still, short motion from a flat image, a lifestyle context for a product shot. The hero tier stays human.