Search & shopping
Google Ads and Microsoft Advertising: account and campaign architecture, keyword and query management, negative strategy, Shopping and Performance Max with feed work, bid strategy selection and budget pacing.
Services
Six service lines. One measurement spine underneath all of them.
You can buy any of these on its own. What you cannot buy from us is media management without the measurement that tells you whether it worked — that combination is how budgets get spent against a number nobody can defend.
01 — Paid media
Media planning, buying and optimisation
Planned and bought in-house. No trading desk in the middle, no undisclosed margin, no platform relationship that decides what we recommend.
Paid social
Meta, TikTok, LinkedIn, Pinterest, Snapchat and Reddit. Audience architecture, creative rotation tied to a testing calendar, placement and bid controls, and catalogue campaigns for retail.
Retail media
Amazon Ads and retailer networks: sponsored products and brands, DSP where it earns its place, and the part most advertisers skip — reconciling retail-media reported sales against your own order data.
Programmatic & display
DV360 and The Trade Desk for prospecting, retargeting and reach extension, with supply-path and inventory-quality review so spend is not quietly leaking into made-for-advertising inventory.
What we do not do: buy on a percentage of spend, accept platform rebates or incentive tiers, or take a margin on media. A flat fee means our recommendation to increase budget is the same whether you act on it or not.
02 — Measurement & analytics
The number everyone agrees on
This is the part that decides whether the other five service lines are worth anything. We build it first, and we build it in your name.
Server-side data collection
A tagging container you own, hosted in your cloud account, receiving events from the browser and from your servers. Conversions are deduplicated, enriched and forwarded deliberately — rather than broadcast to every platform by a client-side tag that nobody has audited since installation.
- Consent-aware: nothing fires before the signal it depends on
- Enriched server-side with order value, margin band and customer type
- Platform delivery controlled per destination, not all-or-nothing
- Full event log you can inspect, replay and diff
Attribution you can interrogate
Rules-based and data-driven models are only useful if their inputs are clean. We build the model on top of the deduplicated event stream and reconcile the output against your finance system — so a channel's claimed contribution is a testable statement, not a dashboard default.
- Multi-touch and position-based models kept side by side
- View-through separated from click-through, always
- Modelled conversions labelled as modelled everywhere they appear
- Reconciliation report: platform claim vs. your own revenue
Incrementality testing
Attribution tells you where credit was assigned. It cannot tell you what would have happened anyway. For that you need a holdout — and we would rather run a smaller honest test than a large misleading one.
- Geo holdouts for always-on channels
- Platform lift studies, read with the right scepticism
- Spend-down and pause tests where the risk is acceptable
- Results stated with their confidence, not rounded up into a headline
DELIVERABLES — MEASUREMENT
- Server-side container
- In your cloud
- Event schema
- Documented
- GA4 property
- Aligned to it
- Warehouse pipeline
- BigQuery
- Attribution models
- 2+ compared
- Reconciliation report
- Monthly
- Incrementality plan
- Quarterly
- Change log
- Every edit
- Ownership
- Yours
Everything above sits in accounts registered to you. If the engagement ends, nothing is revoked and nothing needs exporting.
03 — Audit & evaluation
An independent read, before anyone commits to anything
Most audits are a sales document with a findings section stapled to the front. Ours is a deliverable you can act on with a different agency, or with nobody at all.
Where the money went
- Spend by campaign, channel and query, over 12 months
- Wasted spend: broad match drift, irrelevant placements, cannibalised brand
- Auction overlap and internal competition between campaigns
- Creative fatigue curves and refresh cadence
- Budget pacing against seasonality and margin
Whether the number is real
- Conversion event inventory: what fires, where, and how often
- Duplicate and double-counted events across containers
- Consent handling and what is lost before it is sent
- Platform-reported vs. analytics-reported divergence
- Attribution windows and whether they flatter a channel
Who owns what
- Asset ownership: accounts, containers, properties, cloud projects
- Access review — who still has login to what
- Data-sharing settings per platform, and what they expose
- Retention and PII handling in the event stream
- Continuity risk if a vendor or contractor leaves
Output is a written report, a prioritised fix list with effort and impact estimated, and a walkthrough session. Findings are yours; there is no clause that makes acting on them depend on hiring us.
04 — Reporting & dashboards
Reporting built so it can be argued with
A dashboard is only useful if two people looking at it would compute the same number by hand. That means one definition per metric, written down, with a named owner and a data source — and a visible trail when a definition changes.
- One warehouse — platform APIs and your own systems land in the same place
- Metric dictionary — every number has a formula, a source and an owner
- Weekly delta view — what moved, by how much, and the likely cause
- Monthly narrative — written, not a screenshot deck
- Change log — every edit we made to an account, dated, with the metric it touched
- Reconciliation tab — platform claims set against your own revenue
A MONTHLY REPORT CONTAINS
- Spend and revenue by channel
- Reconciled
- Agreed primary metric
- Trend + target
- Tests concluded
- Result + decision
- Tests running
- Status
- Changes made
- Dated log
- Open questions
- Honest list
- Next month's plan
- With reasoning
05 — Creative & landing pages
Built to be tested, not to be admired
Creative is a variable, not a deliverable. We produce enough of it to run a real testing cadence, and we design the landing path so a lift has somewhere to show up.
Static & motion ad creative
Concepted against the offer and the audience, produced in the variants the platforms actually reward.
Testing calendar
A rolling queue with a hypothesis per test, a success measure fixed in advance, and a decision date.
Landing pages
Fast, accessible pages wired into the same event schema, so a form submit is counted once and counted correctly.
CRO
Funnel instrumentation, friction analysis and experiments prioritised by expected value rather than by opinion.
06 — Data governance
The unglamorous part that prevents most incidents
Consent mode, retention windows, PII handling and platform data-sharing settings are configured deliberately, documented, and reviewed on a schedule. Defaults are not decisions — they are whatever the platform chose on the day the tag was installed.
- Consent signals wired end to end, including server-side
- Retention set per data class, not globally
- PII never forwarded to ad platforms; hashed identifiers only where intended
- Access review each quarter, with a written record
- Vendor and processor list maintained with your legal team
HOW WE ARE PAID
Flat fees. No percentage of media spend. No rebates, no volume incentives, no resale margin on platforms or tools.
Where a tool is genuinely needed we will recommend the cheapest option that does the job, including free tiers, and you buy it directly. We would rather bill you for hours of thinking than for a licence we marked up.
Ad spend always sits in your accounts and on your cards. We never hold client money.
Tell us which of these you actually need
Most teams do not need all six. Send a note describing the problem and we will say which service line addresses it — and which ones you can skip.