Our methods are
public on purpose
Most agencies describe their process in adjectives. These pages describe ours in protocols — the prompt counts, the metric definitions, the prioritisation weights, and the limits we will not pretend away.
Why a methodology page exists at all
A published marketing methodology states what will be measured, how it will be counted, and what the measurement cannot prove — before work begins. It exists so results cannot be reframed after the fact by switching to a friendlier metric. Oneskai publishes three: GEO testing, revenue SEO, and the 90-day execution framework.
Three documented methodologies
Each one covers a different question: are we visible to AI, is organic producing revenue, and is the engagement itself on track.
AI visibility2,500 observations per cycleGEO testing methodology
The full protocol for measuring whether AI engines mention and cite your brand: 100 fixed prompts, 5 engines, 5 repeat runs, seven defined metrics — and a published list of what the method cannot tell you.
Read the protocol
Organic revenueReported to the pipeline lineRevenue SEO methodology
How technical foundations, entity architecture and intent mapping connect to CRM outcomes — including the scoring model we use to decide what gets built first, published so you can argue with it.
Read the protocol
Execution90-day operating rhythm90-day growth framework
The engagement cadence itself: what happens in the first fortnight, what ships by day 45, what gets reviewed at day 90, and the conditions under which we recommend stopping rather than renewing.
Read the protocolSix rules that govern all three
These are the constraints the methodologies inherit. They are also the reasons we occasionally lose pitches.
Baselines are frozen first
Success criteria and starting measurements are agreed and locked before work begins. A baseline set after month three is a baseline chosen to flatter the result.
Every claim maps to a source
Market statistics carry a named source and date. Client outcomes carry permission. Anything we cannot evidence is written as an estimate, or not written.
Information gain over word count
A page ships only if it adds something not already available elsewhere — original data, a practitioner opinion, a worked example. Restating the consensus is not publishing.
Limits published with results
Every measurement section names what it cannot prove. A methodology without a stated boundary is marketing copy wearing a lab coat.
Metrics must change decisions
A number earns a place in the summary only if moving it would change what we do next. Impressions rarely qualify. They go in the appendix.
Protocols are versioned
Search surfaces change faster than agency processes do. Ours are dated, reviewed quarterly, and any revision affecting historical comparability is flagged in client reports.
What the first 90 days look like
The methodologies describe how we measure. This is the rhythm the measuring happens inside.
Baseline and instrument
Technical audit, GEO baseline cycle, CRM attribution wired up, success criteria agreed in writing. Nothing is published in this window — we are establishing what "before" means.
Fix blockers, ship first wave
Crawl and render blockers cleared, entity architecture corrected, first Decision-stage pages and answer capsules shipped. First measurable movement usually appears at the end of this window.
Compound and expand
Template-level schema, internal link architecture, second content wave. Second GEO cycle runs against the frozen prompt set so the comparison is genuine.
Review honestly
Full review against the day-zero baseline. If the numbers do not support continuing, we say so and recommend stopping. That has happened, and it is the reason the baseline gets frozen.
What you receive, and when
- Technical audit + prioritised backlog
- Day 14
- GEO baseline, coded observations
- Day 14
- Attribution live in your CRM
- Day 14
- Frozen success criteria, signed
- Day 14
- First shipped work
- Day 30
- Monthly report #1
- Day 45
- Second GEO cycle
- Day 60
- 90-day review + recommendation
- Day 90
Including the stop recommendation
If the channel is not viable for your economics, the day-90 review says that.
Things we will not do
Published because a methodology is defined as much by its refusals as by its steps.
Declined, every time
- Guaranteed rankings or guaranteed AI citations
- Programmatic pages that swap a city name into a template
- Invented statistics or unattributed “studies”
- Client logos or quotes without written permission
- Reporting that changes metric when results disappoint
- Link schemes and paid placement dressed as editorial
Offered instead
- Forecast ranges with the assumptions attached
- Market pages only where genuine local substance exists
- Every figure sourced, dated and checkable
- Anonymised case data where NDAs apply
- Frozen baselines and consistent metric definitions
- Earned coverage through original research
About the methodology
Why publish your methodology instead of keeping it proprietary?
Because the methodology is not the moat — execution is. Publishing the protocol lets prospects check whether our measurement is rigorous before they sign, gives clients a standard to hold us to, and forces our own team to keep the process honest. Agencies that will not describe how they measure usually cannot.
Do you apply the same method to every client?
The protocol is fixed; the inputs are not. Every engagement uses the same measurement definitions, testing cadence and reporting structure, but the prompt set, keyword universe, competitor list and success thresholds are rebuilt from your market and your CRM data. Standard method, bespoke inputs.
How do you decide what gets measured?
A metric earns a place in the report only if a change in it should change a decision. Impressions rarely qualify. Answer ownership on a commercial query, cost per SQL by channel, and pipeline influenced by organic entry point usually do. Everything else goes in an appendix rather than the summary.
What happens when the method says the work is not working?
We say so in the monthly report and propose a change. Our measurement definitions are set before a campaign starts precisely so that a disappointing result cannot be reframed afterwards by switching to a friendlier metric. Baselines are frozen at kickoff and shared with you.
How often is the methodology itself revised?
We review it quarterly and version it. Search surfaces change quickly — AI Overviews, AI Mode and assistant behaviour have all shifted materially in the last two years — so protocols that never change are a warning sign. Revisions are documented, and any change affecting historical comparability is flagged in client reports.
Where this gets applied
Hold us to the protocol
Bring the methodology to the first call. Ask what the baseline will be, which metrics enter the summary, and what the day-90 stop condition looks like. Those are the questions the pages above are written to answer.