Methodology
A protocol you can inspect before you buy.
Measurement is only credible if the method is visible. This is the protocol every Dgenius diagnostic follows, and the limits we state before any engagement begins.
- Subject-matter expertise→
- Technical research→
- Methodology→
- Measurement→
- Data→
- Engineering→
- Revenue
Protocol
Seven steps, in order.
- 01
Intent mapping
We reconstruct the questions your buyers actually ask at each stage — shortlist formation, comparison, budget qualification, technical fit, procurement risk, migration — and express them as prompt families rather than keywords.
- 02
Repeated cross-model observation
Each prompt family is run repeatedly across multiple systems and sessions. Single observations are not evidence: AI answers vary, and only repetition separates a stable pattern from noise.
- 03
State classification
Every observation is classified into a defined state — recommended, mentioned, cited, absent, competitor-held, uncertain — and stored with prompt, model, timestamp, source and competitor.
- 04
Source and citation analysis
We map which documents, domains and third-party sources the answers draw on, and which of them you control, influence, or have no relationship with.
- 05
Factual consistency review
We check whether the facts about your company are consistent and reconcilable across the sources these systems rely on. Contradiction suppresses confident recommendation.
- 06
Technical accessibility assessment
We assess whether your material can be retrieved, parsed and attributed at all: structure, rendering, access rules, entity clarity, machine-readable information.
- 07
Prioritized intervention roadmap
Findings are ordered by expected commercial effect against implementation cost, and delivered as a sequence, not a list.
Limits
What this method cannot do.
Stating limits is part of the method. These constraints apply to every practitioner in this field, including us.
- No one outside these companies can observe how a model weights its inputs. We measure outputs, not internal ranking mechanics.
- AI systems change without notice. Every observation is a timestamped record, not a permanent state.
- Answers vary between sessions and users. We report distributions across repeated observations, never a single screenshot.
- No intervention can guarantee placement, citation or recommendation in any AI system. Anyone who guarantees this is describing something they cannot control.
- Personalization, geography and account history influence answers. We control what we can and disclose what we cannot.
Run the protocol against your company.
$6,500 fixed · 10–14 days. Fixed scope, defined deliverables.
Start Your AI Discovery Diagnostic