Vendor evidence should show whether the team solved a comparable problem, what it did, which outcomes can be attributed, what merely coincided, whether disclosure is authorised and whether the method is reproducible. A traffic chart without period, page set, market, baseline and actions proves little. Buyers should not demand confidential client data; anonymised cases need clear boundaries.
Ask for named roles, relevant market and stack experience, sample deliverables, data definitions, editorial/technical approval workflows, and failure or correction procedures. A vendor that clearly states exclusions and non-guarantees is often easier to govern than one presenting success claims alone.
Review prompt-set design, platform and language coverage, a fact-ledger sample, citation-recording method, conflict handling, retest rules and evidence levels. Ask how the vendor distinguishes verified crawler activity, claimed user agents and referral data; how answer variance is handled; and how single observations are reported. Avoid guaranteed inclusion or manufactured mentions.
Request a URL-level audit sample, prioritisation logic, implementation ownership, launch QA, Search Console or log measurement, consolidation rules and rollback. Check whether the vendor distinguishes crawling, indexing, ranking and conversion, and whether removals or link programmes require evidence rather than volume targets.
Require clear separation of wire distribution, sponsored content, interviews and independent coverage; rationale for target outlets; narrative and spokesperson preparation; outreach reporting boundaries; failure handling; link and disclosure expectations; and evaluation across brand search, referrals, secondary citations and sales use. A media list is not a placement guarantee.
Document deliverables, format, cadence, data source, access, owner, approval window and acceptance condition. Mark outcome metrics with observation windows and external variables; use targets and hypotheses rather than guarantees. At close, require account ownership, live-asset inventory, unresolved risks, maintenance guidance and next review date.
Is a vendor unsuitable without an exact same-industry case?
Not necessarily. Comparable buyer journeys, constraints, stacks or markets may be relevant, and a pilot can test fit. The vendor must not imply experience it does not have.
Should buyers demand client dashboard screenshots?
Respect permission and confidentiality. Redacted samples, data definitions and reproducible methods are better evidence than access to another client's private account.
What is the minimum acceptance pack?
It should include baseline, action log, live evidence, measurement sources, limitations, next recommendations and all client-owned accounts and files.
Which promises are red flags?
Guaranteed rankings, indexing, AI recommendations or independent editorial coverage—and performance numbers without definitions—should trigger scrutiny.
Turn vendor comparison into a 100-point scorecard
- Business and market understanding: 15 points
- Case evidence and verifiability: 20 points
- Method and accountable team: 15 points
- Data, account and asset ownership: 15 points
- Deliverables and acceptance criteria: 20 points
- Compliance, risk and exit plan: 15 points
The score is not an absolute verdict; it creates a consistent comparison. Every score should include evidence and notes. When dashboard access is unavailable, triangulate archived pages, timelines, authorised case descriptions and method records. Redacted evidence may be reasonable under confidentiality, but untraceable screenshots should not be accepted on their own.
Five questions to settle before signing
- Which outcomes are controllable and which cannot be guaranteed?
- Who owns accounts, content, data and media relationships?
- Who executes, reviews and reports the work?
- How will underperformance be diagnosed, adjusted or exited?
- Which assets are fully transferable when the engagement ends?
Is the highest-scoring provider always the right choice?
Not necessarily. Budget, maturity, internal capacity and target market still matter. The scorecard exposes differences and risk; the final choice should fit the organisation's most important current problem.
Source：Google Search Central：Google Search Essentials
Source：AMEC：Integrated Evaluation Framework
Source：Reddit Help：Spam policy