A method for reviewing B2B social selling products and agencies with provisional desk grades

What to take away

Publish the reviewer, cut-off, candidate pool, criteria and test environment before any verdict.
This edition gives a method plus provisional desk grades, from desk research closed on 1 July 2026. No named reviewer stands behind them yet.
Hard gates stop a purchase whatever the feature total.
Score only candidates with the same need, using one evidence unit per criterion.
Test scripts show workflow completion for a configuration, never sales performance.

Decide the review question and the unit

Start with the decision and the unit. "Choose support for ten England-based account executives to research named accounts and hand accepted conversations into one CRM" is reviewable. "Find the best social selling tool" is not.

Fix the geographic rule. England-based staff, an English registered office and an English operating site are three different populations.

A product review examines a named edition, a service review examines a scope and contracting entity, and a customer-review analysis examines third-party accounts.

Publish the method before the verdict

A reproducible method names who reviewed, the cut-off date, the candidate pool, inclusion rules, conflicts, criteria, weights and test environment.

Keep an evidence ledger holding the claim, the exact source, the access date and the version. The competitor research ledger method shows how to record geography and version beside each claim so a reader can repeat the check.

The CAP Code treats "best" as a claim needing a defined candidate universe. Comparison wording also needs material, relevant and verifiable criteria for identifiable competitors, as the ASA's misleading advertising section sets out.

Publish the review method first

Name reviewers and research cut-off
Record inclusion and exclusion rules
List conflicts and evidence sources
Set criteria, weights, test environment
Distinguish desk research from hands-on work

Grade evidence once, then score

Rank evidence by what it can establish. A current help page or contract schedule proves what the supplier states. A Companies House entry proves a limited identity boundary, not competence or delivery quality.

Independent evidence includes regulator decisions, reproducible testing and a buyer's own logs. Testimonials are leads for questions, not comparable outcomes.

Label each item observed, supplier-stated, reviewer-tested or unresolved. Grade evidence 0 to 4 before any weighted score, and state that rule once.

Set hard gates before scores

Gates stop a purchase regardless of feature totals. Typical gates are correct legal identity, an acceptable contract, lawful data use, authorised account access and a workable exit route.

Check identity on the Companies House register, then link the website to the legal entity through a company number or a contract draft. The selection checklist carries that check through to a written decision record. The ICO sets contract terms where a supplier acts as processor, and the NCSC supplier assurance questions cover governance, incidents and personnel.

Platform rules are gates too. Dux-Soup, Expandi and Waalaxy automate LinkedIn activity, and LinkedIn's prohibited software guidance bars scraping and unauthorised automation.

Write test scripts, not promised results

No buyer-run test results are published here. What follows is the script a buyer should run.

Write test scripts

Use the proposed edition in a test tenant with synthetic or specifically authorised data.
Record browser, role, integration, region, settings and test time.
Run an ordinary case and a duplicate, each with an expected result and an evidence location.
Add access refusal, deletion, failed connection and leaver scenarios.
Repeat the script across every shortlisted candidate before scoring.

A short trial shows whether a workflow completed for one sample. It cannot show annual reliability, adoption or incremental revenue.

Score like with like

Only score candidates that satisfy the same need. Give each criterion one common evidence unit.

Criterion	Common evidence unit	Pass question
Required workflow	completed scripted task	Can an authorised role finish it without a workaround?
Access control	role-to-action matrix	Can each role do only the approved actions?
Data handling	field and transfer map	Are purpose, destination, retention and deletion clear?
Continuity	timed recovery exercise	Can the team recover from absence or failure?
Commercial scope	normalised written quotation	Are users, term, support and exit on the same basis?

Criteria, evidence units, pass questions

Criterion

Required workflow: Completed scripted task
Access control: Role-to-action matrix
Data handling: Field and transfer map
Continuity: Timed recovery exercise
Commercial scope: Normalised written quotation
Evidence quality: Dated evidence grade

Evidence unit

Required workflow: Finish without excluded workaround?
Access control: Each role only approved actions?
Data handling: Purpose, destination, retention clear?
Continuity: Recover from absence or failure?
Commercial scope: Same basis for users, term, taxes?
Evidence quality: Observation, assertion or open gap?

Pass question

Required workflow
Access control
Data handling
Continuity
Commercial scope
Evidence quality

Weights follow the use case. Where CRM writeback is mandatory, a product without the supported integration fails and extra features cannot compensate.

The next table records provisional desk outcomes. Every row is supplier-stated or unresolved, so no row carries a buyer-run grade.

Candidate	Job	Gate outcome	Provisional verdict
Dux-Soup	LinkedIn automation	Fails platform rules	Reject for automated outreach
Expandi	Cloud LinkedIn outreach	Fails platform rules	Hold
Waalaxy	LinkedIn plus email prospecting	Fails on automated features	Reject unless manual only
PhantomBuster	Extraction and automation	Fails on scraping	Reject for LinkedIn scraping
Lusha	Contact data enrichment	Passes only with a lawful basis	Gate before any pilot
DLA Ignite	Social selling training	Identity check open	Shortlist for behaviour change
The Social Hire	LinkedIn training and workshops	Statement of work outstanding	Shortlist for workshops

No price or plan figure is carried here. No dated first-party price page was captured for Dux-Soup, Expandi, Waalaxy, PhantomBuster or Lusha.

Review agencies through delivery evidence

Start with the contracting entity and the exact service. Obtain a statement of work naming deliverables, client inputs, authorised identities, approvals and sub-contractors.

Ask the provider to walk through one anonymised evidence chain, from account criteria through research to CRM hand-off and suppression. Then run a finite pilot with acceptance tests agreed in advance.

UK guidance warns marketing companies not to create or arrange fake reviews. Check the CMA's guidance on reviews for businesses and agencies. For a desk-only agency verdict worked through, see the Bee Social Marketing desk review.

Common questions

Does this edition contain finished reviews?

No. It gives a method and provisional desk grades for seven named candidates, based on platform rules and supplier material.

Why are no prices listed?

No dated first-party price page was captured for these products, and an undated figure cannot be checked by a reader.

Can a short trial prove a tool works?

It shows one scripted workflow completing in one configuration. It cannot establish reliability over a year or incremental revenue.

What if a supplier refuses a gate question?

Record it as an open gate, not a pass, and hold the purchase until evidence arrives.