Use the checklist before the sales call with the vendor
A cold calling vendor can look excellent in a demo because the cleanest workflow is the one being shown. The buying risk lives in the messy cases: duplicate records, wrong roles, objections, uncertain number types, cross-border lists, failed integrations, callbacks, recordings, and the moment a qualified conversation must move to your sales team.
Before comparing price, decide which operating evidence the vendor must provide. A credible partner should be able to explain how data enters the system, who can call it, how suppression works, how calls are identified, what is stored, how quality is reviewed, and what you can take with you when the relationship ends.
Data provenance and permitted use
Ask where phone data comes from and what rights attach to it. “Verified” is not enough. You want to know the verification method, freshness, number type when known, geographic scope, and whether the data license permits the intended use. If the vendor combines multiple sources, ask whether source-level provenance survives in the exported record.
Then test correction. Give the vendor a record with an outdated title and see how the change is handled. Ask what happens when two sources disagree. Ask whether your team can mark a number as bad and prevent it from quietly returning in a later refresh.
A useful contract or operating statement should also distinguish data you supply from data the vendor licenses. That distinction affects portability and what happens after termination.
Suppression and objection handling
This is one of the fastest ways to separate a mature operation from a superficial one. Ask the vendor to demonstrate, not merely describe, how an objection becomes suppression.
Test at least five paths: a prospect says “do not call”; a manager adds a number manually; the same contact appears in a second campaign; an old CSV is re-imported; and another representative tries to call the account. The expected behavior should be clear.
Cross-border programs need additional routing. The UK requires B2B live marketing callers to respect TPS/CTPS registrations and prior objections subject to applicable rules. Canadian B2B calls are exempt from the National DNCL portion of the UTRs but remain subject to other telemarketing rules and internal do-not-call obligations. U.S. B2B calling has its own federal and state considerations. A vendor should support your policy; it should not sell “global compliance” as a magic setting.
Dialing method and connection experience
Ask exactly how calls are placed. Is it one rep to one number? A power queue? Parallel-assisted dialing? Predictive logic? Prerecorded audio? An artificial voice? The answer affects both customer experience and compliance review.
If multiple numbers can be dialed at once, ask what happens when more than one person answers. Measure delay before the live representative joins. Ask how callbacks reach the correct team. Check how caller ID is selected and whether local-presence features accurately represent the caller.
If any feature uses an artificial or prerecorded voice, require a separate explanation. The FCC has made clear that AI-generated human voices fall under TCPA restrictions for artificial or prerecorded voice calls. Do not let a broad “AI sales agent” label blur that distinction.
Identity, scripts and claims
You need control over what the caller says the company is, why it is calling, and what claims are permitted. Ask whether the vendor can lock or version required disclosures while still letting representatives speak naturally.
This matters because the risk is not only calling the wrong number. The FTC's 2024 Telemarketing Sales Rule changes expanded prohibitions against material misrepresentations in business-to-business telemarketing. A caller who invents pricing, authority, customer results or urgency can create a problem even if the list itself is acceptable.
Request sample talk tracks, objection handling, and the vendor's process for approving changes. If the vendor trains with “whatever works,” ask how managers detect misleading statements.
Recording, transcription and AI review
Call recording and transcription can be powerful for coaching, but the vendor should not assume one recording rule applies everywhere. Ask which calls are recorded, how notice or consent is handled where required, how long recordings are retained, and whether customers can configure retention by market.
For AI summaries, ask what the model is allowed to infer. A summary that converts “maybe next quarter” into “qualified opportunity” can distort the pipeline. The original call outcome and rep-entered disposition should remain distinguishable from machine-generated interpretation.
Also ask whether recordings or transcripts are used to train vendor models, whether that can be disabled, and what happens to derived data when your contract ends. Those are operational and data-governance questions, not just security questions.
Quality assurance and coaching
A vendor that promises experienced callers should be able to show how experience is maintained. Ask what percentage of calls managers review, how calls are sampled, which behaviors are scored, how calibration works across coaches, and what triggers retraining.
Avoid scorecards that reward only script adherence. Good cold calling requires listening, relevance, truthful positioning and clean next-step decisions. A representative should be able to stop when the premise is wrong instead of forcing every conversation toward a meeting.
Ask for examples of how a weak campaign was diagnosed. The answer should distinguish data problems, connect problems, message problems, skill problems and handoff problems rather than blaming “bad leads” or “bad reps” by default.
CRM integration and auditability
Have the vendor demonstrate the full record lifecycle. A call should create or update the correct person, company, activity, disposition, notes and next step without duplicating ownership or losing prior objections.
Then break the integration on purpose. Disconnect it briefly. Change an account owner. Add a duplicate contact. Ask what retries look like. Ask which system is authoritative if the vendor and CRM disagree.
An audit-friendly workflow keeps timestamps, campaign identity, caller identity and enough history to reconstruct what happened. That record is useful for management even when no legal retention rule applies. It is essential when a company needs to investigate a complaint or reconcile a disputed lead.
Metrics and acceptance definitions
Before launch, write definitions for live connect, relevant conversation, qualified conversation, meeting booked, sales-accepted next step and opportunity. Ask the vendor to report against those definitions.
A per-meeting contract can create bad incentives if “meeting” is loosely defined. Require evidence that the right role agreed to a meaningful next step. If your sales team rejects the meeting, record why. That feedback should return to the calling operation.
The vendor dashboard should allow segmentation by list source, caller, account type and period. One overall average is too easy to game and too weak for diagnosis.
Commercial structure and hidden cost
Compare the total operating cost, not just seat price or per-meeting fee. Include onboarding, data, telephone minutes, number rental, recording storage, AI credits, integration services, manager time, minimum commitments, overages and exit fees.
Ask what happens during ramp. If the first month is mostly training and setup, a long commitment can make the apparent monthly price misleading. For outsourced teams, clarify whether replacements, retraining and quality management are included.
Also ask who owns the playbook created during the engagement. If your company pays to discover objections, talk tracks and segment insights, those learnings should not disappear when the vendor relationship ends.
Security and access
Cold calling systems touch contact data, notes and sometimes recordings. Ask about role-based access, administrator controls, export permissions, account offboarding, authentication, incident handling and data location where relevant.
The practical test is employee turnover. When a caller leaves on Friday, can access be removed immediately without breaking ownership of the records? Can you see what they exported? Are API keys and integrations tied to named owners rather than a shared account?
Security review should match the sensitivity of the data. Do not let a long questionnaire substitute for understanding the actual data flow.
Pilot before commitment
Give finalists the same bounded pilot. Use a representative segment with clean success criteria and a small number of deliberately difficult records. Observe preparation time, connection quality, note quality, suppression, CRM behavior, manager effort and sales acceptance.
Do not change the target, offer and qualification rule for each vendor. Otherwise the test compares campaigns rather than systems.
At the end, ask three questions: Did the vendor create more useful conversations per unit of total effort? Did the operating controls behave as expected? Can your team explain why the results happened? If any answer is no, more volume will not fix the design.
Reference checks that are actually useful
Vendor references are most useful when you ask about operating friction rather than satisfaction. Speak with a customer whose team size and market resemble yours. Ask what had to be rebuilt after the first month, which promise was harder to implement than expected, how often the customer needs vendor support, and what happens when data or integrations fail.
Ask specifically about quality drift. Did results change after the initial launch team handed the account to normal operations? How stable is caller staffing? How quickly are weak representatives coached or replaced? If the service uses offshore or distributed teams, how are language, time zone, supervision and account context handled? None of those models is inherently good or bad; the point is whether the vendor has a repeatable answer.
For software references, ask how much of the product is used in practice. A buyer can pay for a broad platform and still run the daily workflow in spreadsheets because the core integrations are inconvenient. That is a warning sign no feature matrix will reveal.
Red flags during procurement
Several answers should slow the process down. “Our database is 99% accurate” without a definition of accuracy is not useful. “We handle compliance for you” without jurisdiction, recipient type and calling method is too broad. “Our AI qualifies every lead” without showing the underlying evidence invites pipeline inflation. “Unlimited calls” can hide telephone restrictions, quality trade-offs or fair-use limits.
Another red flag is reluctance to provide sample exports. If a vendor cannot show the shape of call history, suppression records, notes and configuration before contract signature, portability is uncertain. Likewise, a platform that cannot explain how it handles a do-not-call request across duplicate records is not ready for scale.
The procurement team should reward precise limitations. A vendor that says “this feature is not appropriate for that market” may be more trustworthy than one that claims every feature works everywhere.
Final contract test
Before signature, compare the written contract with the operating promises made in demos. Check service levels, data rights, support scope, subcontractors, retention, termination, and any limits on exports. A feature promised verbally but excluded by the order form is not an operating control.
Assign an internal owner for the vendor relationship. Someone must reconcile quality, policy changes, suppression, data sources and sales feedback after launch. Procurement ends at signature; governance begins there.
The 25-question sign-off
Before signing, make sure someone can answer: data source, verification date, usage rights, number type, geographic routing, suppression ownership, preference-service screening, caller identity, dialing mode, artificial/prerecorded voice status, recording rules, script ownership, prohibited claims, disposition definitions, callback routing, CRM authority, retry behavior, coaching sample, calibration process, sales acceptance, reporting denominators, data export, recording export, deletion/retention, and termination assistance.
The exact legal answer will vary by market. This checklist is an operational procurement framework, not legal advice. Use qualified local counsel or regulators for jurisdiction-specific decisions.
Sources
- https://www.ftc.gov/news-events/news/press-releases/2024/03/ftc-implements-new-protections-businesses-against-telemarketing-fraud-affirms-protections-against-ai — FTC 2024 rule update on B2B telemarketing representations and records.
- https://www.ftc.gov/business-guidance/resources/complying-telemarketing-sales-rule — FTC Telemarketing Sales Rule compliance guidance.
- https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf — FCC AI-generated voice declaratory ruling.
- https://crtc.gc.ca/eng/phone/telemarketing/tobligations/rules-regles.htm — CRTC key telemarketing rules.
- https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing/ — ICO business-to-business direct marketing guidance.