A pay-per-call buyer should not have to scale a source based on a blended publisher average, a handful of promising calls, or a vague assurance that the traffic is “good.”

The buyer should be able to see how a defined source performs in the buyer’s own operation.

That means separating the source from the broader publisher relationship, measuring the same events consistently, keeping buyer handling visible, and making the next decision at the right level: source, campaign, target, geography, schedule, and call type.

Source-level metrics do not remove uncertainty. They make uncertainty easier to manage.

They help a buyer answer practical questions:

Is this source reaching the intended destination?
Are calls connecting?
Are conversations lasting long enough to reach the qualification event?
Are agents converting the calls?
Are outcomes stable across days, targets, and geographies?
Are disputes or duplicates concentrated in one traffic path?
Is the buyer ready to increase the cap?
Should the source stay limited while the buyer fixes an internal problem?
Should the source be paused or disabled?

The goal is not to create a universal score for call quality.

The goal is to give the buyer enough source-specific evidence to scale deliberately instead of scaling on reputation, blended averages, or emotion.

What are source-level metrics in pay-per-call?

Source-level metrics are performance measures tied to a specific, consistently identified traffic source.

A source may represent:

An owned-and-operated website.
A paid-search campaign.
A social advertising funnel.
A consumer-initiated inbound call path.
A live-transfer operation.
A particular landing page and creative set.
A defined sub-publisher.
A specific call center or transfer floor.
A campaign limited to a certain product, language, or geography.

The source definition should be narrow enough that the buyer knows what changed when the metric changes.

“Publisher A” is often too broad.

“Insurance traffic” is almost always too broad.

Even “consumer-initiated inbound” may be too broad when several domains, offers, or caller experiences are grouped together.

A useful source identity connects the same traffic path across:

Review and approval.
Buyer enablement.
Incoming pings or calls.
Routing decisions.
Connected-call records.
Qualification.
Conversion feedback.
Disputes and adjustments.
Buyer billing.
Ongoing performance review.

When the source label changes constantly, or several materially different paths share one label, source-level reporting becomes blended reporting with a more precise-looking name.

That is why source identity is not a cosmetic field. It is the unit of measurement.

For the operating model behind this, see what source enablement means in pay-per-call.

Why blended averages make scaling harder

Blended reporting can answer a broad question:

How is this publisher relationship performing overall?

It cannot reliably answer:

Which traffic path should receive more buyer capacity?

One publisher may have three sources:

A direct consumer-initiated inbound source that has run for months.
A new live-transfer source from a separate team.
An aggregated source that combines several upstream partners.

Suppose the established inbound source performs well enough to lift the publisher’s overall average. The new transfer source may look stronger than it is because the buyer sees only the blended result.

The opposite can happen too. A weak or poorly matched source can pull down a strong source and cause the buyer to pause the entire publisher relationship.

Blending hides the decision the buyer actually needs to make.

It can also hide important operational patterns:

One source sends calls during staffed hours while another arrives near closing.
One source fits the primary target while another is reaching overflow agents.
One source has a stable caller journey while another changed its creative.
One source produces longer conversations but more disputes.
One source converts after a longer reporting delay.
One source is strong in a narrow geography but weak outside it.
One source has a duplicate problem that does not exist elsewhere.

A buyer does not need perfect attribution to make every decision.

It does need enough separation to avoid rewarding or punishing unrelated traffic.

This is also why a buyer should not be required to accept every source by default. Source-level controls and source-level metrics belong together. The control defines what may route; the metrics explain what happened after it routed. Read more in why buyers should not have to accept every source by default.

Metrics should follow the call through the full operating chain

Call platforms can record useful communications facts.

For example, Google Ads call reporting can track details such as call duration, start time, and whether a call connected. Google also allows advertisers to count calls meeting a specified duration as conversions.

Likewise, Twilio’s Call resource includes call status, timestamps, direction, and duration for completed calls.

Those fields matter. They are not the whole business outcome.

A serious pay-per-call operation should distinguish the stages a call moves through:

Offered: The call opportunity was presented to a routing or bidding path.
Routed: A destination was selected.
Connected: The caller and receiving destination established a live connection.
Qualified: The call met the campaign’s defined qualification rule.
Billable: The buyer can be charged under the agreed rule.
Payable: The publisher can be paid under the applicable rule.
Converted: The buyer recorded the defined downstream result.
Disputed or adjusted: The initial treatment was challenged or changed.
Settled: The financial outcome was finalized.

A source dashboard that collapses these stages into “calls” or “conversions” can create false confidence.

A source may have a strong routing rate and a weak connection rate.

It may connect consistently but fail the qualification rule.

It may qualify but convert poorly.

It may convert well but create an unusual dispute pattern.

Each stage points to a different operating question.

The most useful source-level metrics for buyers

No single metric proves source quality. The buyer needs a small set of measures that describe delivery, handling, qualification, outcome, and financial impact.

The exact set should reflect the commercial model and vertical, but the following measures form a practical starting point.

1. Offered or attempted volume

This is the number of source opportunities presented during the measurement window.

It answers:

How much supply was available?
When did it arrive?
Was the arrival pattern steady or concentrated?
Did the source attempt more volume than the buyer could accept?
Did volume change after a creative, schedule, or routing change?

Attempted volume should not be confused with delivered volume.

A source can present 500 opportunities while only 100 are eligible for the buyer. The difference may reflect geography, schedule, source enablement, caps, concurrency, duplicates, or no acceptable bid.

That gap is not automatically a source-quality problem.

It may be a routing-fit problem.

2. Routed volume

Routed volume is the number of calls the routing system assigned to the buyer or target.

This metric should be reviewed with exclusion reasons.

Without exclusion reasons, “not routed” becomes a vague bucket that can hide:

Source not enabled.
Campaign mismatch.
Target mismatch.
Geography mismatch.
Schedule closed.
Cap exhausted.
Concurrency full.
Duplicate exclusion.
Destination unavailable.
No accepted bid.
Reservation or connection failure.

A source should not be blamed for calls the buyer intentionally or operationally could not receive.

3. Connected-call rate

The connected-call rate shows how many routed calls reached a live connection under the operation’s definition.

The denominator matters.

A team may calculate connected rate as:

Connected calls divided by routed calls.
Connected calls divided by accepted bids.
Connected calls divided by attempted calls.

Those are different measures.

The dashboard should state the denominator instead of presenting a percentage without context.

A weak connected rate can point to:

Destination health problems.
Buyer answer delays.
Failed bridges.
Abandoned calls.
Caller hang-ups.
Transfer handoff problems.
Incorrect destination configuration.
Calls arriving outside actual staffed capacity.

Before a buyer reduces source volume, it should determine whether the loss happened before or after the source completed its part of the path.

4. Average billable talk time

Average billable talk time can help a buyer understand whether connected calls are reaching meaningful conversations under the campaign’s billing rules.

It is more useful than raw call duration when the operation applies a defined transformation or excludes time that should not count toward qualification.

But talk time is not a quality score.

A longer call can reflect:

Genuine consumer interest.
A complex intake.
Product comparison.
A successful sales conversation.
Agent hold time.
Repeated explanations.
Consumer confusion.
A poor transfer handoff.
An agent attempting to rescue a mismatched call.

A shorter call can reflect:

A clear disqualifier.
Wrong geography.
Existing-customer service intent.
A caller who immediately recognizes a mismatch.
A buyer agent ending the call efficiently.
A bridge or audio failure.

Average talk time becomes useful when it is interpreted beside qualification, conversion, dispositions, and recordings or QA findings.

It should also include:

The measurement window.
The number of calls in the average.
The exact talk-time definition.
Whether the value is raw, connected, qualified, or billable duration.

5. Qualification rate

Qualification rate shows how often calls meet the agreed campaign rule.

The rule must be defined before the metric is trusted.

A duration-based campaign may qualify after a stated number of billable seconds.

Another campaign may require:

A consumer in an accepted geography.
A defined service need.
A licensed-agent conversation.
A completed intake event.
A buyer disposition.
Another explicitly agreed event.

Qualification rate should not be reverse-engineered from the invoice after the fact.

The buyer, publisher, and operator should understand:

The qualifying event.
The denominator.
The exclusions.
Whether duplicates can qualify.
How transfers and consumer-initiated inbounds are treated.
How technical failures are handled.
Whether a later adjustment can change the result.

A source may appear to improve or decline when the real change is a qualification-rule update.

6. Conversion rate

Conversion rate is often the metric buyers care about most.

It is also one of the easiest to misread.

The buyer should define:

What counts as a conversion.
Which calls are eligible for the denominator.
How long the conversion window remains open.
Whether conversions can be reversed.
How multi-step outcomes are handled.
How quickly agents or systems return dispositions.
Whether missing outcomes are treated as failures or unknowns.

Possible denominators include:

Routed calls.
Connected calls.
Qualified calls.
Billable calls.
Completed opportunities.

A 20% conversion rate from connected calls is not directly comparable with a 20% conversion rate from all routed calls.

Delayed reporting matters too.

A source can look weak on the same day and improve after the buyer’s conversion window closes. A different source may produce immediate outcomes. Comparing them too early favors the faster-reporting source, not necessarily the better source.

Conversion should be reviewed by cohort. Calls from the same delivery period should receive the same opportunity to mature before the buyer compares them.

7. Dispute, adjustment, and duplicate patterns

Disputes should not be treated only as finance cleanup.

They are source-level operating evidence.

A source may have an elevated dispute pattern related to:

Wrong geography.
Repeat callers.
Existing customers.
Caller intent.
Transfer disclosures.
Call duration.
Dead air.
Recording quality.
Misrouted service calls.
Incomplete buyer dispositions.
A disputed conversion event.
A mismatch between source marketing and buyer handling.

The buyer should examine both the dispute rate and the reasons.

A low dispute rate is not automatically proof of quality. It can also reflect a buyer that rarely reviews calls.

A high dispute rate is not automatically proof of bad traffic. It may reveal unclear qualification rules or inconsistent buyer handling.

Duplicate reporting also needs a defined scope and lookback window. A caller can be a duplicate for one campaign and new for another. The source metric should reflect the actual policy rather than a universal label.

8. Billed total and buyer price context

A buyer needs to know what the source cost under the applicable buyer price.

Useful financial views may include:

Billable call count.
Billed total.
Average buyer price per billable call.
Cost per recorded conversion.
Adjustments.
Disputed amount.
Settled amount.

These measures should reconcile to call-level records.

They should not expose publisher payout, exchange margin, or another buyer’s economics.

Price also does not prove quality.

A higher buyer price may reflect scarcity, vertical economics, geography, traffic type, commercial negotiation, or qualification terms. It should be evaluated against the source’s actual buyer-specific outcome.

Every metric needs provenance

A source metric should tell the buyer where it came from.

There are at least four common evidence tiers.

Buyer-specific history

This is performance calculated from the buyer’s own calls.

It is usually the most relevant evidence after enough traffic and outcomes have accumulated because it reflects the source and the buyer operating together.

It still needs context. A buyer-specific result can be distorted by:

A target change.
Staffing problems.
An unusual time period.
A changed source.
Delayed conversion reporting.
A small sample.
A campaign-rule change.

Operator-published benchmark

A benchmark can help when the buyer has little or no history with the source.

It should be labeled as a benchmark, include an as-of date, and describe what it measures.

A benchmark is a starting reference, not a promise.

For a deeper discussion, see how benchmarks help new sources get buyer confidence.

Publisher-provided figure

A publisher may provide historical performance, expected volume, talk time, or conversion information.

That can be useful when the publisher states:

The time period.
The sample size.
The traffic context.
The metric definition.
Whether the figure comes from another buyer.
Whether the source changed afterward.
Whether the number is measured or estimated.

Publisher-provided data should not be displayed as buyer-specific live history.

Blended or network-level figure

This is usually the least specific evidence.

It may include different buyers, campaigns, sources, geographies, or qualification rules.

A blended figure can describe the broader supply relationship. It should carry less weight in a source-level scale decision unless the buyer can confirm that the underlying context is comparable.

Provenance is part of the metric.

Without it, the buyer cannot know how much confidence to place in the number.

Sample size matters, but there is no universal magic call count

A buyer may ask:

How many calls do we need before we can decide?

There is no honest universal answer.

The needed sample depends on:

The event being measured.
The baseline rate.
The size of the change the buyer cares about.
The amount of uncertainty the buyer can tolerate.
How quickly outcomes arrive.
How variable the source and buyer handling are.
The cost of scaling too early.
The cost of waiting too long.

The NIST guidance on sample sizes for testing proportions shows why sample size depends on the baseline proportion, the change the analyst wants to detect, significance, and statistical power.

A pay-per-call buyer does not need to turn every source test into a formal statistical study.

It should avoid pretending that ten calls and one hundred calls carry the same evidentiary weight.

Useful reporting should show:

The number of calls behind each metric.
The date range.
Whether the source was stable during the period.
Whether outcomes are complete.
Whether the calls were concentrated in one day or spread across several operating conditions.
Whether the buyer changed targets, scripts, schedules, or qualification rules.

The number alone is not enough.

One hundred calls delivered to the wrong team may teach less about source quality than a smaller controlled test handled consistently.

Buyer handling must remain visible

Source-level metrics are not purely source metrics.

Once a call enters the buyer’s operation, the result reflects both sides.

A source can send a well-matched caller and still perform poorly because:

The buyer did not answer.
The call waited too long.
The wrong target received it.
The receiving agent was not trained.
The buyer lacked the required geography or product.
The agent opening did not match the caller’s expectation.
The buyer hit concurrency.
The destination was technically unhealthy.
The disposition was missing.
The conversion was reported late.
The source was enabled outside the intended schedule.

This does not mean the source should be excused automatically.

It means the buyer should compare source results with handling metrics.

Useful buyer-side measures include:

Answer rate.
Answer speed.
Abandonment before agent connection.
Target and queue assignment.
Agent or team.
Time of day.
Hold time.
Disposition completion.
Conversion-reporting delay.
Capacity status when the call arrived.

The strongest source report helps the buyer separate three questions:

Did the source send a relevant opportunity?
Did the routing system deliver it to the intended operating path?
Did the buyer handle it competently?

When those questions are blended, the team can make the wrong correction.

It may pause a useful source when the real problem is staffing.

It may add more supply when the real problem is weak agent handling.

It may blame agents when the caller journey was misleading.

Source-level metrics should make diagnosis possible before scale makes the problem larger.

Segment before you conclude

A source can look acceptable in aggregate and fail in one important segment.

Before a buyer scales, it should review performance by dimensions that materially affect the call.

Campaign

The same source may fit one campaign and not another because the offer, qualification rule, price, or conversion event differs.

Target

One destination may have more experienced agents, different licensing, a different script, or better answer speed.

Geography

Performance can vary by state, service area, local demand, licensing, competition, product availability, and buyer coverage.

Schedule

Weekday mornings may perform differently from evenings or weekends. A source’s strongest volume period may be the buyer’s weakest staffing period.

Call type

Consumer-initiated inbound calls and live transfers have different caller journeys and handling requirements.

Traffic or supply type

Direct supply and aggregated supply should not be assumed to have the same control, consistency, or change process.

Source version

A material source change should create a new reporting boundary or at least a visible annotation.

Material changes can include:

A new domain.
A new creative.
A new landing page.
A changed transfer script.
A different upstream supplier.
A new call center.
A new geography.
A changed product promise.
A change from direct to aggregated supply.

A source should not keep the benefit of old history after becoming materially different traffic.

Use metrics to control the test, not merely describe it

Metrics are most useful when they are tied to a defined test plan.

A disciplined source test should specify the following before traffic begins.

The source

Use one stable source identity.

Do not start several unrelated sources and expect the blended result to explain each one.

The allowed campaign and targets

Enable the source only where the buyer is prepared to receive it.

A source may be appropriate for an experienced-agent target and inappropriate for a training or overflow target.

The schedule

Match delivery hours to actual staffed capacity.

A published schedule is not useful if the buyer routinely has no prepared agents during part of it.

Caps and concurrency

Set a starting cap that is large enough to learn from and small enough to stop without creating an avoidable operational or financial problem.

Concurrency matters because twenty calls spread across a day are different from twenty calls arriving in five minutes.

See how caps, schedules, and concurrency shape call flow.

Definitions

Lock the meanings of:

Routed.
Connected.
Qualified.
Billable.
Payable.
Converted.
Duplicate.
Disputed.
Adjusted.

Also define the talk-time measure and conversion denominator.

The review window

Choose when the team will review the test.

Account for outcomes that arrive later.

Immediate-stop conditions

Some events justify a pause before the planned review:

A materially misleading caller journey.
A serious routing mismatch.
An unexplained volume spike.
A repeated technical failure.
A concentrated complaint pattern.
Calls reaching unprepared or unauthorized destinations.
A material source change that has not been reviewed.

A test plan should not force the buyer to continue traffic through a clear risk.

The decision owner

Someone should be responsible for increasing the cap, revising the test, pausing the source, or disabling it.

Otherwise, metrics can become a dashboard everyone watches and no one acts on.

A practical scale, hold, diagnose, or pause framework

A source review should end with an operating decision.

Scale

Scale when:

The source remains the same traffic path that was reviewed.
Calls are reaching the intended target.
The buyer has enough relevant history to support the decision.
Qualification and conversion definitions are stable.
Outcomes have matured.
Buyer handling is consistent.
Dispute and duplicate patterns are understood.
The team has capacity for the next increment.
The next cap or schedule change is documented.

Scaling does not have to mean turning the source on everywhere.

A buyer can increase one dimension at a time:

Raise the cap.
Add hours.
Add a geography.
Add a prepared target.
Increase concurrency.
Expand a product segment.

Incremental scaling preserves the ability to explain what changed.

Hold

Hold the current limits when the early result is directionally acceptable but incomplete.

Reasons to hold include:

Too few calls.
Delayed conversions.
A short measurement window.
A temporary staffing issue.
Incomplete disposition reporting.
A recent source or target change.
Performance that is acceptable but unstable.

Holding is a decision. It is not indecision when the evidence is not mature.

Diagnose

Diagnose when the metrics disagree.

Examples include:

Strong talk time and weak conversion.
Strong conversion and high disputes.
High attempted volume and low routed volume.
High routed volume and low connection.
Good performance on one target and poor performance on another.
Stable qualification with a sudden duplicate increase.
Good daytime performance and weak evening performance.

Conflicting metrics often identify the next question more clearly than a single score could.

Pause or disable

Pause when the operation needs investigation but expects the source may return.

Disable when the buyer no longer wants the source active for the applicable target or campaign.

Possible reasons include:

The source no longer fits the buyer.
The caller journey changed materially.
The source cannot be isolated.
Repeated routing or technical failures persist.
Qualification rules remain disputed.
Performance is consistently below the buyer’s acceptable level after buyer handling is accounted for.
The buyer lacks capacity to serve the source properly.
The source documentation or review status is no longer current.

A source-level control lets the buyer stop one traffic path without ending every relationship associated with the publisher.

A hypothetical source-level review

Consider a hypothetical buyer testing two consumer-initiated home-services sources.

The example is illustrative, not a benchmark.

Both sources are offered to the buyer. The buyer enables each source for the same campaign but uses separate source identities and controlled targets.

After the review window, the dashboard shows this pattern:

Measure	Source North	Source South
Routed calls	80	75
Connected calls	68	69
Average billable talk time	4:10	5:05
Qualified calls	51	49
Recorded conversions	18	12
Disputed calls	2	1

A quick reading might say Source North is better because it has more conversions.

A useful review asks more.

The buyer finds:

Source North reached an experienced team during normal hours.
Source South reached a newer team for part of the test.
Several Source South outcomes are still inside the conversion window.
Source South produced longer conversations, but recordings show agents repeating intake questions already answered on the landing page.
Neither source has a meaningful dispute problem under the defined window.

The buyer does not combine the sources.

It does not scale Source North everywhere.

It also does not reject Source South.

A reasonable decision may be:

Increase Source North’s cap modestly on the existing target.
Keep Source South at the same cap.
Move Source South to the experienced team.
Correct the intake script.
Wait for the remaining conversions.
Review both sources again under comparable handling conditions.

The metrics did not provide a final quality verdict.

They made the next decision more specific.

Common source-metric mistakes

Using one metric as the answer

Long talk time, high conversion, low disputes, or high volume can each be useful.

None is a complete quality score.

Hiding the denominator

A percentage without its numerator and denominator is hard to evaluate.

Ignoring sample size

A source should not receive unlimited volume because its first few calls performed well.

Comparing different windows

A seven-day source result should not be compared casually with a six-month average.

Ignoring delayed outcomes

Conversion rates should not be finalized while a meaningful part of the cohort remains unresolved.

Blending buyer handling with source behavior

A missed call is not the same as a poor caller.

Treating price as proof

Buyer price reflects commercial terms. It does not prove the source will convert.

Failing to separate traffic types

Transfers and consumer-initiated inbounds should not inherit the same conclusions merely because they come from the same publisher.

Letting source identity drift

Old history becomes misleading when the underlying traffic changes materially.

Using a benchmark after buyer-specific history is stronger

General expectations should give way to sustained buyer-specific evidence.

Treating metrics as compliance proof

Performance does not prove that advertising, consent, disclosures, licensing, recording, or other legal requirements were satisfied.

Metrics can identify a pattern that deserves review. They cannot certify the entire caller journey.

Scaling beyond actual capacity

A source can perform well at a low cap and poorly after the buyer overwhelms the team that handled the test.

For a broader readiness framework, see what makes a pay-per-call buyer actually ready to scale.

What publishers should expect from source-level measurement

Source-level reporting should not be used as a one-sided surveillance tool.

It can create a fairer operating relationship when both sides understand the definitions.

Publishers should expect buyers and operators to:

Keep source identities stable.
Separate materially different traffic.
Define the qualification rule.
Return dispositions consistently.
Distinguish routing failures from source failures.
Explain disputes at the call level.
Avoid judging a source before delayed outcomes mature.
Avoid exposing private publisher identity or upstream relationships unnecessarily.
Share useful source-safe feedback.
Record material buyer-side changes that affected the test.

Publishers should also provide:

Consistent source labels.
Accurate traffic-type descriptions.
Current creative, landing-page, or transfer materials.
Honest expected volume.
Change notifications.
Metric definitions for historical figures.
Sample sizes and date ranges.
Clear limits on what prior performance can predict.

For the publisher side of this topic, see why source-level reporting matters for publishers.

The buyer does not need unrestricted access to every private detail behind the source.

It does need a stable buyer-safe identity and enough evidence to make a scoped decision.

The buyer source-metrics checklist

Before scaling a source, review the following.

Source integrity

Is the source defined narrowly?
Is the buyer-facing identity stable?
Are materially different traffic paths separated?
Has the source changed during the measurement window?
Is the traffic type clear?

Metric definitions

Are offered, routed, connected, qualified, billable, payable, converted, disputed, and adjusted events separate?
Is each rate’s denominator stated?
Is talk time raw, connected, qualified, or billable?
Is the conversion event defined?
Is the duplicate policy defined?

Evidence quality

Is the metric buyer-specific, operator-published, publisher-provided, or blended?
Is the date range visible?
Is the call count visible?
Are delayed outcomes complete?
Is the comparison period relevant?
Is the source stable enough for the history to remain meaningful?

Buyer handling

Did calls reach the intended target?
Was the target open and staffed?
Was answer speed acceptable?
Were agents prepared for the caller journey?
Were dispositions returned consistently?
Did the buyer change routing, scripts, schedules, or qualification rules?

Scale readiness

Is the result stable enough for the decision?
Is there capacity for the next increment?
Will only one important variable change?
Is the next cap, schedule, geography, or target expansion documented?
Is a review date set?
Can the source be paused or disabled without disrupting unrelated supply?

A buyer that can answer these questions is not guessing from a dashboard.

It is running a controlled source decision.

How Dependable Calls is approaching source-level metrics

Dependable Calls is being built around curated source enablement rather than unrestricted source discovery.

The current implementation supports a two-gate source model:

Dependable Calls decides which reviewed sources are appropriate to offer to a buyer.
The buyer decides which offered sources to enable for a specific target.

Both gates must be satisfied for a curated source to route.

The current implementation also supports buyer-scoped per-source history including call volume, average billable talk time, billed total, and recorded conversion counts. Those buyer-specific figures are kept separate from DCE-published source benchmarks for buyers with little or no history.

Buyer-facing source labels are designed to provide continuity without exposing protected publisher identity. Source controls operate at the target level so a buyer can test a source with one prepared destination instead of enabling it across the entire operation.

These capabilities have been implemented and tested in the current software. They remain subject to live campaign validation, data completeness, source-by-source review, and continued hardening.

They do not guarantee that a source is compliant, profitable, or ready to scale.

The operating principle is narrower:

Give buyers enough source-specific evidence to make a scoped decision, preserve the limits of that evidence, and let buyer-specific history replace broad assumptions over time.

That is how source-level metrics help buyers scale more confidently.

Not by removing risk.

By making the next change easier to explain.

Want access to curated call sources? Apply to join the Dependable Calls buyer beta.