Short answer
Sequential dialling allows contact review and call control; parallel dialling requires available agents and rules for discarded attempts. Compare useful conversations and complete follow-up as well as time, and confirm the documented mode in your plan.
Key verification: Verify what someone hears when answering a discarded attempt before a live campaign.
Sources and limitationsChoose sequential dialling if you need to review each contact, prepare the conversation and control when the call is placed. Consider a parallel mode when you have a large list of contacts you may lawfully call, the agent is ready to answer immediately and you can control what happens to discarded attempts, including the person who picks up while the agent is already talking to someone else.
For a small team returning quote requests, the sensible starting point is one call at a time. Any reduction in waiting from a parallel dialler has to be weighed against lost preparation and the experience of people who receive an attempt with no conversation.
What each mode does
| Criterion | Sequential | Parallel |
|---|---|---|
| Active attempts per agent | One at a time | Several at once, depending on configuration |
| Preparation | Allows a walk-through focused on each record | Must be ready for whichever contact answers |
| Pace | Moves contact by contact | Moves in batches of attempts |
| Critical check | When the next call starts | What happens to the other attempts when one connects |
| Outcome to compare | Useful conversations and complete follow-up | The same outcomes, plus discarded attempts and incidents |
“Sequential” here means one line at a time. The next call may start manually or automatically depending on the product. Do not infer behaviour from words such as power, smart or AI: watch the whole path from the contact record to the logged outcome.
Which versions providers document
CloudTalk presents AI Sales Dialer as its current product, with sequential and parallel modes. It describes Power Dialer as an older version receiving no new features, and Smart Dialer as a separate tool for calling from web pages or a CRM, not a campaign dialler. CloudTalk's current versions and modes.
In CloudTalk's current sequential mode, the agent reviews the contact, starts the call or lets a countdown place it, and logs an outcome before moving on. Its parallel mode allows up to five lines, and up to ten on request for some accounts. It connects the first contact who answers and cuts the other attempts; if two people answer almost at the same moment, the second is also cut. It detects answering machines on each line before connecting. It requires an add-on and a Parallel Calling seat per agent. The documentation does not explain what the person who is cut off hears. Sequential mode and parallel mode, consulted on 27 October 2026.
Aircall documents Power Dialer within Professional as a call list with time between conversations for notes and tags before the next number is dialled automatically. It is another implementation of contact-by-contact work. The documentation cited does not mention parallel dialling. Aircall Professional features.
Estimate the improvement that would matter for your process
Example of a sequential block of work: twelve four-minute conversations, twelve one-minute follow-up logs and thirty unanswered attempts of twenty seconds each. Those components add up to 48 + 12 + 10 = 70 minutes, before any extra preparation or breaks.
In that example, unanswered attempts account for ten minutes. Even if you removed all of that waiting, at least sixty minutes of conversation and logging would remain. That is an arithmetic limit of the example, not a prediction of any software's performance.
Second scenario: what parallel costs in contacts
Simplified model with sample data: 25% of attempts are answered, each failed attempt takes 20 seconds and the aim is twelve conversations. Sequential dialling needs 12 / 0.25 = 48 attempts, of which 36 fail. With three lines, each batch has a 1 − 0.75³ = 57.8% chance that someone answers, so 12 / 0.578 = 20.8 batches and about 62.3 attempts are needed.
| Item | Sequential, 1 line | Parallel, 3 lines |
|---|---|---|
| Attempts for 12 conversations | 48 | 62.3 |
| Waiting on unanswered attempts or batches | 12 min | 2.9 min |
| Conversation and logging | 60 min | 60 min |
| Block total | 72 min | 62.9 min |
| Contacts with an attempt but no conversation | 36 | 50.3 |
| People who pick up and are cut off, at most | 0 | 3.6 |
Parallel saves 9.1 minutes, 12.6% of the block, at the cost of 14 more contacts receiving a call with no conversation. The last row is an upper limit: it assumes that everyone who would have answered in a batch picks up before the cut. On average 0.75 people answer per batch and only one can be connected, leaving 0.172 people cut off per batch. Using Ofcom's formula, explained further down in the sections on power or predictive diallers, that limit equals 3.6 / (3.6 + 12) = 22.9% abandonment. The real figure will be lower and depends on how long each person takes to answer, but the calculation shows why you must ask the provider for it.
With five lines the same model raises the limit to 39%. To try different paces and contact rates, the dialler and abandonment simulator works out the effect on conversations and abandoned calls.
What it means in Spain
Article 66.1 of Spain's Ley 11/2022 on telecommunications gives people the right not to receive unwanted sales calls unless they have consented or another basis under article 6.1 GDPR applies. The AEPD's Circular 1/2023 requires the caller to state at the start of every call the company's identity, the sales purpose and the right to object. Someone who picks up and meets a cut-off receives none of that information, so parallel dialling raises more than a question of experience. Before loading the list, apply the do-not-call exclusions (in Spanish).
Set up a comparison that respects the contact
Use comparable lists of people who have asked to be contacted or for whom you have verified that the call is lawful. Keep time slots, reasons and agents equivalent. A dialler being technically available does not grant permission to call.
- Define a useful outcome: need understood and next step agreed.
- Record attempts, people reached, useful conversations and working time.
- Note cut-offs, silences, unexpected call-backs and requests not to be called again.
- Check that the history lets you avoid repeating a contact already handled.
- Set a stop condition if an experience appears that your team cannot resolve.
Do not present “calls made” as if they were all conversations. With parallel dialling, ask explicitly what someone hears when answering a discarded attempt and how it is logged. Use authorised test contacts to check this before a live campaign.
Progressive or predictive: how each mode sets the pace
| Aspect | Power (progressive) | Predictive |
|---|---|---|
| When it dials | When an agent becomes free or finishes wrap-up | Before the agent becomes free, based on a forecast |
| Attempts per free agent | One | Several, adjusted by the system |
| Data it uses | Agent status | Contact rate, talk and wrap-up duration, agents logged in |
| What the person answering hears | The agent, who was already on the call | The agent, a message or, if the message is missing, silence |
| Main risk | Agent time spent listening to ringing | Contacts with no agent available |
Aircall's documentation describes its Power Dialer as a list dialled in order, with optional wrap-up time after each call before moving to the next number: using the Aircall Power Dialer, consulted on 22 October 2026. By contrast, Sinch's predictive configuration includes a maximum percentage of calls dropped because no agent is free, with a default value of 1.0, and warns that some countries restrict that value: Sinch predictive parameters, consulted on the same date.
Abandoned calls, silent calls and their rate
An abandoned call is one that a person answers but the system cannot pass to an agent, so it hangs up or plays a message. A silent call is the same situation without a message: the person answering hears nothing and does not know who was calling. It can also be caused by an answering machine detector that mistakes a person for a machine.
Ofcom, the UK regulator, calculates the rate per calendar day and per campaign or call centre: abandoned calls divided by the sum of abandoned calls and calls answered by a person and connected to an agent. Unconnected attempts and calls to answering machines are excluded from the calculation. Do not confuse this rate with abandonment in an inbound queue, where the person hanging up is the customer who was waiting.
Predictive dialler: what the rules say in the UK and Spain
Ofcom's persistent misuse policy, published on 20 December 2016 and applicable from 1 March 2017, is on its page about silent and abandoned calls. Its main points:
- Silent calls are the top enforcement priority.
- It no longer sets a maximum rate: the 3% in the 2010 policy was a criterion for prioritising cases, and Ofcom states that it was never a permitted margin.
- The message on an abandoned call should identify the caller, explain the attempt, give a basic rate number to ask not to be called again and contain no marketing. Ofcom considers whether it starts as quickly as possible and within no more than 2 seconds.
- Calling again within the following 72 hours without guaranteeing an available agent aggravates the case.
- Letting the phone ring for less than 15 seconds or withholding the calling number also aggravates the harm.
In Spain, article 66.1 of Law 11/2022 recognises the right not to receive automated calls without human intervention for commercial purposes without prior consent, and not to receive unwanted commercial calls unless there is consent or another legal basis under article 6.1 of the GDPR. AEPD Circular 1/2023 requires, among other safeguards, that each call begins by stating the identity of the company, the commercial purpose and the right to object. The rules cited do not set a Spanish abandonment rate; do not apply the British percentage as if it were a Spanish rule. A silent call cannot provide that information, and if you plan to use a recorded message, ask your adviser whether it falls under the automated calls provision.
In addition, the Resolution of 14 April 2026 allocates the 400 range to commercial calls by companies covered by Law 10/2025. Those numbers can only make outbound calls and cannot receive calls. If someone wants to call back to object, they will need another channel. Confirm with your operator whether your company is obliged to use it.
Progressive or predictive: when each mode fits
| Situation | Recommended mode | Reason |
|---|---|---|
| One to five agents | Power | Few agents free at the same time; the forecast fails easily |
| Short list or recent enquiries | Power | Each contact is worth more than the minutes saved |
| Conversations that require preparing the record | Power | The agent needs to see the context before the person answers |
| Large team, large list and stable contact rate | Evaluate predictive | The saved waiting time may pay off if abandonment is measured every day |
| Vendor that does not report abandonment per campaign and day | Power | Without that figure you cannot demonstrate control |
Before loading the list, check it with the guide to preparing a CSV for the dialler and apply the do-not-call exclusions (in Spanish).
Calls to other countries: legal hours and data protection in Spain
If some of your outbound calls reach consumers in Spain, note that article 96.2 of the consolidated General Law for the Protection of Consumers and Users states that telephone calls must never be made before 9:00 or after 21:00, and bans them on public holidays and weekends. That window sets a legal limit for recipients in Spain; for a recipient in another country, the applicable reference is that of their own jurisdiction, which this guide neither sets nor replaces.
A named person's time zone and expected location are personal data: article 4.1 of the GDPR explicitly cites "location data" as an example of an identifier. Storing that data in the CRM or in a scheduling record means handling it under the regulation: stating the purpose, limiting it to scheduling the agreed call and not keeping it longer than necessary. Knowing who confirmed the data and when also helps justify that processing if the recipient themselves raises a query.
Common mistakes and how to spot them
| Mistake | Symptom | Check |
|---|---|---|
| Measuring only calls per hour | Activity rises but useful conversations do not | Compare useful outcomes per agent hour in both modes |
| Not logging cut-offs | Call-backs from numbers nobody remembers handling | Ask for a report of attempts cut after pick-up |
| Agent without context | Hesitant first seconds and questions already answered | Listen to a sample and check whether the record appears before connection |
| Calling again someone already cut off | The same person receives several calls within a few days | Review history by normalised number, not by campaign |
| More lines to make up for a poor list | Contact rate falls and abandonment rises at the same time | Clean the list before adding lines |
Decide with a purchase sheet
The sheet can record product version, available mode, licences required, preparation time, handling of discarded attempts, logging in the CRM and the procedure for pausing. Ask the provider to demonstrate every point that is essential.
| Situation | Sensible mode | Reason |
|---|---|---|
| Returning recent requests | Sequential | Each contact expects the call and preparation adds value |
| Waiting on unanswered calls is shorter than logging | Sequential | The possible saving is small; improve logging first |
| Large list, low contact rate and short script | Evaluate parallel | Waiting weighs more, provided cut attempts are measured |
| Provider with no report of cuts after pick-up | Sequential | You cannot show what experience the contact receives |
Stay with sequential mode if individual preparation and continuity of conversation are worth more than the waiting you could save. Afterwards, review the team metrics against the same outcome criterion.
Where a predictive dialler fits
A predictive dialler does not multiply lines per agent: it calculates how many attempts to launch for the team based on the agents who will become free. The comparison between the two modes and how to measure their abandonment are further up, in the sections on the power or predictive dialler. If you compare providers, ask for the same test list and compare conversations handled, not just attempts.
From the author's experience
After eleven years implementing and supporting contact centre software, I would not publish a percentage of cut-offs after pick-up without having measured it myself. With two or three lines per agent, I would watch closely for simultaneous answers, disconnected calls, and the time between pick-up and an agent actually being on the line. What the cut-off person hears in parallel mode depends on the dialler and how it handles attempts it cannot assign: in validation I would test whether they hear silence, a recording or a straight cut, and check what that particular version lets you configure.
There is no list size that by itself justifies moving to parallel dialling; I would consider it when idle time from unanswered attempts is significant and there is real capacity to absorb contacts without leaving anyone hanging. The first configuration risk I would check with several lines is turning on too many for the agents available without adjusting simultaneous-connection limits, along with agent states and the rules deciding when another call is launched.
To compare sequential against parallel honestly, I would build two comparable groups of records, with similar timing, agents and contact criteria, and measure useful conversations per hour, abandonment, cut-offs after answer and complaints, not just the number of attempts made.
Sources and limitations
Documentary review: . Content type: Documentation-based comparison.
Scope: Comparison of dialing modes, with an example calculation. The functions and versions are verified in documentation, without testing in accounts.
- CloudTalk's current versions and modeshelp.cloudtalk.io
- Sequential modehelp.cloudtalk.io
- parallel modehelp.cloudtalk.io
- Aircall Professional featuressupport.aircall.io
- Article 66.1 of Spain's Ley 11/2022www.boe.es
- Circular 1/2023www.boe.es
- using the Aircall Power Dialersupport.aircall.io
- Sinch predictive parametersdocs.cc.sinch.com
- silent and abandoned callswww.ofcom.org.uk
- Resolution of 14 April 2026www.boe.es
- consolidated General Law for the Protection of Consumers and Userswww.boe.es
- GDPRwww.boe.es
Check current terms
Consider these options if they solve the problem described. Confirm features, limits and availability in your country.
CloudTalkOfficial plans and termsAircallOfficial plans and termsEstos enlaces llevan a las páginas oficiales de cada proveedor. CallsIQ no usa enlaces de afiliación. Our commercial policy.
Official sources, explained calculations and clearly labelled examples. Read about our methodology and use of AI in writing.