Cloud dialer vs self-hosted dialer: which deployment model fits your team

Cloud dialer or self-hosted? It is the second question every team asks after "which dial mode" and the answer drives 60% of your total telephony cost over three years. This post is the structured trade-off, with the actual numbers and the operational realities — written so you can decide without a 90-minute vendor pitch.

What each model means

Cloud (SaaS) dialer

The vendor runs the dialer software, the database, the recording storage, and the SIP gateway. You access via browser softphone and the admin UI. Carrier minutes are either bundled (vendor pays the carrier on your behalf and marks up) or passthrough (you bring your own carrier contract).

Typical setup time: 3-7 days from signed contract to first live call.

Self-hosted dialer

You run the dialer software (typically Asterisk + Postgres + an admin layer) on your own infrastructure — AWS, GCP, on-prem datacentre, or a colocated server. You hold the SIP trunk contract directly. Recordings live on your storage.

Typical setup time: 2-4 weeks for a team with basic Linux ops capability.

Hybrid

Control plane (admin UI, queue management, analytics, reporting) on the vendor's cloud; media servers and recording storage on your infrastructure. Combines fast onboarding with data residency for sensitive workloads.

The cost math at three team sizes

The math shifts as you scale. Here is the same workload — agents making outbound calls in India at typical talk-time and minute volumes — priced across both models. Carrier minutes are at ₹0.50/min for self-hosted (direct trunk) vs ₹1.10/min for bundled cloud (the typical markup).

Small (10 agents, ~110,000 outbound minutes/month)

Cloud (bundled): Platform fee ₹15,000-25,000 + minutes 110k × ₹1.10 = ₹121,000 = ~₹140,000/month

Cloud (passthrough): Platform fee 10 × ₹1,200 = ₹12,000 + minutes 110k × ₹0.50 = ₹55,000 = ~₹67,000/month

Self-hosted: Server ~₹8,000/month + DIDs ₹3,000 + minutes ₹55,000 + ops time (5 hrs/month × ₹500 = ₹2,500) = ~₹68,500/month

At 10 agents, the cloud-passthrough and self-hosted models cost the same. The differentiator is operational overhead (about 5 hours/month of Linux work) vs vendor management.

Mid-market (50 agents, ~550,000 outbound minutes/month)

Cloud (bundled): Platform fee ₹60,000-1,00,000 + minutes 550k × ₹1.10 = ₹605,000 = ~₹680,000/month

Cloud (passthrough): Platform fee 50 × ₹1,200 = ₹60,000 + minutes 550k × ₹0.50 = ₹275,000 = ~₹335,000/month

Self-hosted: Servers ~₹25,000/month + DIDs ₹15,000 + minutes ₹275,000 + 1 part-time SRE hour/day × ₹800/hr × 22 days = ₹17,600 = ~₹332,000/month

At 50 agents, the savings between bundled-cloud and self-hosted exceed ₹3.5 lakh/month. That pays for a senior SRE plus their hardware budget several times over.

Large (200 agents, ~2.2M outbound minutes/month)

Cloud (bundled): Platform ₹2-3 lakh + minutes ₹24.2 lakh = ~₹27 lakh/month

Cloud (passthrough): Platform 200 × ₹1,200 = ₹2.4 lakh + minutes ₹11 lakh = ~₹13.4 lakh/month

Self-hosted: Servers ~₹80,000 + DIDs ₹50,000 + minutes ₹11 lakh + 1 full-time SRE ₹1.5 lakh = ~₹13.8 lakh/month

At 200 agents, the bundled-cloud premium is ₹13.5 lakh/month — that funds an entire SRE + DBA team plus tooling. Self-hosted is the obvious choice past ~50 agents if you have any engineering bench.

When cloud wins

• Team < 8 agents. Operational overhead amortises poorly at this scale.
• No engineering bench. Hiring a telephony-aware SRE in India is slow and expensive (₹15-25 lakh CTC for someone competent on Asterisk + Linux + SIP).
• You need to be live in a week. Self-hosted timelines are 3-4 weeks minimum.
• Variable / seasonal load. Cloud lets you scale up an Election-season campaign and scale back without server provisioning.
• Geographic spread. If you have agents in 5 countries, dealing with regional SIP trunks + multiple data centres is operationally heavy. Cloud abstracts that.

When self-hosted wins

• Team > 50 agents. The cost savings dominate the operational overhead.
• Compliance requires data residency. DPDP Act, RBI for BFSI, IRDAI for insurance, sectoral regulators for healthcare — all push call audio onto your infrastructure. Self-hosting on an Indian cloud region or your DC is the cleanest answer.
• You need custom call routing. Complex IVRs, language detection, regional routing, dynamic outbound trunk selection. Self-hosted Asterisk dialplans handle anything; cloud platforms work within their API model.
• You already have an SRE team. The marginal cost of one more Linux service is small if you already operate Postgres + a web app at scale.
• Multi-tenant. You are running contact-centre-as-a-service for multiple clients with isolated workloads. Self-hosted gives you the isolation primitives.
• Long contracts with bundled-minute providers feel expensive. The break-even is usually obvious at ~30 agents.

When hybrid wins

• You need data residency (audio on your infra) but want fast onboarding (admin UI managed)
• Mid-market growth phase where you do not yet have the SRE capacity to run a full self-hosted stack but you cannot stay on a marked-up cloud platform forever
• BPOs operating for end clients with mixed compliance needs — hybrid lets you isolate one client's media to a dedicated environment while keeping the rest on the shared cloud

Operational realities you should know

Self-hosted operational tax

• Linux patching (kernel + Asterisk security advisories) every 4-6 weeks
• Postgres tuning + replication setup (one primary, one standby is standard)
• SIP trunk monitoring (one-way audio, codec mismatches, jitter)
• Backup + restore drills (quarterly)
• Capacity planning for peak hours
• On-call rotation (telephony fails differently than web — outages are loud)

Total: roughly 20-40 hours/month of skilled SRE time for a 50-agent setup once it is steady-state.

Cloud operational tax

• Vendor escalation when something breaks (you do not have root access to debug)
• Per-incident dependency on vendor support response times (usually 4-24 hours)
• Lock-in to the vendor's data formats; portability is real but always painful

The trade is "your time vs vendor's time", and the right answer depends on which is cheaper for your specific team.

Migration path — start cloud, move to self-hosted

A pattern we see often: start on cloud (managed) for the first 6-18 months while building product / market fit. Then migrate to self-hosted (or hybrid) once the agent count crosses ~30 and the per-minute economics become impossible to ignore.

If your dialer vendor offers both modes from the same codebase (dialque does), the migration is a configuration change, not a re-platform. If your vendor is cloud-only, the migration is a full vendor switch — 6-10 weeks of phased rollout, DID porting, agent retraining.

This is the strongest argument for picking a vendor that supports both modes from day one, even if you start cloud. The optionality is worth a small premium.

Compliance + data residency

Cloud dialers route your call audio through the vendor's infrastructure. For Indian customers in regulated sectors:

• RBI's "Storage of Payment System Data" directive requires payment-related call audio in India only. Many cloud vendors are headquartered in the US or use US regions; verify their India region story before signing.
• DPDP Act 2023 allows cross-border transfer unless on the government's restricted list, but the cleaner path is "all personal data stays in India".
• Sector regulators (IRDAI, SEBI, TRAI, sometimes MoHFW for healthcare) impose their own residency rules.

Self-hosted in an Indian cloud region (AWS Mumbai, GCP Delhi NCR, Azure South India) or on-prem in India trivially satisfies these. Hybrid does too if the media servers are in India.

Frequently asked questions

Can I start cloud and migrate to self-hosted later?

Yes if your vendor supports both modes from the same codebase — most do not, so verify before signing. If your vendor is cloud-only, the migration is a full vendor switch.

What is the smallest team where self-hosted makes sense?

~30 agents if you have engineering bench. ~50 agents if you are hiring an SRE specifically for this. Below that, cloud-passthrough beats self-hosted.

Does self-hosted mean on-prem?

No. Self-hosted just means you run the software. AWS Mumbai is the most common deployment target; on-prem datacentres are less common and only used by very compliance-bound customers.

What about pure on-prem with no cloud touchpoint?

Possible but expensive at small scale. Requires DC space, redundant power, redundant networking, on-call hardware support. Rare outside of large enterprises with existing DC operations.

Does cloud guarantee higher availability?

No. Both models can hit 99.9-99.99% uptime; the difference is who is responsible for hitting it. Cloud vendors publish SLAs (usually 99.9%); self-hosted SLAs are whatever your team commits to.

dialque ships the same software in all three modes. Start cloud-managed if you need to be live in a week. Move to self-hosted or hybrid when your team scale or compliance posture warrants it. The codebase is the same — only the operations boundary changes.