Short answer: An automated NPS call costs roughly 38 paise a minute of carriage on a published Indian domestic rate, billed in 30-second pulses — so a 35-second call bills a full minute. The number that matters is cost per returned score, not per dial. At a 10% response rate that is ₹2.09 a score, against ₹10.90 by WhatsApp marketing template.
Net Promoter Score by phone is one of the few survey jobs where an automated call is structurally cheaper than a message, and the reason is not the headline rate: Meta’s own rules will not let a generic feedback survey be priced as a utility template. We sell voice and messaging campaigns, so treat the comparisons below as interested.
What the carriage actually costs
| Provider | Route | Published rate | Billing increment |
|---|---|---|---|
| Plivo | India domestic outbound voice | ₹0.38 per minute | “Pulse is 30 seconds” |
| Plivo | Browser SDK / WebRTC | ₹0.25 per minute | As above |
| Twilio | India mobile, outbound | $0.0496 per minute | Not stated on the page |
At the European Central Bank reference rate of ₹95.28 to the dollar, dated 13 October 2026, Twilio’s $0.0496 is ₹4.7259 a minute — 12.44 times Plivo’s published domestic rate. We do not assert an explanation; the two pages describe their routes differently, and a buyer should ask in writing which route their traffic takes. Our work on bulk voice call pricing shows why that question decides the bill.
The pulse is the first place your price moves
A 30-second pulse means you buy time in half-minute blocks, rounded up. An NPS call is short by design — greeting, one question, one keypress, thank-you — and short calls are where rounding bites hardest.
| What happens on the dial | Connected audio | Pulses billed | Carriage at ₹0.38/min |
|---|---|---|---|
| Respondent presses a key | 35 seconds | 2 | ₹0.38 |
| Respondent hangs up on the question | 15 seconds | 1 | ₹0.19 |
| Nobody answers | 0 seconds | 0 assumed | ₹0.00 |
Those durations are our illustrations, and the last row is an assumption — we found no published statement of how unanswered attempts are billed, so confirm it with your provider. What is published is the pulse, and it converts a 35-second call into a 60-second invoice. Shaving five seconds off a 40-second call does nothing; shaving six off a 36-second call halves the bill.
WhatsApp will not price a generic NPS survey as utility
This is the finding that decides the channel, and it is Meta’s own wording. On template categorisation, read 14 October 2026:
“Specificity of the order or interaction to which these relate is necessary.”
“A general/generic survey or request for feedback will not be approved as utility.”
An NPS question — how likely are you to recommend us — is tied to no order number and no particular interaction. On Meta’s published rules it therefore cannot be a utility template, and is priced as marketing. On the only Indian card we could verify on 14 October 2026, AiSensy’s, effective 1 October 2026:
| Category | India rate per delivered message |
|---|---|
| Marketing | ₹1.09 |
| Utility | ₹0.145 |
| Authentication | ₹0.145 |
| Service | ₹0.145, after 1,000 free per number per month |
Marketing is about 7.52 times utility. That multiple is the whole argument: “how was order 4471?” can plausibly sit at ₹0.145, while a brand-level NPS question sits at ₹1.09 for the same keystroke from the same customer. The rules are set out in WhatsApp utility vs marketing templates.
Cost per returned score, not cost per dial
Neither a dial nor a delivered template is an outcome. The outcome is a returned score:
Cost per returned score = total campaign cost ÷ scores actually returned
For WhatsApp that is simple, because Meta charges on delivery: cost per returned score is the list rate divided by your reply rate. For IVR it is better than it looks, because dials that produce no score are mostly cheaper than those that do.
| Share returning a score | IVR, carriage only | IVR from 17 Nov | WhatsApp marketing | WhatsApp utility, if it qualified |
|---|---|---|---|---|
| 42.9% | ₹0.6329 | ₹0.7494 | ₹2.5408 | ₹0.338 |
| 30% | ₹0.8233 | ₹0.99 | ₹3.6333 | ₹0.4833 |
| 20% | ₹1.14 | ₹1.39 | ₹5.45 | ₹0.725 |
| 10% | ₹2.09 | ₹2.59 | ₹10.90 | ₹1.45 |
| 5% | ₹3.99 | ₹4.99 | ₹21.80 | ₹2.90 |
Across the range, a WhatsApp NPS round at the marketing rate costs four to five and a half times an IVR round on published carriage — and the ratio worsens as response rates fall, because the message is charged whether or not anyone replies.
Worked example: 20,000 customers, one round
Carriage only, no platform fee, no GST. At a 10% return rate, giving 2,000 scores:
- IVR on published domestic carriage: ₹4,180.00, or ₹2.09 a score
- The same round after 17 November, if the regulation 35A charge reached the invoice: ₹5,180.00, or ₹2.59
- WhatsApp marketing template: ₹21,800.00, or ₹10.90 a score — 5.22 times the IVR round
- WhatsApp utility, if a transaction-specific version qualified: ₹2,900.00, or ₹1.45
At the one published completion figure cited below, 42.9%, giving 8,580 scores: IVR ₹5,430.20 (₹0.6329 a score) against WhatsApp marketing ₹21,800.00 (₹2.5408), a multiple of 4.01.
Add the 18% GST that providers’ cards apply and ₹4,180.00 becomes ₹4,932.40, ₹21,800.00 becomes ₹25,724.00. We have no dated primary citation for that 18%, so we describe it as the rate providers apply rather than asserting it.
From 17 November your number series becomes a price
TRAI’s Third Amendment to the TCCCPR, gazetted 18 September 2026, inserts regulation 35A. Verbatim:
“The Terminating Access Provider (TAP) may charge the Originating Access Provider (OAP) upto Rs. 0.05 (five paisa only) per minute on per minute pulse basis for A2P calls”
with a proviso that there shall be no such charge on
“any A2P calls made using number resources assigned from 140xx, 1600xx, 1601xx or any other series designated by the Authority for Commercial Communications from time to time”
Regulation 18, which inserts 35A, sits in the sixty-day commencement bucket. Counted from the gazette, that is 17 November 2026.
On the 20,000-dial example at a 10% return rate, the ceiling adds ₹1,000.00 a month — every call of up to 60 seconds counts as one chargeable minute, so that is 23.92% on top of carriage, or ₹12,000.00 a year — and nil from a designated series. We are not claiming it will reach your invoice: it is a ceiling one operator may levy on another, and no aggregator publishes a pass-through.
The second reason the series matters is not a price. TRAI’s press release of 18 September 2026 says the special series including 140xx, 1600xx and 1601xx “shall not be flagged as suspected spam for recipients”, and that Call Management Applications are “prohibited from blanket blocking, filtering, or spam tagging calls from number series designated”. A call that cannot be spam-tagged is a call that can be answered — delivery reality arriving as a regulation rather than a rate. Whether your NPS call is a service call on an existing relationship, or promotional and so on the DND-blockable 140xx series, turns on definitions we read but will not rule on; see the 140 and 1600 series explained. This is a reading of published regulations, not legal advice.
What the published evidence does and does not say
We publish no response rate of our own. What exists in the dated, peer-reviewed literature:
- JMIR Formative Research, 2025 — an IVR patient-satisfaction survey in Addis Ababa, fielded June–July 2024. Of 2,415 enrolled patients, 2,236 (92.59%) picked up, 1,584 (65.59%) answered at least one question, and 1,037 (42.94%) completed the full survey.
- Chakraborty et al., DEV’13, India, data collected 2012 — IVR against a live operator: “IVR incurs a 4.0% error rate (95% C.I. 2.5% – 6.1%)”, and 6 of 26 people with the opportunity to take the survey, 23.08%, were lost to interface problems.
The 42.9% carries a caveat that matters more than the number: those patients were approached in person, screened and consented before any call was placed. That is a recruited cohort, not a cold customer list. Do not budget a cold NPS round at 42.9% — it is the top of a range, not the middle. Use the formula with your own historical rate; if you have none, run a few hundred dials and measure first. Our comparison of IVR, CATI and WhatsApp surveys applies the same reasoning across three methods.
The layers on an NPS campaign bill
| Layer | On an NPS round | Published today? |
|---|---|---|
| 1 — Platform list rate | ₹0.38/min carriage, or ₹1.09 a marketing template | Yes |
| 2 — Intermediary markup | Whatever your aggregator adds on top | Rarely |
| 3 — Platform and seat fees | Monthly plan, per-agent charges, recording storage | Usually |
| 4 — Compliance and tax | DLT registration, 18% GST, regulation 35A from 17 Nov | Partly |
| 5 — Delivery reality | Pulse rounding, spam-tagging, DND, who presses a key | Never |
Layer five moves the cost per returned score by multiples rather than percentages, and it is the only layer that never reaches an invoice. The full frame is in our WhatsApp pricing pillar and in SMS vs WhatsApp vs voice.
Frequently asked questions
Is an NPS call promotional or a service call? It turns on whether you have an existing relationship and matching consent. The Third Amendment revised the definition of explicit consent and kept “inquiry” within the definition of relationship, limited to seven days from that inquiry. Take the classification to your access provider in writing: it decides both your number series and your DND exposure.
Can I send the NPS question as a utility template instead? Not a generic one. Meta’s page says a general or generic feedback request “will not be approved as utility”. A request tied to a specific order may qualify — a decision about what you ask, not a billing trick.
Does the 30-second pulse apply everywhere? It is published on Plivo’s India voice page. Twilio’s India voice page states none. Ask for the increment in writing — on a 35-second call it is worth as much as the rate.
What does regulation 35A do to my budget? On the example above, 23.92% on top of carriage from 17 November 2026 — or nothing from a 140xx, 1600xx or 1601xx number. Size it as a risk rather than booking it as a cost.
What response rate should I plan for? Not one we will give you. The only dated, peer-reviewed figure here, 42.9%, comes from a consented, in-person-enrolled cohort and is not a cold-call rate. Measure a few hundred dials of your own list.
Is IVR always cheaper than WhatsApp for feedback? No. If your request is specific enough to be approved as utility, WhatsApp at ₹0.145 undercuts IVR carriage at every rate in the table. The generic NPS question is where IVR wins, precisely because it cannot be a utility template.
Where these numbers come from
Every rate, fee and regulation here was verified against a live primary source on 14 October 2026, attributed to the company or authority that publishes it. We publish no engagement statistics of our own, and where we could not verify something we say so rather than estimating.
What we could not verify. A dated CBIC citation for the 18% GST on telecom services. How unanswered call attempts are billed by any Indian provider. Whether any aggregator passes the regulation 35A charge through; none publishes it. Any permitted calling-hours window for commercial voice calls — the gazetted amendment sets out customer time-band preferences in Schedule-II but no sender-side window, and we found no primary TRAI source for the widely repeated 9am–9pm figure, so it does not appear above. Plivo’s India domestic SMS rate, which that page marked “Not Supported” on 14 October 2026.
Sources. Plivo India voice and India SMS pricing pages; Twilio India voice pricing page; AiSensy pricing page (India card, effective 1 October 2026); Meta’s WhatsApp template categorisation guidelines; TRAI, Telecom Commercial Communication Customer Preference (Third Amendment) Regulations, 2026, gazetted 18 September 2026; TRAI Press Release No. 119 of 2026; European Central Bank reference rate via the Frankfurter API, rate dated 13 October 2026; JMIR Formative Research 2025 (Addis Ababa IVR study); Chakraborty et al., DEV’13. All read 14 October 2026.
To see this arithmetic run against your own list and your provider’s card, our pricing page explains how we work, and our AiSensy comparison shows the method applied to a published card.