Customer data platform for telecom: ePrivacy and CPNI consent rules, data residency, architecture at tens of millions of subscribers, and the vendor landscape.
Updated
industries
Customer Data Platform for High-Load Commerce and Telecom
By Art Kukharenko · Published · Updated
A customer data platform (CDP) for high-load commerce and telecom has to ingest thousands to tens of thousands of events per second, resolve identities across tens of millions of profiles, and activate to many destinations, without per-identity pricing turning growth into cost. It is usually deployed in the retailer's or operator's own cloud or data centre, sized and load-tested for peaks, with consent and residency rules enforced in the data layer.
This is the entry page for PrivyCDP's material on retail, marketplace and telecom workloads. It links to three practical guides and to the pricing and total-cost tools.
What "high-load" changes
Marketers at smaller companies can treat the data platform as a subscription. At high volume, five things change the answer.
- Cost shape. Meters that count monthly tracked users, profiles or credits grow linearly with audience size. At Twilio Segment's published rate above 100,000 MTUs ($10 per 1,000 per month), 50 million MTUs would list at about $6 million a year; at Salesforce's published $240 per 1,000 profiles per year, 50 million profiles is $12 million 12. These are list-price extrapolations, not quotes.
- Anonymous traffic. Segment counts anonymous visitors as MTUs 1, so the visitor-to-customer ratio drives the bill.
- Peaks. Promotions and launches multiply event rates for hours. The question is not average throughput but p99 latency and backlog drain at a multiple of forecast peak.
- Destination limits. Email, SMS, ad and CRM destinations throttle, and the platform has to isolate them from each other.
- Residency and consent. Telecom operators face traffic-data and location-data rules and national storage conditions.
Guides for high-load teams
Why per-profile pricing breaks at scale
The published meters from Segment, Salesforce and Adobe, cost curves at 1, 10, 50 and 100 million identities with stated assumptions, the anonymous-traffic multiplier, break-even maths against a flat licence, and a negotiation checklist. Read Why Per-Profile Customer Data Platform Pricing Breaks at Scale.
Peak-load activation test plan
A workload model, synthetic event and profile generators, eight scenarios (ingest spike, segmentation lag, activation fan-out, destination rate limits, replay, failure injection, consent withdrawal, soak), pass criteria and tooling. Read Peak-Load Activation Test Plan for a Customer Data Platform.
Customer data platform for telecom
Data classes in telecom, ePrivacy and CPNI consent rules by region, residency and in-country cells, architecture at subscriber scale, number recycling, and the vendor landscape. Read Customer Data Platform for Telecom: Residency, Scale, Consent.
For the full pricing background across vendors, see CDP pricing explained.
Metered or flat: how the models compare
Scroll horizontally to compare all columns ↔
| Dimension | Metered SaaS (per MTU, profile or credit) | Customer-operated platform with flat licence |
|---|---|---|
| Cost shape | Grows with identities, events or credits | Licence by deployment tier; infrastructure and operations scoped separately 3 |
| Peak behaviour | Higher MTU count and throughput limits at peak; credit burn on activation work | Bound by your own capacity, which you can test |
| Time to start | Days to weeks | Weeks to months |
| Operations | Vendor | You or a partner |
| Residency and keys | Vendor-determined, with region options | Your boundary and your keys |
| Best at | Small to mid-scale, no residency constraint, no ops team | Tens of millions of identities, spiky traffic, strict residency |
PrivyCDP is a customer-owned, single-tenant customer data platform with no per-profile or per-event billing, an immutable append-only audit ledger, customer-held keys, and deployment on-premise, in a private cloud in your own AWS, Azure, GCP or OCI account, or hybrid. It is not the right choice when volumes are modest or you have no operations capacity; metered SaaS is often cheaper and simpler at that scale. Put your own volumes into the TCO model before deciding.
Before you shortlist
- Pull your peak-month MTUs or profiles, split into known and anonymous.
- Compute list-price cost at 1x, 3x and 10x that volume for every shortlisted vendor.
- Get the contract definition of the billing unit in writing.
- Define peak targets (events per second, p99 latency, backlog drain) and test them.
- For telecom, separate raw network and billing records from derived attributes allowed into marketing.
- Choose the deployment model with the deployment qualification.
Prices are published vendor figures as read on 17 October 2026 and are not quotes.
Explore this collection
high load commerce
Customer data platform pricing per MTU and per profile: cost curves from 1M to 100M identities using Segment and Salesforce price pages, plus a checklist.
Updated
Load testing a customer data platform for peak traffic: workload model, synthetic generators, eight scenarios, pass criteria and k6, Locust and Kafka tooling.
Updated
Frequently asked questions
What is a customer data platform for high-traffic e-commerce?
It collects web, app and order events, resolves them to customer profiles, applies consent, and activates audiences to marketing and advertising tools. At high traffic the differences are ingestion capacity, anonymous-visitor handling and pricing that does not scale with audience size.
Why does CDP pricing per MTU become expensive?
MTU meters count distinct monthly users including anonymous visitors 1, so cost follows traffic, not revenue. Extended at published rates, 50 million MTUs would list at about $6 million a year, and per-profile pricing grows the same way. Negotiated prices differ, so ask for the contract definition of the billing unit.
Can we load test a customer data platform before Black Friday?
Yes, with synthetic events, open-model arrival rates and mocked destinations. The test plan sets out eight scenarios and example pass criteria. Hosted vendors may restrict load testing in their terms, so check your contract; a platform you run yourself can be sized and tested at full scale.
Does a telecom operator need a different platform?
Often the same software with a different architecture: in-country cells where residency rules require them, raw network and billing records kept out of marketing, and consent keyed to purpose. The telecom guide explains the regional rules, number recycling and the vendor landscape.
When is a metered SaaS customer data platform the better choice?
At modest volumes, with no strict residency rule and no operations team. At 100,000 monthly tracked users, Segment's published Team rate extends to about $13,500 a year, which is small next to the people cost of running a platform. The case for a flat licence appears at the high-load end, with tens of millions of identities.
Sources
- Twilio Segment pricing Open source
- Salesforce Data 360 pricing Open source
- PrivyCDP licensing and TCO pages Open source · Additional source
Put this evaluation into practice.
Assess your deployment requirementsThis article is general information, not legal advice.
Further reading
- Deployment readiness checklist
- CDP Pricing Explained
- CDP for Telecom: Residency and Consent
- Load Testing a CDP: Peak Test Plan
- CDP Pricing per MTU: Why It Breaks
Published: · Updated:
Publication date recorded for the approved release. Editorial updates do not establish legal approval or a customer result.