We built it on Snowflake with full data ownership. 16.4 million unified customer profiles, RFM scoring, churn prediction, and discount suppression that identified £8.3M in margin protection. We still run it.
CDP vendors charge premium prices for generic segmentation. They give you personas, not profit. Most businesses are paying £30-100K per year for a platform that can't answer the most important question: "which customers should NOT get a discount?"
When we audited the existing CDP at a major UK e-commerce retailer — processing 2 million daily requests across 16.4 million customer profiles — we found a platform that could segment customers by demographics and purchase history but couldn't classify price sensitivity, predict churn at the individual level, or identify which customers were being over-discounted. The vendor charged £45K per year for this capability gap.
The data existed. It was scattered across six platforms — e-commerce, email, SMS, support, analytics, and payment — but no single vendor could unify all of it into a view that answered margin questions rather than marketing questions. So we built one.
We don't default to building. But when vendor costs exceed £30K per year and the business has specific margin questions no off-the-shelf platform can answer, building is the right call. Here's what we did.
We found data across 6 platforms — e-commerce, email, SMS, support, analytics, payment. Most CDPs can't unify all of these without costly custom connectors. We built the pipelines ourselves.
Which customers buy at full price? Which ones only convert with discounts? Which are about to churn? These are margin questions, not marketing questions. They drove every design decision.
RFM scoring (recency, frequency, monetary), churn prediction, price sensitivity classification. All built on Snowflake with dbt transformations — no vendor lock-in, full data ownership.
We push segments directly to email, SMS, and ad platforms. Four discount tiers: suppress (full price), light, standard, deep. The suppress tier is where the money is.
By analysing purchase patterns across 16.4 million profiles, we identified 90,000 customers who consistently buy at full price regardless of discount availability. These customers were receiving the same promotional emails as everyone else — the business was giving away margin unnecessarily.
We calculated £8.3M in potential margin protection by suppressing discounts to this segment alone. Even the conservative estimate — that 17.6% of these customers were using coupons they didn't need — represented £727K per quarter in leakage, or £2.9M annually.
This is the single highest-ROI use case in customer data. Not personalisation. Not journey mapping. Simply stopping the practice of giving discounts to people who would pay full price. Our pricing work operates at the product level, while discount suppression works at the customer level — together they protect margin from both sides.
Not every customer should get the same offer. We built a four-tier classification that determines discount strategy at the individual level.
| Tier | Customer Behaviour | Action | Est. Population |
|---|---|---|---|
| Suppress | Buys at full price regardless | No discount codes, no promotional pricing | ~90,000 |
| Light | Converts with minimal incentive | 5-10% discount, loyalty rewards only | ~180,000 |
| Standard | Needs moderate incentive | Standard promotional calendar | ~2.1M active |
| Deep | Only converts with heavy discount | Aggressive markdown, clearance first access | ~850,000 |
We built and deployed the CDP in under 8 weeks. Here is what it delivered.
The CDP runs on Snowflake as the data warehouse with dbt Cloud managing all transformations. CDC (Change Data Capture) pipelines replicate data in near real-time from the e-commerce platform, so the customer view is always current.
RFM scores are recalculated daily. Churn prediction models run weekly against the full customer base, flagging at-risk customers before they lapse. Price sensitivity classification uses 12 months of transaction history, discount code usage, and session behaviour to assign each customer to one of the four discount tiers.
Segments are pushed directly to the marketing activation layer — email and SMS platforms receive updated lists automatically, so campaigns always target the right tier with the right offer. No manual CSV exports. No stale data. We maintain and optimise the whole system ongoing.
Build when your margin questions are specific to your business, when you need full data ownership, when vendor costs exceed £30K per year, or when your data spans more than 3-4 platforms that no single vendor integrates well. Buy when you need speed-to-market and your data is already consolidated.
In this engagement, the data lived across six platforms with non-standard schemas. No off-the-shelf CDP could unify all of it without expensive custom work — at which point you're paying vendor pricing for a bespoke build anyway.
Discount suppression identifies customers who buy at full price and removes them from promotional campaigns. Most businesses send the same discount to everyone — meaning they're giving margin away to customers who would have paid full price anyway.
This is the single highest-ROI CDP use case we've deployed. It doesn't require complex AI or months of development. It requires a clean customer view, transaction history analysis, and the discipline to treat different customers differently.
RFM scores each customer on three dimensions: Recency (when they last purchased), Frequency (how often they purchase), and Monetary (how much they spend). Combining these creates segments like "high-value loyal" (recent, frequent, high spend) vs "lapsed VIP" (not recent, was frequent, high historical spend).
Each segment gets different marketing treatment. A lapsed VIP might get a personal re-engagement offer, while a high-value loyal customer gets early access to new products — not discounts.
At minimum: a data warehouse (Snowflake, BigQuery, or Redshift), a transformation layer (dbt), CDC pipelines for real-time data sync, and API connections to your activation platforms (email, SMS, ads).
Total infrastructure cost is typically £500-2,000 per month — a fraction of vendor CDP pricing. The real cost is the build itself, which is a one-time investment. After deployment, we maintain the platform with minimal overhead and zero licence fees.
A production CDP with RFM scoring, churn prediction, and discount suppression can be built and deployed in 6-8 weeks. The first 2 weeks are data mapping and pipeline setup, weeks 3-4 build the intelligence layer, and weeks 5-8 are activation, testing, and optimisation.
The timeline depends on data complexity. If your data is already in a warehouse, we can cut 2 weeks. If you have 10+ source systems with non-standard schemas, add a week for mapping.
Ours identified £8.3M in margin protection for one client. If you're spending £30K+ per year on a CDP that can't tell you which customers to stop discounting, let's talk.