Udemy
    •  
    •  
    •  
    •  
    •  
    •  
    •  
    •  
Turn what you know into an opportunity and reach millions around the world.
Learn More
Your cart is empty.
Keep shopping
Yespo Certification: Personalization & AI in a CDP
New
6 students

Yespo Certification: Personalization & AI in a CDP

Master Velocity, promo codes, and AI recommendations, then automate work itself with AI agents.
New
Created byYespo Team
Last updated 8/2026
English

What you'll learn

  • Understand the four levels of personalization depth and why depth is set by data, not by which tool you use
  • Master Velocity, merge tags, and Display Conditions to build content that adapts to each recipient
  • Use the preprocessor to send unique, file-driven content to every subscriber in a single mass campaign
  • Choose the right delivery method among seven ways to issue promo codes, from shared codes to real-time API
  • Understand how transformer-based AI recommendations work and how to implement them on-site and in campaigns
  • Learn how predictive segmentation and AI agents shift work from rule-based guessing to automated, data-driven decisions

Course content

4 sections • 9 lectures • 1h 25m total length

  • Four Levels of Depth7:45

    Personalization has four levels of depth. Attribute-based — using static fields (name, city, gender). Behavioral — based on actions (viewed, abandoned cart). Interest-segment — based on accumulated interest (preferred category). AI — a model that learns from all behavior.

    Depth is determined by data, not by the tool. Each next level requires more and deeper data: static attributes → event stream → accumulated history → the full behavioral dataset. The best tool is powerless if there is nothing to learn from.

    Behavioral personalization is impossible without web tracking. The most profitable triggers — cart, browse, price drop — are built on behavioral events. No behavioral data collection means staying at level one.

    AI works at scale, not just in triggers. In a single mass campaign to a million people, each person sees a block tailored to their own history. The transformer model understands context and sequence, not just “bought together.” Liki24 case: approximately +70% conversion.

    AI is controlled. Industry rules (such as the prohibition on recommending prescription drugs) are applied to the model without loss of effectiveness — it works within your expertise, not instead of it.

    Personalization starts with data collection, not with tools. Start with an audit: what do you already know about the customer. Then choose a depth that matches what you have.

  • Variables and Velocity: From a Name to Logic7:25

    All content personalization runs on one tool — Velocity — from a simple name to complex logic. A merge tag is already Velocity in its simplest form.

    The syntax for an empty field is critical. A variable without an exclamation mark shows the literal variable text and is only suitable for testing; with an exclamation mark, an empty field renders as nothing; with a default value, it substitutes fallback text. For live campaigns — use the exclamation mark or a default value.

    Field names in merge tags are case-insensitive. An event value takes priority over a contact record value.

    Velocity offers three capabilities beyond simple substitution. A condition (different content based on data), a loop (repeat a block for a list of products), and a custom variable (calculate something inside the email). This is the same if-else logic as workflow branching — just operating inside the email.

    Display Conditions is that same conditional logic without code. Show different people different blocks in one email using a toggle in the editor.

    Case sensitivity in Display Conditions is a trap. For contact fields, the name is case-insensitive but the value must match exactly; for event parameters and external sources, the name itself is also case-sensitive. If there is a mismatch, the block does not appear — so always test.

    Complex cases go beyond the visual editor. Nested conditions, else branches, and loops are written in HTML as full Velocity.

  • The Preprocessor: Unique Content in a Mass Campaign7:11

    The preprocessor closes a specific gap. A mass campaign with unique content for each subscriber, when the data is neither in the contact record nor in an event, and building a live integration is not on the table.

    The mechanics are simple. You prepare a data file in advance, upload it, and the system inserts each person's portion at send time.

    Two types solve different problems. Key matches data to the recipient by email address — this is one-to-one personalization. Random distributes random selections from a shared pool — this is even catalog coverage when products are many and similar.

    The file is prepared in JSON format. Uploaded in the account settings, and archives (ZIP and TAR.GZ) up to 200 MB are also accepted. Files have a limited lifespan — this is a campaign tool, not a data store.

    Rendering the data in the email uses the same Velocity. The preprocessor only brings the data; the loops and variables work exactly as you have already learned.

    Testing is mandatory. Use the preview with test JSON substitution, and always check three cases: a full dataset, an incomplete one, and a contact who is not in the file at all.

    The data in the file is static as of the time of upload. Generate the file as close to the send time as possible; for rapidly changing data, look for a different tool.

  • Promo Codes: Seven Ways to Give a Customer a Discount7:50

    Promo codes are not a separate tool — they are an application of what you have already learned. Variables, the preprocessor, and workflow blocks. The complexity comes from choosing the right method, not from the substitution itself.

    The main fork is one shared code versus unique codes for each person. A shared code is simple but can spread; a unique code is controlled but requires generation and distribution. Almost all the complexity comes from the unique side.

    In mass campaigns, there are three paths. A code from the contact field (pre-assigned), a code from the promo code database (a pool with no pre-assignment), or the preprocessor (via file, for email).

    In triggered workflows, a Task block handles it. Retrieve a code from a pre-prepared pool, which guarantees unique consumption; or generate a code on the fly with the discount parameters encoded into it.

    Via API is used when the code is generated by your external loyalty system. That system remains the owner; the sending platform only delivers.

    Gamification is a composition of already-familiar parts. The game generates the code, the workflow delivers it immediately, and a timer plus a usage check manage the reminder. A won code is valued more than a gifted one.

    Choosing a method comes down to two questions: is the code shared or unique, and where does it come from? The answers almost always determine the mechanism.

Requirements

  • No prior technical background is required, though completing the earlier modules will help, since this module builds directly on data, workflows, and channel concepts already covered.

Description

This course contains the use of artificial intelligence.

"Dear [Name]" is the entry point to personalization, not its ceiling. This module covers everything above it: from a simple merge tag to a model that decides, on its own, who should get an email at all.

You'll start with the core principle that depth is set by data, not by tools, before mastering Velocity — the single technology behind every level of personalization, from a name to full conditional logic and product loops. From there, you'll cover two tools for edge cases standard variables can't reach: the preprocessor, which delivers unique file-driven content to every subscriber in a mass send, and the seven ways to issue a promo code, from a shared discount code to a workflow-generated code with the terms encoded inside it.

The module then moves into AI recommendations — how transformer models evolved past "bought together" logic to understand context and sequence, how they solve the cold-start problem for new products and new visitors, and how to actually implement them: web tracking, product feed, data sources, and page-by-page placement strategy. You'll see how the account dashboard quietly outperforms every other placement on the site.

From there, predictive segmentation flips the question from "what did this person do" to "what will they do next" — with real campaign economics showing the difference between a guessed segment and a modeled one. The module closes with two very different AI agents: one built into the workflow editor that builds and diagnoses automations from a plain-language prompt, and an external agent that can edit templates, audit deliverability, and run analytics across your entire account, always under your confirmation.

By the end, you'll be able to personalize at every level the platform supports, and know exactly which AI tool fits which job.

Who this course is for:

  • Marketers who already understand CDP data and workflows and want to move from generic content to true 1:1 personalization
  • CRM specialists managing promo code logistics across mass campaigns, triggers, and gamification
  • Marketers evaluating or implementing AI recommendation blocks on-site or in email
  • Anyone curious about using AI agents to speed up template editing, deliverability audits, and reporting

Report abuse