Personalize with Liquid and Adaptyle

Use Liquid for facts that must stay exact, and Adaptyle for wording that benefits from customer context. For a new coffee announcement, Liquid can insert the customer’s name and product link while Adaptyle writes a short introduction based on recorded preferences.

Start with your first personalized email for a visual walkthrough. This reference explains how to choose the right layer and check its output.

The two stages of personalization

Liquid resolves variables and conditions first. Adaptyle can then generate or modify marked passages using that resolved content and the available data. A prompt-free email does not need the Adaptyle generation pass.

Writing an email with maxinja is a separate activity: the assistant helps author the reusable message. Adaptyle adapts marked content when recipient messages are rendered.

Liquid variables

Open Data and read this email’s property names. If the audience property is customer, start with:

Hi {{ customer.firstName | default: "there" }},

With firstName set to Maya, the result is “Hi Maya,”. With no first name, it is “Hi there,”. A different property name needs a different expression; contact.firstName does not work unless the email exposes contact.

Use Liquid for names, totals, URLs and conditions. Select required Evaluated fields in Data. On-demand fields resolve their lookup; scheduled fields read a complete publication without recalculating because you selected them. An enabled calculation with no complete or expired publication blocks expansion; a valid publication with no result for this customer can return null. An ordinary missing field can render empty, so define a fallback and inspect sample data. A Liquid fallback does not repair an unavailable evaluation.

Let maxinja finish the expression

Dynamic Composer and Plain Composer offer AI autocomplete. Accept a suggestion with Tab or dismiss it with Escape, then check the property it references. Suggestions use credits. An Admin controls Settings → maxinja → Editor autocomplete for the space; code editors do not provide this autocomplete.

Adaptyle prompts

Use Adaptyle when the best wording depends on the recipient. Keep offer terms and essential instructions in fixed copy or Liquid outside the passage. Select the source data the passage may use, and define what to say when it is missing.

Generation prompts

A generation prompt asks for new copy:

<<<Write one warm sentence introducing our new roast. Use a recorded coffee preference if available; otherwise use a general introduction. Do not invent a purchase, discount or product claim.>>>

The generated passage replaces the prompt in the personalized result. Check that its claims come from the selected data; a clear instruction is not proof that every result is correct.

Modification prompts

A modification prompt supplies starting copy after three colons:

<<<Make this warmer and shorter. Preserve its meaning and add no new facts.:::Thank you for shopping with us. We hope you enjoy your coffee.>>>

Use this for a controlled rewrite of a passage you have already approved. Review the result alongside the original and the customer data.

Branding is injected automatically

Adaptyle receives the selected published brand context. Its tone and writing guidance help shape the result. Review the chosen brand kit and representative outputs; selecting a brand does not guarantee every passage follows every instruction.

When to use which

Before sending, compare a customer with useful history, one with sparse data and one outside the usual pattern. Check facts, links, tone and fallback text, not only whether the paragraph sounds natural.

How rendering and cost work

Open Preview. Base renders Liquid and layout without Adaptyle generation. Under Personalized, Generate preview produces a personalized sample when prompts are present and can use credits. Send test in the toolbar sends the rendered subject and body already available; it does not generate another variation for the test address.

Live rendering uses the recipient’s real data. Generated wording can differ from your sample. AI usage, quality-review usage and email delivery have different billing effects; inspect Personalize → Estimate in the composer and the credits guide.

For broadcasts, previewing samples does not approve every generated message. If preparation produces a hold, inspect the exact message and evidence in Quality review.

Let maxinja help

Tell maxinja the customer outcome, the available facts and the passage you want adapted. For example: “Draft a new-roast announcement. Keep the product link fixed, use the customer’s first name with a fallback, and adapt only the introduction from recorded coffee preferences.” Review the saved draft and sample outputs before scheduling.

Frequently asked questions

Why does a fallback not fix an unavailable Evaluated field?

A fallback handles a missing value in valid rendering data. An enabled scheduled field with no complete or expired publication cannot be expanded for that decision. Inspect its status, Published value and Data through, then resolve that evaluation before rendering.

Does editing with maxinja personalize every future email?

A maxinja edit changes the saved draft. An Adaptyle instruction stays in that draft and uses the email’s configured runtime model when each recipient message is prepared. Keep the two actions separate when reviewing cost and saved content.

Can a runtime instruction change the button or fetch more history?

The renderer uses the declared data and selected expansions. Its checks protect links, attributes and surrounding fixed content. Supply the required customer context and make structural changes while authoring the draft, then generate another sample.