Capabilities and use cases

SparkleTree is a context-aware marketing platform. This page walks through what it can do, organized by the outcome you’re after, with short examples of each capability in use.

Run campaigns that learn

Every campaign can hold multiple variants — alternative creatives for the same goal. With Adaptive CVG selection (the default), SparkleTree learns which variant performs and shows winners more often. You don’t schedule A/B tests or pick a winner by hand; the campaign keeps a small share of traffic exploring while the rest goes to what’s working. If you prefer predictable rotation instead, switch the campaign to Cycle mode on its Configure face.

Use case: a retail brand launches a spring campaign with three headline variants. Within days, Adaptive CVG is routing most traffic to the variant that actually gets clicks, while still sampling the others in case context shifts. The team checks Dashboard → CVG to see the evidence, not to make the call.

Ads that answer questions

A Squid is your campaign’s living, conversational ad — it greets, talks, and adapts. Instead of a static banner, visitors see a kicker, headline, and body, plus a CTA pill and suggested-question chips that open a real conversation about your products. You edit all of it in place: flip into Shaping mode, click the text, type, and it saves.

Use case: a restaurant’s Squid greets visitors, answers “do you have gluten-free options?” and “are you open Sunday?”, and nudges toward a booking — questions that a static ad would have left to a phone call.

Ads that know where they are

Campaigns can adapt to weather, location, and time. Language adapts too: set a campaign’s language mode to Local language by context and the Squid speaks the visitor’s local language, or pin it to English or a specific language. Context awareness needs no extra configuration per surface — it travels with the campaign.

Build once, deliver everywhere

One campaign can appear across every surface you run — one recipe each in the Deployment cookbook:

• Website embed — one <script> tag (/api/embed/v1/script.js), edge-served.
• Brand page and campaign pages — your published brand home and its campaign pages.
• QR codes — download from Brand → Publishing; scans are tracked and tagged by campaign and surface.
• Physical screens — register Surfaces in Dashboard → Network (kiosks, retail screens, vending, transit and more), group them into Zones, and click Publish to push to the global edge network. A native Android app runs signage and kiosks.
• Meta ads — deploy a campaign to Facebook and Instagram from campaign detail.
• Shopify product pages — an app block renders the winning variant on product pages, with checkout attribution.

Note: The Shopify integration is in beta with limited availability.

Use case: an e-commerce merchant creates one campaign, embeds it on their site, adds the app block to Shopify product pages, and deploys the same campaign to Meta. Analytics reconcile clicks and conversions back to the variant across all three.

Use case: a retail brand with in-store screens registers each screen as a Surface, groups locations into Zones, and assigns the campaign once. Publishing pushes to every screen globally; QR codes on the screens carry shoppers — and the conversation — onto their phones.

Be findable by AI, not just search engines

Your published brand page ships GSO (Generative Search Optimization — SEO’s counterpart for AI assistants) surfaces automatically: llms.txt, machine-readable agent cards, structured data, markdown twins of your pages, and a public /ask concierge that answers questions about your brand. When someone’s AI assistant researches “best pizza near the station,” your brand can answer in its own words.

Generate video without an editor

SparkleTree’s video generation is not a timeline editor. You submit a brief, and an AI authoring agent builds a short on-brand video inside a locked sandbox, with your brand palette injected automatically. Before you ever see the result, a machine judge checks it for legibility, exposure, and duration — so what reaches you is already usable.

Use case: the restaurant asks for a short venue-board video announcing weekend brunch. The result lands ready for its screens and socials, no editing session required.

Get help from Florence

Florence is the AI marketing assistant, available on paid plans. She ships in Observe mode: she watches your socials and campaigns and reports back — she never acts alone. Before Florence may use a variant on socials, you must star it for autonomous use. See Florence for the full picture of her autonomy modes.

In preview: Assist, Operate, and Auto modes appear in the UI but are still in preview.

Let your own AI operate the platform

Connect Claude, ChatGPT, or any MCP-capable agent to your workspace and it can list campaigns, create variants, pull analytics, and render video on your behalf — with the same role and workspace limits you have. See MCP connector setup for the details, and the analytics guide for how to read what all of this produces.