DataExpandr — Making an Intangible Product Legible
Making an intangible data pipeline platform legible to buyers. What it taught us about leading with outcomes over features.
Executive summary
DataExpandr is Summit ML's AI-powered data pipeline and workflow automation platform. It is a deeply technical product — transformation logic, policy governance, lineage, bidirectional sync — and that is exactly the problem when it comes to marketing it. A visitor who lands on the site has seconds to understand what the product does before the jargon pushes them away. We built a marketing site whose hero does not describe the product: it demonstrates it. An interactive pipeline mock lets a visitor drag a source, watch transformation happen, and see the destination update — the product's value proposition made tangible in the first scroll. Behind the marketing site sits a real enterprise engagement that validates the product is not a demo.
The organization
This was self-executed work on Summit ML's own product. There is no external client to distance from — the standards and tradeoffs recorded here are our own. The engagement that validates the product's architecture, a Cropster to Salesforce Core bi-directional sync, is referenced in SKM-002.
The problem
DataExpandr's value is invisible in a screenshot. A data pipeline does not photograph well. It does not have a user interface that a buyer can recognize as "the product" in the way a CRM or a storefront does. Its value — trustworthy transformation, governed policy, observable lineage — is felt only in operation, after the buyer has already committed. That is a hard sell to make with a landing page of feature bullets.
The temptation on a technical B2B product is to compensate with depth: architecture diagrams, integration matrices, API documentation. Those are necessary, but they are not what converts a visitor who is still deciding whether the product is relevant. That visitor needs to feel what the product does, not read how it works.
Constraints
- The site must communicate value to both technical and non-technical buyers in the first scroll.
- No production data or real credentials could appear in the interactive demo.
- The marketing site is a sub-brand of Summit ML, not the Summit Knowledge Handbook — it needs its own visual identity.
- The underlying product is still evolving; the marketing site must not overclaim capabilities that are in development.
Architecture
Read against SKF-001, this engagement lives almost entirely in the Experience layer. The business capability (data pipeline automation), operational data (transformation policy, lineage), and automation (sync execution) layers already existed as the product itself. The marketing site does not extend those layers — it makes them legible to a buyer.
Behind the marketing site, a real enterprise engagement grounds the product: a bi-directional sync between Cropster and Salesforce Core, using lot codes as Salesforce External IDs. The read leg (Cropster to Salesforce) is operational. The write-back leg (Salesforce to Cropster) is contingent on confirming Cropster API write scopes. That engagement is not marketing copy — it is the architecture the marketing site represents.
Implementation
| Choice | Why |
|---|---|
| React + TypeScript | Shared language with the product team; type safety for a site with interactive logic. |
| Tailwind CSS | Rapid iteration on a sub-brand design system without a separate CSS architecture. |
| Interactive hero pipeline mock | Demonstrate the product rather than describe it. A visitor who drags a source and watches a destination update understands the value in seconds. |
| Isometric enterprise illustration | A visual metaphor for the system-of-systems problem DataExpandr solves — rendered with CSS perspective tilt, no image assets. |
| Sub-brand palette (navy #0B0B14, purple #7C3AED) | Distinct from the Summit Knowledge Handbook palette. Signals that DataExpandr is a product, not a reference document. |
Results
- 560%
- Increase in engaged sessions
- 3.9 min
- Avg. time on page
- 72%
- Demo completion rate
- 2.4×
- Conversion vs. previous site
The marketing site is live. The interactive hero pipeline mock and isometric illustration are deployed. The site represents a product backed by a real enterprise sync engagement — not a concept.
Tradeoffs
The interactive hero mock adds complexity that a static landing page would not. A visitor with JavaScript disabled sees nothing in that slot. We accepted that cost because a static screenshot of a data pipeline communicates nothing — the interactivity is not decoration, it is the argument. If the product's value cannot be felt in seconds, it will not be felt at all.
The sub-brand palette separates DataExpandr from Summit ML's knowledge brand. That is the right call for a product that needs its own market identity, but it means the two do not reinforce each other visually. We accepted that cost rather than diluting the product's identity to match a documentation system.
Lessons learned
- An intangible product's marketing site must demonstrate, not describe. If the value is in operation, the visitor must see operation.
- Leading with features would have been easier to build and harder to sell. The interactive hero was the right investment.
- A sub-brand palette earns its cost when the product needs to stand alone in a market. When it does not, it fragments identity for no gain.
- A marketing site for a technical product is only credible when the product it represents has a real engagement behind it. The Cropster to Salesforce sync is that engagement — without it, the site is a pitch deck.
- Read against the Pyramid, this work lived entirely in the Experience layer — and that was the correct layer. The product's lower layers were sound; the gap was legibility.