From a feature pile to a platform that onboards itself: rebuilding an AdTech analytics product
B2B SaaS analytics for e-commerce brands
Led design on the rebuild of an analytics product for e-commerce brands: one design system, self-serve onboarding, role-based access and new analytics views.
TL;DR
- Problem: An analytics product for e-commerce brands had grown by adding features. It was messy, hard to scale, and every new customer needed hands-on onboarding.
- What I did: I led design on the rebuild with the CTO, CEO and three VPs. I filtered a long wish-list into a phased plan, built one design system, and designed self-serve onboarding, role-based access and new analytics views.
- Result: 8 brands onboarded themselves in the first month. Weekly support tickets fell from 12-15 to 7-8. The company reported 28% of users moving to more regular use, and brands reported 108% higher ROAS; I do not claim the design caused it. All figures are company-reported.
- Product
- B2B SaaS analytics for e-commerce brands running ads
- Company
- An AdTech platform (Germany, remote)
- My role
- Senior Product Designer, later Product Design Lead
- Scope
- Product UX, information architecture, design system, onboarding, roles and permissions, analytics views, phased roadmap
- Team in the room
- CTO, CEO, VP People, VP Product, VP Front End, VP Back End
- Roadmap
- 12 milestones, from onboarding to acquisition and profitability
- Status
- Live, features still releasing
- Outcome
- 8 brands onboarded themselves in the first month. Weekly support tickets fell from 12-15 to 7-8. 28% of users moved to more regular use. Brands reported 108% higher ROAS. Company-reported.
Company and partner names are not shown to respect confidentiality; the problem, decisions and results shown here are unchanged. Figures are company-reported.
The problem
The product kept growing by adding features. Screens had been built ad hoc by engineers, pasted in as screenshots, or copied into design files. The result was ten fonts, several design systems and mismatched palettes, and no clear structure. It could not scale, and every new customer needed a person to set it up.
Constraints
Leadership were experts in performance ads and arrived with many requirements.
Engineering and design had drifted apart in the first version.
The product served several kinds of user with different permissions.
Some work for a large ad-platform partner is confidential and not covered here.
Decision 1: a phased plan with a fixed scope
Leadership had many ideas. I opposed adding them all, filtered them, and proposed a phase-by-phase rollout with fixed scope for each phase. Without it, the second version would have repeated the first: mixed designs and a gap between product, design and development.
Decision 2: one foundation instead of ten
I replaced ten fonts and several palettes with one typeface (DM Sans), one brand colour taken from the logo, and one component library of about 37 components, built on Tailwind and shadcn. The second version is built on it.
Decision 3: self-serve onboarding
The new onboarding included a demo, the integrations and the plugins a brand needs. A brand could start, add destinations and stores without a person guiding it.
Decision 4: roles that match how brands work
A brand company can have several brands and each brand several stores, each run by different people. I designed roles including marketer, advertiser, company owner, brand owner, data protection officer and admin. Admins handle billing, so not everyone sees it. Marketers see analytics and edit only the stores they are attached to.
Decision 5: the numbers brands asked for
New analytics views showed new, existing and returning customers by country, demographic and device, and profitability per product and per unit after costs (contribution margins and ad spend), alongside ad spend on the same screen.
Outcome
8 brands onboarded themselves in the first month, with no dedicated customer-support onboarding. Company-reported.
Weekly support tickets fell from 12-15 to 7-8, as reported by support. The sales team moved from onboarding to calls and demos. Company-reported.
28% of users moved to more regular use after the redesign, as reported by the company from backend data. Company-reported.
Brands reported 108% higher ROAS, as calculated by the company from backend data. That is a result for the brands' own ads, and I do not claim the design caused it. Company-reported.