Native mobile · ChargePoint · 2025
Installer App
Connecting the field to the platform: an end-to-end self-serve installation native mobile app.
- Role
- Lead UX/UI Designer
- Duration
- 6 months · 2 iterations
- Domain
- B2B SaaS · Native Mobile
- Client
- ChargePoint



Outcome
A station commissioned in the field now shows up in Polaris ready to activate, with no CX involved. White-glove activation stays available as a chargeable premium service rather than the default, which turns agent cost from overhead into a revenue line.
- 0
- CX touches on the self-serve path
- 1
- activation event per site & org, not per station
- 2
- clean activation routes: enterprise pre-assign + VAR hyperlink
Context
ChargePoint runs one of the world’s largest commercial EV charging networks. Four years ago it moved from direct sales to a Value-Added Reseller model, and today roughly 90% of sales flow through partners like EATON. The activation process never caught up.
Two people live in two separate worlds. The electrician mounts the hardware, scans serial numbers, confirms connectivity, and leaves, all through the Installer App. The org admin never touches hardware; they manage policies, pricing, and activation afterward in Polaris Suite. Before this work, a Customer Experience agent had to bridge every handoff between them by hand.
The problem
The field app and the platform never shared a clean data handoff, so every activation stalled at the seam between them.
- No owner: stations arrived on the platform with nothing linking them to the organization that bought them.
- Manual bridging: a CX agent stepped in on every activation to find the owner and connect the stations.
- No grouping: installing 10 stations at one site meant running the same flow 10 times.
Research & Discovery
I interviewed the CX agents who had been bridging every handoff, then mapped the data flow between field and platform end to end.
The critical insight: the first challenge wasn’t screens, it was timing. At what point in the physical workflow is org identity even knowable? Mapping the deployment showed three gaps the CX agents had been covering by hand, and those became the first release.
Iteration 1: MVP
What I built:
- Sites: a physical address becomes the unit of work, so stations installed together stay together.
- CMS declaration upfront: ChargePoint cloud or a third-party system (be.ENERGISED, Studio); the two paths split immediately.
- Org pre-assignment: org identity pushed from Salesforce and shown on the installer’s summary as a read-only signal.
- Email-triggered activation: a regional link sent to the org admin the moment the job is complete.
What worked: the release closed the three gaps CX had been bridging by hand.
What didn’t: stakeholder review surfaced four failures.
- Silent pre-assignment failure: when the API org link broke, the installer saw “Org ✓” while the admin received nothing.
- No asset visibility: admins couldn’t confirm from Polaris what had been installed.
- Per-station email didn’t scale: a 30-station site fired 30 separate emails.
- Meaningless site names: an auto-generated string like 450haciendacalifornia meant nothing to the admin.



Iteration 2: Refinement
Pre-assignment leaned on Salesforce data that was missing in about 30% of VAR transactions, and per-station emails broke at scale. Iteration 2 fixed both.
What I built:
- Org-agnostic completion: the install finishes cleanly without Salesforce data. The station queues in Polaris as “Ready for Activation”, and the admin gets a regional link to attach it to their org.
- Job Summary: all installed devices grouped into one submit event.
- Add more stations + cluster devices: installers loop back and add each station to the same job before submitting.
What worked: stations now arrive in Polaris ready to activate, with no CX touch and no dependency on Salesforce data.






How AI fit in
Microsoft Copilot (company-approved): summarising interview notes, drafting specs and Jira tickets, and first drafts of UX copy.
Figma Make: prototyping installer screens to test flow variations quickly. It was new to the team, so the first weeks went into learning what it could and couldn’t do.
What I learned
Both iterations came down to letting go of an assumption. The first version assumed org identity had to be resolved at the moment of install, and that one belief made the whole flow fragile. Once I stopped defending it, the design got simpler and sturdier.
This app is the upstream half of a two-part system. The activation flow in Polaris only works because this app captures the right data, in the right shape, at the right moment. That seam, not the app, is what I was designing.
- Design for the data you have, not the data you need. Designing from what was reliably present (the station MAC, the site address, the installer’s job record) produced a flow that survived reality.
- The handoff point is the design. Designing the installer-to-admin handoff explicitly, instead of routing it through a CX agent, is what made self-serve viable.
- If I did it again: I’d get engineering into the PRD earlier. The “sites” grouping shaped two iterations before engineering review found it wasn’t buildable as written.





