Show useful pricing earlier.
The observed quote route requires contact details before prices. Test whether customers can get a meaningful estimate before optional follow-up.
Inspect the contact gate →A customer should be able to understand the offer, trust the numbers and complete the next task without starting over.
Expert UX teardown · Desktop focus · 12 September 2026
Version 1 remains unchanged. This version adds interpretation and proposals.
Ordered by observed friction and likely task impact. These are qualitative priorities, not a quantified business case.
The observed quote route requires contact details before prices. Test whether customers can get a meaningful estimate before optional follow-up.
Inspect the contact gate →Revalidate the UAT balance and usage issues. Make amount due, dates, credits and missing data understandable before adding richer forecasts.
Inspect the account view →Fix Fern’s observed Home-command dead end. Make feedback optional and ensure useful exits and human support from every chat state.
Inspect Fern’s dead end →Recognisable service choices. The homepage states electricity, gas and broadband, and gives returning customers a portal entry.
Useful shortcuts. Address lookup, an ICP fallback, password recovery and a callback option accommodate different needs.
Multiple services together. The supplied account capture combines electricity and broadband; Fern supports topics and free text.
Some uncertainty is disclosed. Terms, sign-up stages, field errors and payment-processing delay guidance are visible.
Expert hypotheses to test with customers. These are not interview quotes or measured satisfaction scores.
| Customer job | What helps today | Likely friction | Opportunity |
|---|---|---|---|
| Find an offer | Clear join entry and assisted comparison | Comparison label leads to contact collection | Make immediate versus assisted outcomes explicit |
| Get a price | Property autocomplete and carried-forward details | Contact details required before plan prices | Explain necessary inputs; test earlier price visibility |
| Access an account | Login, registration and recovery | Online registration may be confused with joining supply | Clarify audience and complete recovery states |
| Understand money | Balance, statements and usage in one portal | Competing figures and unexplained data gaps in UAT | One dated account state and clear calculation scope |
| Move home | Public form covers service information | Long form and re-entry; fulfilment unobserved | Relevant questions, review and trackable confirmation |
| Get help | Fern topics, text and team option | Feedback state blocks a promised navigation command | Persistent exits and a tested support handoff |
All 28 captured screens, with strengths, friction and an improvement hypothesis. Click a screen for its original evidence. Customer questions are illustrative; no participant interviews were conducted.
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionWhere do I start for my task?
The hero names the services and offers a clear joining action; returning customers have a My Account entry.
Multiple acquisition and support routes compete. A first-time visitor may need to learn the navigation before comparing an offer.
Keep one primary Check prices action, with visible Pay a bill, Move home and Get help shortcuts. Carry the selected task through the next domain.
Run a first-click study for joining, paying and moving; check keyboard access to menus.
First-click accuracy and time to the correct task.
Meridian · account and moving help · Nielsen Norman Group · usability heuristics
Likely customer questionWill asking for help trigger a sales call?
A callback offers an assisted route and the pop-up can be dismissed.
An automatic contact request interrupts browsing and asks for several details before the customer knows the price. The extent of follow-up is not demonstrated.
Make callback an explicit choice. State contact purpose and an operationally achievable response window; ask only what the chosen channel requires.
Compare opt-in help with the automatic pop-up; verify focus containment and restoration.
Callback completion, unwanted-contact complaints and quote progression.
Likely customer questionDoes Compare rates show prices now?
Join and Compare rates recognise two useful customer intentions.
Compare rates leads to a contact request rather than an immediate comparison. The label may create a different expectation from the next page.
Distinguish See prices for my home from Ask us to compare a bill. State what each path delivers and when.
Ask prospects to predict each destination before opening it.
Expectation match and backtracking from the comparison form.
Nielsen Norman Group · usability heuristics · Meridian · public plan comparison
Likely customer questionCan I compare without waiting or uploading a bill?
Bill upload supports a tailored assisted comparison for customers who need it.
Contact fields, comments and CAPTCHA create effort before a result. The captured page does not provide a self-service comparison.
Offer an address-led estimate where pricing permits. Keep optional bill-based comparison with a clear turnaround and alternative to uploading.
Confirm quote-engine inputs before removing fields; test error recovery and an accessible CAPTCHA alternative.
Time to useful comparison, qualified sign-ups and assisted resolution time.
GOV.UK · question pages · W3C WAI · accessible forms · Meridian · public plan comparison
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionWhat will I pay after the joining incentive?
The address-first entry is relevant to energy eligibility and the offer links to conditions.
A prominent initial credit may distract from ongoing costs, term and exit conditions. Later price screens were not reached, so their disclosure quality is unknown.
Alongside the offer, preview the information customers will get: ongoing rates, daily charge, estimated total, contract conditions and promotion expiry.
Test offer comprehension, including the post-promotion position, without prompting.
Correct understanding of ongoing cost and offer conditions.
Nielsen Norman Group · usability heuristics · GOV.UK · check answers · Meridian · public plan comparison
Likely customer questionWhat is an ICP and where would I find it?
An ICP fallback can help when address matching fails.
Technical terminology adds recall effort; no explanation is visible beside this control.
Explain ICP in plain language, show where to find it on a bill and retain an address-help route.
Test with people unfamiliar with energy terminology; label and announce format errors.
Successful recovery from unmatched addresses.
GOV.UK · question pages · W3C WAI · accessible forms · Nielsen Norman Group · usability heuristics
Likely customer questionIs this definitely my property?
Autocomplete returns a selectable match and reduces manual entry.
The displayed location string differs from the later property summary. Even a valid lookup may look unreliable when address formatting changes.
Use a consistent formatted address and ask customers to confirm the supply point. Provide No match and Wrong property recovery.
Check apartments, rural addresses, duplicate street names and keyboard autocomplete.
Wrong-property corrections and successful address matches.
W3C WAI · accessible forms · Nielsen Norman Group · usability heuristics
Likely customer questionWhy do you need my contact details to show my price?
Property details are carried forward and the four-step journey names what comes next.
The first stage includes contact collection and user-type choices before price visibility. Customers may not understand why these are needed or whether Standard User is suitable.
Test a price preview before optional follow-up details. Explain any truly required contact field and make tariff assumptions visible and editable.
Validate pricing and contact requirements with operations; compare comprehension and completion in a prototype.
Price-view rate, completed sign-ups and quote accuracy.
GOV.UK · question pages · Nielsen Norman Group · usability heuristics
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionCan I continue without agreeing to follow-up?
Empty submission produces individual errors for the required contact fields.
The flow blocks price discovery until name, mobile and email are supplied; the page also says Pulse may contact incomplete applicants. This is a trust and conversion hypothesis, not measured abandonment.
Separate necessary quote inputs from optional follow-up permission where feasible. Keep entered values, an error summary and clear field guidance.
Audit actual data dependencies; test with prospective customers and measure before/after only after instrumentation.
Quote-to-price completion; guardrails: lead quality, complaints and pricing errors.
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionWhich route gets this account task done?
The menu exposes a broad range of support and account tasks, including moving and payment options.
Information pages and actions share a large menu. Customers may not know which paths require login.
Group around customer tasks and label account-only actions. Keep public help available without making login a universal gate.
Tree-test bill, account, moving and broadband tasks.
Task-find success and unnecessary login detours.
Electric Kiwi · Kiwi Central · Meridian · account and moving help · Nielsen Norman Group · usability heuristics
Likely customer questionAm I joining Pulse or accessing my existing account?
Show Password, registration and recovery are visible; the portal explains its value.
A separate domain and a registration route can blur the difference between buying energy and enabling online access.
Use Already a Pulse customer? and Register online access labels. Explain the domain transition and support password managers and pasted credentials.
Test login with keyboard and password managers; accessible-authentication support is unverified.
Successful login, recovery completion and misrouted new customers.
W3C · accessible authentication · Nielsen Norman Group · usability heuristics
Likely customer questionWhere is my account number?
Registration asks for just email and an existing account number.
The account-number dependency can strand people who do not have a bill handy. A disabled button alone does not explain missing or invalid inputs.
Show a bill example locating the number, explain email matching, and offer a recover-account route. Confirm the next verification step.
Test with first-bill and multi-account customers; inspect labels and status announcements.
Activation completion and account-number-related support contacts.
W3C WAI · accessible forms · W3C · accessible authentication
Likely customer questionWhat happens after I request a reset?
Password recovery is short and provides a route back to sign-in.
Only the empty state was observed. Delivery, expiry, resend and success states remain evidence gaps rather than confirmed defects.
Design a clear, privacy-preserving confirmation, resend route and expiry recovery. Preserve the task after successful sign-in.
Capture the complete authorised reset journey; test email delay and expired links.
Reset completion and time to regained access.
W3C · accessible authentication · Nielsen Norman Group · usability heuristics
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionDo I owe money, and what should I do now?
A calm card layout combines payment and service usage; direct debit is near the payment action.
The UAT capture shows Pay Now with zero due, several money figures and an overflowing greeting. The chat launcher competes with View more.
Lead with one dated account state and a state-appropriate action. Keep ledger detail expandable; label the property switcher and make navigation consistent.
Recheck in production, then test whether customers identify amount due, date and next action without explanation.
Balance comprehension, accidental payment starts and billing contacts.
Octopus UK · usage in money and units · Nielsen Norman Group · usability heuristics · Supplied My Account audit · 8 August 2026
Likely customer questionWas that bill paid, and which period did it cover?
Statements, bank details and SmoothPay information appear on one page.
The UAT list has repeated primary buttons, no visible paid/due status or billing periods, a large historical gap and an email route to SmoothPay.
Use readable statement rows with period, amount and reconciled payment status. Add year grouping and preset filters; explain payment-plan eligibility before setup.
Confirm billing data can support accurate statuses; validate credit and partial-payment cases.
Correct bill retrieval, status comprehension and payment-plan completion.
Mercury · SmoothPay eligibility and setup · Meridian · account and moving help · Nielsen Norman Group · usability heuristics · Supplied My Account audit · 8 August 2026
Likely customer questionWhat did my electricity cost, and why did it change?
Daily/monthly controls and a year filter expose basic usage exploration.
The UAT view is kWh-only, with unexplained empty months and a Compare label that does not describe its comparison. Static evidence cannot establish keyboard or tooltip behavior.
Provide money/units views, a named comparison period, fresh-data labels and a table alternative. Reconcile cost scope with the bill.
Verify rates, GST, daily charges and meter coverage; test chart interpretation and assistive technology.
Cost comprehension, reconciliation accuracy and usage-related contacts.
Mercury · usage and cost breakdown · Octopus UK · usage in money and units · AGL · usage and projection calculations · W3C · WCAG 2.2 · Supplied My Account audit · 8 August 2026
Likely customer questionIs zero usage real, or is the data missing?
Broadband usage and the named plan are visible together.
The UAT summary shows zero while earlier months have usage; no clear last-read state resolves the uncertainty. Fine print and number formats vary.
Distinguish no reads from zero consumption; show last update and a next step. Add relevant connection troubleshooting and plan-management routes.
Recheck real data states and the service API; test stale, disconnected and newly connected scenarios.
Correct state interpretation and avoidable broadband contacts.
Electric Kiwi · Kiwi Central · Nielsen Norman Group · usability heuristics · Supplied My Account audit · 8 August 2026
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionWhy am I re-entering details you already know?
The moving form covers account/current-property details and directs urgent cases to a phone channel.
A long public form requires account information without an observed authenticated shortcut. Date defaults may be mistaken for an intentional selection.
Offer a signed-in route with confirmed saved details and an accessible guest fallback. Start with urgency and current/new-property dates.
Verify authenticated moving before assuming it is missing; test renters, owners and overlapping properties.
Move completion, re-entry effort and date corrections.
Meridian · account and moving help · AGL · track a new connection · GOV.UK · question pages
Likely customer questionWhich questions apply to my move?
Service and access questions could support a coordinated connection.
Many questions and preselected No answers demand sustained attention; medically dependent or unusual-property cases may be missed.
Group by current home, new home and services. Reveal relevant questions progressively and require deliberate answers to consequential questions.
Review defaults with service operations; test conditional logic and recoverable back navigation.
Incomplete requests and corrections to dates, access or support needs.
Likely customer questionHas my move been arranged, and when will power be on?
Terms and submission are visible at the end of the public form.
Review, confirmation and status were not observed. The long-form submission boundary provides no evidence of what assurance customers receive afterwards.
Add a check-answers step and a receipt with reference, confirmed versus requested dates, remaining actions and trackable progress.
Observe an authorised completed move before claiming any current confirmation is absent.
First-time-complete requests and where-is-my-move contacts.
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionCan this assistant help with my problem?
Fern is discoverable and stays alongside the page; the conversation persisted when reopened in this session.
A floating launcher and callback pop-up can compete for space and attention. Session persistence beyond this test is unknown.
Keep a predictable safe area and a clear scope label. Preserve the current page and task when chat opens or closes.
Check zoom, small desktop windows, focus and collision with important controls.
Assistance discovery and obscured-control incidents.
Electric Kiwi · Kiwi Central · W3C · minimum target size · Nielsen Norman Group · usability heuristics
Likely customer questionWhy must I open chat twice?
The welcome panel clearly provides a Message us action.
The extra welcome screen delays reaching a useful prompt after the visitor has already opened Fern.
Test opening directly to a compact task menu and text field, with assistant identity and scope in a short header.
Compare time to first useful response; verify panel focus and close behavior.
Time to first question and abandonment before first input.
Likely customer questionWhich topic describes my issue?
Customers can use topic buttons or free text. Fern identifies itself as a virtual assistant.
Multiple introductory messages precede the menu. The Home-anytime promise creates an expectation the tested feedback state fails to meet.
Lead with tasks such as Compare prices, Understand a bill and Moving home. Keep necessary disclosures available without burying the task menu.
Test task wording and whether new messages are announced without moving keyboard focus unexpectedly.
Correct topic selection and time to useful answer.
Electric Kiwi · Kiwi Central · W3C WAI · accessible forms · Nielsen Norman Group · usability heuristics
Likely customer questionWhy do I need to type a number for a simple question?
Clarification gives Fern a way to avoid answering the wrong joining topic.
The clarification switches from buttons to numbered text, adding recall and a new interaction convention.
Answer clear joining intent directly; when uncertain, offer short tappable choices and an escape route.
Run paraphrase tests, including spelling errors; evaluate intent resolution rather than message count alone.
Correct first answer and unnecessary clarification rate.
Original screen order retained. Annotations are expert hypotheses, not observed customer reactions.
Likely customer questionWill I have to restart on another site?
Fern provides the correct sign-up destination and a way to indicate whether it helped.
The generic link moves the customer into a separate flow and the Yes/No prompt asks about usefulness before task completion.
Use a descriptive Check prices for my home link, explain the next step and retain safe task context. Make feedback optional.
Check continuity and referral state across domains; do not pass private data in URLs.
Chat-to-price completion and repeated information.
Likely customer questionWhy does Home not work when you said it always would?
A Yes response followed by Home did restore the menu in the observed conversation.
Home is rejected at the feedback prompt. This is a directly observed navigation inconsistency, not a hypothetical chatbot risk.
Handle Home, Back, Cancel and human-help intent before the feedback state. Accept the next question without requiring a rating.
Regression-test escape commands from every conversational state, including errors and handoffs.
Navigation dead ends and successful task continuation.
Likely customer questionCan I get straight to the billing task?
The menu recognises concrete tasks such as balances, refunds and updating details without demanding an account number first.
A broad menu still requires users to classify their problem; the tested balance path supplies guidance, not an account balance.
Keep common tasks visible and state when secure sign-in is needed. Retain free-text input and a clear human route.
Test mixed intents such as a paid bill still showing as due.
Correct routing and repeat explanations after login or transfer.
Electric Kiwi · Kiwi Central · Meridian · account and moving help · Nielsen Norman Group · usability heuristics
Likely customer questionCan I reach a person without explaining everything again?
Fern links to My Account, states a payment-processing delay and offers team contact and navigation.
Agent availability, transfer success and context retention were not tested. A portal link alone may not resolve a disputed balance.
Show available channels and realistic wait expectations. With customer agreement, carry a short conversation summary into a human handoff and confirm the next step.
Complete an authorised handoff and test unavailable-agent recovery; measure outcome rather than containment alone.
Resolved tasks, repeated-contact rate and handoff completion.
Electric Kiwi · Kiwi Central · Nielsen Norman Group · usability heuristics
No screens match these filters.
What other energy providers make clear—and how Pulse could make it useful for customers.
Fresh captures · 12 September 2026 · Desktop research, including clearly labelled mobile-app previews and a public demo.
A useful comparison can begin before a sales conversation.
Pulse carries the property forward and names the sign-up stages, but the observed price route stops at required contact fields. Prices beyond that gate were not inspected.
A shared comparison structure can reduce memory load and make plan fit easier to judge. This is a design rationale to test, not a measured competitor advantage.
Show household fit, term, recurring charges and offer expiry in consistent rows. Then request only the inputs needed for an accurate property price.
Do not make promotional credits or export buy-back rates look like the cost of imported electricity. Do not claim an address-only quote is feasible until pricing inputs are confirmed.
Ask prospects to choose a plan and explain first-year versus ongoing cost. Measure correct understanding and address-to-price completion.
Evidence and principles: Meridian · public plan comparison · GOV.UK · question pages
The most useful balance is a sentence, not just a number.
The supplied Pulse UAT dashboard brings several services together. Competing money figures need production revalidation before any redesign decision.
Words such as “in credit” reduce the need to infer meaning from a sign or colour. Pairing state with the next relevant task can make the screen easier to act on.
Give amount due, available credit, payment processing and next scheduled payment distinct labels. Show the effective date and one relevant primary action.
Do not treat “in credit” as proof that no future payment is scheduled. Do not copy an illustrative competitor balance as a complete state model.
Use four scenarios: credit, amount due, zero due and pending payment. Ask what is owed, when, and what action is needed.
Evidence and principles: Octopus UK · usage in money and units · AGL · app billing and usage previews · Supplied My Account audit · 8 August 2026
Customers need a period, a unit and an explanation—not a chart alone.
The supplied Pulse UAT electricity screen shows kWh and period controls. The audit flags data clarity issues; production behavior and available reads remain unverified.
Keeping period, total and unit beside the graph can help users answer “when did this happen?” and “what did it cost?”. A good empty state is as important as the normal chart.
Show money and energy where calculations support them. State included charges, update time and missing reads; provide a table alternative.
Do not compare unlike totals. Mercury and Octopus document different cost inclusions. Do not show missing data as zero or promise live data without the required meter feed.
Ask users to find a high-use day and reconcile the view with a bill. Include keyboard-only use and a missing-data scenario.
Evidence and principles: Mercury · usage and cost breakdown · Octopus UK · usage in money and units · Genesis · monthly usage demo · Genesis · daily usage demo · Supplied My Account audit · 8 August 2026
An account becomes useful when information leads to a relevant task.
Pulse already places broadband alongside electricity in the supplied UAT account. That is a foundation to retain; the next question is what users can do from each service.
Service-specific actions can save users from translating a problem into the retailer’s organisational structure.
Give broadband a clear route to connection help or troubleshooting, and electricity clear bill and usage tasks. Retain a coherent overview across services.
Do not add decorative service tiles with no useful destination. A router speed test needs technical support; it is not just a frontend control.
Start with “my internet is slow” and “why did electricity cost more?”. Watch where users go and whether support keeps the selected service context.
Evidence and principles: Electric Kiwi · Kiwi Central · Contact · energy and mobile app previews · AGL · app billing and usage previews · Supplied My Account audit · 8 August 2026
Forecasts need a clear distinction between known cost and uncertainty.
The inspected Pulse states do not establish whether forecasts or alerts exist. This is a capability exploration, not a confirmed production gap.
A cost-to-date and projected-total pair can help customers anticipate a bill. Opt-in alerts could make that useful without requiring repeated account checks.
First separate accrued usage cost from amount due. If forecasts are reliable, show estimate assumptions and let customers choose a threshold and channel.
Do not make a forecast look like a bill or an enforced budget cap. Do not copy PG&E’s eligibility rules or Genesis’s sample values into a NZ service design.
Measure forecast error by customer segment, then test whether customers distinguish actual from estimated values. Include opt-out and stale-data behavior.
Evidence and principles: Genesis · public Energy IQ demo · PG&E · Bill Forecast Alert · AGL · usage and projection calculations
Submitting a move starts a service process; customers need to see what happens next.
Pulse’s public move form gathers property and service details. The confirmation, tracking and operational outcome were not submitted or inspected.
A durable reference and a return route can reduce uncertainty after a request leaves the browser. The status is only useful if it reflects operations accurately.
Show received, awaiting information, scheduled and confirmed states where operationally valid. Give the customer one secure status link and the next required action.
Do not label a submitted request as a completed connection. Do not transplant Australian safety or connection instructions into Pulse copy.
Walk through delayed and changed-date cases with operations; test whether customers can find status again from an email and their account.
Evidence and principles: AGL · track a new connection · Contact · energy and mobile app previews
Good assistance lets customers change direction and recover.
Fern offers topics and free text, but the captured Home command was rejected during feedback. A human transfer was not completed in the audit.
Customer-language questions improve recognition. A clear escape and handoff path can keep help from becoming an additional task.
Use shared task labels across help and account navigation. Make feedback optional; give Fern persistent restart, back and human-help routes with accurate availability.
Do not infer that a competitor’s promised seamless handoff works in practice. Do not pass account details or chat history without a secure, appropriate context transfer.
Test a wrong answer, topic change, feedback refusal and after-hours request. Measure confirmed resolution and successful recovery, not bot containment alone.
Evidence and principles: Contact · task-based account help · Electric Kiwi · AI and human support expectations · Nielsen Norman Group · usability heuristics
Numbered for reference. Select a screenshot for the strength, tradeoff, Pulse application and validation test. App images are published examples; they are not logged-in captures.
The preview connects balance, weekly cost and daily usage in one visual story.
Monthly, Daily and Hourly controls stay together; clicking Daily changed the chart.
The selected Daily state is visible and its chart differs from Monthly.
Plan names and short use cases give customers a starting point for comparison.
Consistent rows align term, fit and offer across plans.
Three labelled previews distinguish service overview, usage and account administration.
Power, mobile and broadband examples pair information with relevant controls.
Published guidance states human choice and wait-time expectations.
An app image sits beside practical billing, usage and moving tasks.
A question expands in place and gives a direct route to more specific instructions.
Period tabs, totals and a money/energy switch sit with the chart.
The balance example states that the account is in credit and shows the recurring payment.
The guide documents tracking through both the account and confirmation email, including an estimated date.
The app illustration groups a bill amount, payment help, direct debit and a breakdown.
The tutorial illustrates an alert switch with on/off text and a customisation link.
Household tiles start a public demo without real account credentials.
The demo puts accrued cost, an estimated bill and time remaining together.
No captures match both filters. Reset to explore all providers.
Five New Zealand retailers and three international references. This compares inspected patterns, not market leadership or measured customer satisfaction.
| Provider | What we inspected | Transferable strength | Limit / tradeoff |
|---|---|---|---|
| Meridian NZ | Public plan cards and a desktop comparison table; in-page navigation inspected. | Consistent plan attributes and household-fit language. | Property-specific price and downstream sign-up were not tested. |
| Mercury NZ | Official app preview captured on desktop; usage documentation reviewed. | Preview the account and connect balance to usage. | App imagery does not establish desktop parity or universal meter coverage. |
| Electric Kiwi NZ | Official app previews and AI support guidance. | Service-specific actions and stated human choice. | No app tasks or live human transfer tested. |
| Contact NZ | App overview and account-help disclosure opened. | Task-language help and concrete self-service benefits. | Separate energy and mobile apps; authenticated tasks untested. |
| Genesis NZ | Public sample-household demo; setup and Monthly → Daily inspected. | Try-before-activation and cost-to-date versus estimate. | Simplified mobile-shaped demo; inconsistent sample dates. Not current production proof. |
| Octopus UK | Official app illustrations and usage guide. | Plain-language credit state; time and unit controls. | Live data and charts have conditions; app is not desktop parity. |
| AGL Australia | Public tracking instructions and app preview. | Return route after a move; payment help beside bill details. | Actual tracker and payment process untested. Localise operational rules. |
| PG&E US | Official illustrated alert setup instructions. | Explicit on/off state and customisation. | Restricted eligibility; estimates are not limits. No alert preference changed. |
“Unknown” means outside the inspected Pulse states. App examples inform the desktop direction; they do not establish a like-for-like desktop feature gap.
| Customer outcome | Pulse evidence | Documented reference | What to explore |
|---|---|---|---|
| Understand plan fit | 03–04: Compare rates leads to a contact request | Meridian · public plan comparison | Public plan differences, plus an explicit path to property-specific pricing |
| Know the cost behind usage | A3: kWh-only UAT capture; production unverified | Mercury · usage and cost breakdown · Octopus UK · usage in money and units | Money/units, period and included charges |
| Choose manageable payments | A2: SmoothPay email route in UAT | Mercury · SmoothPay eligibility and setup · Meridian · account and moving help | Eligibility-aware setup and a clear exception path |
| Manage multiple services | A1/A4: electricity and broadband present | Electric Kiwi · Kiwi Central | Preserve this strength; add practical service actions |
| Track a move | 14–16: public form only; confirmation unknown | AGL · track a new connection | Status link, estimated date and next action |
| Avoid a bill surprise | Forecasts/alerts not observed | AGL · usage and projection calculations · PG&E · Bill Forecast Alert · Octopus UK · Balance Forecast | Validate data before opt-in projections or alerts |
| Get help in context | 17–24: chat and portal handoff observed | Electric Kiwi · Kiwi Central | State-aware navigation and tested human transfer |
A target service experience to validate, not a redesigned screen set or a claim about current capabilities. Copy examples below are illustrative; dates, costs and timings must come from real service data.
A new visitor sees Check prices, Move home and Existing customer sign-in. An existing customer sees familiar account tasks.
One intent follows the customer through website, quote, account and help.
Task labels are understood without explanation; switching routes keeps useful context.
Nielsen Norman Group · usability heuristics · Meridian · account and moving help
Confirm the property, then see a comparable estimate and ongoing rates before optional sales follow-up, where feasible.
State usage assumptions, daily charges, GST, offer duration, contract and applicable exit conditions. Explain any required data.
The customer can explain the ongoing cost and edit an assumption; no invented exact price is shown.
Provide necessary details in short groups, choose services, review answers and submit once.
Show why details are required, save progress safely and separate optional preferences from essential service information.
Back retains answers; errors are actionable; review edits return to the review page.
GOV.UK · question pages · GOV.UK · check answers · W3C WAI · accessible forms
Receive a reference, status link, requested/confirmed dates, next actions and account-access guidance.
A pending application is not presented as an active supply. Delays and exceptions have a clear owner and next update.
Customers distinguish received, in progress and confirmed; status is available outside email alone.
AGL · track a new connection · Nielsen Norman Group · usability heuristics
One headline says what is due and when. Pending payments and credits are dated; service and property context is clear.
A zero balance has an appropriate next action. Bills show period, amount and actual reconciled status.
Customers identify amount, date and action; totals reconcile with the source bill.
Meridian · account and moving help · Octopus UK · usage in money and units · Nielsen Norman Group · usability heuristics
Inspect cost and usage with a named period, compare like-for-like, manage eligible payment options and move home with saved details.
Missing, estimated, stale and zero data are distinct. Forecasts are labelled and optional alerts explain their scope.
The customer understands the displayed cost and can complete an eligible change without repeating known information.
Mercury · usage and cost breakdown · Mercury · SmoothPay eligibility and setup · AGL · usage and projection calculations · PG&E · Bill Forecast Alert
Fern answers routine questions, honours navigation commands and makes human assistance easy to find.
Sensitive account actions move through secure authentication. Feedback never blocks the next task; transfers carry agreed context.
The user can exit, change topic or reach an available channel from every state. Handoff resolution is measured.
Electric Kiwi · Kiwi Central · W3C · accessible authentication · Nielsen Norman Group · usability heuristics
Today, observed: Address → property form → name, mobile and email → prices not reached.
Proposed: Address → necessary tariff inputs → transparent price comparison → optional saved quote or application → review → status receipt.
Dependency: the pricing team must confirm what can be quoted without contact details. If contact remains essential, explain why before entry.
Today, observed: Fern → account link → login boundary. The UAT account shows multiple money figures.
Proposed: Secure account → dated balance and pending-payment state → relevant bill → help with this payment, carrying agreed context.
Dependency: payment reconciliation and identity checks. A pending payment must not be labelled settled.
Today, observed: Long public form covering old home, new home, services and terms.
Proposed: Confirm saved property → choose dates → relevant service/access questions → review → request reference → track confirmation.
Dependency: operations owns fulfilment status. Keep a guest and assisted route for customers unable to sign in.
Explain each question; allow backtracking; reuse prior answers. A shorter form is useful only if it still gathers necessary service information.
Use persistent labels, specific inline errors and an error summary; retain entered values and announce state changes.
Provide check answers with targeted edit links and a clear final action. Confirmation must distinguish request receipt from fulfilment.
Permit password managers and paste; provide an authentication approach consistent with WCAG 2.2 accessible-authentication requirements. Test the actual flow.
Assess 24×24 CSS-pixel targets or applicable spacing/exceptions; inspect focus, contrast, zoom/reflow and non-colour status cues. Screenshot estimates are not conformance results.
Keep exits predictable, labels consistent, data dated and next steps clear. These are heuristic principles to test in Pulse’s context.
P1 = core task, trust or accessibility risk. P2 = clarity and avoidable effort. P3 = capability exploration. Priorities and S/M/L effort are provisional expert estimates, not delivery commitments.
| Order | Opportunity | Effort / owner | Dependency | Success measure | Guardrail |
|---|---|---|---|---|---|
| 1 P1 | Remove Fern navigation dead ends Screens 22, 20–24 | S Chat owner + service team | State-machine ownership and approved support routing | Escape-command failures; handoff success | No fall in answer accuracy or loss of context |
| 2 P1 | Clarify the account balance and current action Screens A1 | S–M Portal product + billing | Revalidate the UAT issue in production; reliable balance states | Amount/date/action comprehension | No misleading pending-payment or credit state |
| 3 P1 | Test earlier price visibility Screens 04, 08–09 | M–L Acquisition + pricing + privacy | Confirm minimal pricing inputs and follow-up requirements | Address-to-price and price-to-completed-sign-up rates | Quote accuracy, service eligibility and unwanted-contact complaints |
| 4 P1 | Fix measured accessibility blockers Screens All; especially forms and A1–A4 | M Design system + engineering | Live DOM, keyboard, zoom and assistive-technology audit | Task completion using keyboard and screen reader | No blanket compliance claim from screenshots |
| 5 P1 | Make bills and usage tell a consistent story Screens A2–A4 | M–L Billing + data + portal | Statement metadata, reconciliation and freshness signals | Bill retrieval and cost interpretation | Status accuracy and no false-zero displays |
| 6 P1 | Simplify moving and expose progress Screens 14–16 | L Service operations + portal | Confirm existing authenticated route; connection status integration | Completed requests and status-related contacts | No missed dates, access or medically dependent support needs |
| 7 P2 | Align task labels and reduce unsolicited interruption Screens 01–03, 10, 17–19 | S–M Website + content | First-click study and callback operations | Task-find success and time to useful answer | Assisted help remains easy to discover |
| 8 P2 | Improve activation and recovery Screens 11–13 | M Identity + support | Observe verification and expired-link journeys | Activation/reset completion | Privacy-preserving errors and accessible authentication |
| 9 P3 | Explore forecasts and opt-in budget alerts Screens A3–A4; unobserved capability | L Data + service product | Reliable meter data, rates, eligibility and explicit opt-in | Forecast calibration and useful-alert ratings | False alarms, alert fatigue and unsupported precision |
Inspect production account states and the authorised quote, activation, recovery and move-confirmation paths. Capture zero due, credit, arrears, pending payment, no reads and stale data.
Output: confirmed defects versus outdated UAT findings.
Start with 6–8 purposively selected participants across prospects, bill-payers and movers, including people with access needs. This is a formative sample, not a statistical estimate.
Tasks: compare an offer, explain a balance, find a paid bill, recover access, arrange a move and escape a Fern feedback loop.
Record step-level completion and errors across domains using privacy-reviewed identifiers. Pair funnel data with support reasons; do not record private form values or raw chat content by default.
Output: baselines, denominators and segment coverage.
Test price-preview ordering, a single-state balance card and Fern’s persistent navigation. Choose quantitative targets after observing the baseline and available traffic.
Output: evidence to keep, revise or reject each proposal.
| Measure | Definition | Interpretation |
|---|---|---|
| Address-to-price rate | Sessions reaching a usable price ÷ sessions with a matched eligible address | Segment by device and address eligibility; pair with final sign-up quality |
| Account comprehension | Participants correctly identifying due amount, date and action ÷ participants tested | Report sample size and scenario; do not generalise small samples |
| Move first-time completeness | Move requests needing no customer correction ÷ submitted move requests | Separate form quality from operational connection delays |
| Assistance resolution | Supported tasks confirmed resolved ÷ supported tasks with an observable outcome | Report unknown outcomes separately; containment alone is not success |
The original board is the current-state record. Version 2 layers expert interpretation and proposals over that record, plus public benchmark research.
24 desktop screens from 12 September 2026: discover, quote entry/validation, account access, moving form and Fern. Reused from the user-approved version 1.
The authorised property lookup reached a contact-data gate. No enrolment, payment, move submission, reset email or human chat transfer was completed.
4 logged-in UAT captures dated 8 August 2026. They support discussion of visible layout and information; they do not establish the current production implementation.
The source’s accessibility and interaction claims require live checks. Its proposed redesigns are not current-state evidence.
8 providers across NZ, UK, Australia and US. The expanded library adds 17 inspected captures: public desktop pages, published app images, setup illustrations and a simplified Genesis demo.
Public page disclosures and sample Genesis demo controls were inspected. No authenticated competitor account, quote submission, mobile-app session or controlled usability test was performed. Published imagery is not performance evidence.
Customer questions, risk assessments, ideal journeys, priorities and effort are expert hypotheses. There are no analytics baselines, participant quotes or measured uplift in this board.
Unobserved downstream screens, identity controls, accessibility behavior and operations capacity remain open questions.
Public reference pages reviewed 12 September 2026. Capture dates for Pulse evidence are labelled separately.
B1Mercury · usage and cost breakdown
B2Mercury · SmoothPay eligibility and setup
B3Electric Kiwi · Kiwi Central
B4Meridian · account and moving help
B5Octopus UK · usage in money and units
B6AGL · track a new connection
B7AGL · usage and projection calculations
B10Meridian · public plan comparison
B9Octopus UK · Balance Forecast
P4W3C · accessible authentication
P6Nielsen Norman Group · usability heuristics
UATSupplied My Account audit · 8 August 2026
B11Contact · energy and mobile app previews
B12Contact · task-based account help
B13Genesis · public Energy IQ demo
B14Genesis · monthly usage demo