Designing a customer effort score framework for the Utility Warehouse app

Sector: Utilities / Customer experience measurement.
Our role: Product design, CES/feedback capture framework.

The challenge

Utility Warehouse runs several digital journeys at once (switching a customer’s broadband, ordering a new SIM, setting up a direct debit, raising a support request) and every product team had its own idea of whether those journeys were actually easy to use. There was no shared, ongoing way to measure customer effort across the app, only the risk of each team bolting on its own one-off survey.

Left alone, that risk is a real cost: ask customers for feedback too often, in too many disconnected ways, and the feedback itself becomes a source of effort and irritation. The brief was to design a single, coherent Customer Effort Score (CES) framework, sometimes referred to internally as a Digital Effort Score, that any team could plug a journey into, with a consistent survey mechanism, a shared scoring model, and rules for when a customer should (and shouldn’t) be asked.

The initiative sat under a wider “Digital Ease Measurement” workstream, with a clear top-line target: measure and increase Customer Effort Score for digital interactions by 25%, giving the business a credible way to prioritise fixes and prove the experience was actually improving over time.

Defining the problem before designing the solution

Before any screens were drawn, the work started with the questions a measurement framework lives or dies on: what, exactly, are we trying to find out, from whom, and why does it matter to the business? That meant setting out the hypothesis, the objectives, clear success criteria and, critically, a method for identifying which journeys were worth measuring first, rather than trying to instrument everything on day one.

A simple decision framework captured how a feedback request should actually get triggered:

  • Customer does something worth knowing about: a self-serve form, an existing-customer interaction, a new sign-up.
  • Category of thing: which kind of service and customer relationship it sits under.
  • Type of feedback request: a post-support call, an onboarding journey, a self-serve form, or something else.
  • Context: did the customer complete it successfully, start and abandon it, or something in between?
  • Feedback request history: has this customer already been asked recently? Rules here stop any one customer being pestered by multiple teams.
  • Decision: send the request now, queue it for later, or don’t send it at all.

That last point mattered as much as any individual survey design: the framework had to work for the whole customer journey rather than the slice one team cared about, and it had to actively prevent feedback fatigue rather than leave it to chance.

Designing the survey itself

With the measurement principles agreed, the next phase was the survey mechanism customers would actually see. Several question formats were explored side by side: a numbered 1–7 scale (the traditional CES format), a Likert agreement scale, a 5-star rating, an emoticon scale, and a simple binary thumbs up/down, each weighed against how quickly a customer could answer on a phone screen, how much nuance it captured, and how well it would hold up across very different journey types, from a quick SIM order to a multi-step broadband switch. The two mockups below are concept explorations from that phase, not final shipped UI. See “The finished in-app journey” further down for how the mechanism actually ended up looking once built.

A consistent follow-up pattern sits behind every scale format shown above: a low or middling score always triggers a lightweight “what went wrong?” question (difficulty finding the form, difficulty understanding it, too many steps, or something else), plus an optional free-text field. That’s what turns a CES number into something a product team can actually act on, rather than a score with no explanation attached.

In-context feedback, not a delayed survey

The most important design decision in the whole project was when to ask. Rather than a generic, delayed survey email days after the fact, the framework places the feedback prompt at the moment a task completes, while the effort involved is still fresh in the customer’s mind. A trigger strategy distinguished between automatic triggers (the system detects a completed or abandoned journey and surfaces the prompt itself) and user-initiated triggers (a customer opts in to give feedback, for example from a “was this helpful?” link), with clear rules for which journeys should use which.

A mobile broadband confirmation screen reading 'Your switch is complete! Welcome to Utility Warehouse broadband', with an inline prompt below reading 'How was your switch? Help us to improve'.
Mockup: an automatic, in-context trigger on a broadband switch: the feedback prompt sits directly inside the confirmation screen itself, rather than arriving as a separate, delayed survey.

This extended to designing the component as a true, reusable piece of UI rather than a one-off screen. The feedback card needed defined states (open, minimised, closed), had to remember where a customer left off if they minimised it partway through, and had to behave consistently whether it appeared on the web journey or inside the app.

A Figma canvas showing an annotated feedback component spec for a web confirmation page, with Interaction, Content and Note callouts documenting how the expandable feedback card behaves across its trigger, steps and confirmation states.
Component-level interaction design in Figma: every state of the expandable feedback card (trigger, step-through, and confirmation), specified and annotated for handover, alongside notes on edge cases like what happens if a customer minimises it mid-flow.

The finished in-app journey

Rather than a blocking pop-up, the shipped entry point sits quietly on the home screen as a dismissible card alongside other prompts: a deliberately low-pressure version of the “ask at the right time” principle, giving a customer the option to give feedback without interrupting whatever they came to the app to do.

Three sequential app screens: a 'Your services are live!' confirmation screen, the home screen 'Good morning!' with a 'How was your experience? Help us to improve' card above a 'Save time every month' prompt, and the resulting onboarding-experience rating screen with a blank 5-point emoji scale.
Final design, step by step: the services-live confirmation, the home screen surfacing a “how was your experience?” card alongside (not instead of) another prompt, and the rating screen it opens into.
Three sequential app screens continuing the journey: the onboarding-experience rating screen with a dissatisfied emoji selected, a 'Thank you' confirmation screen, and the home screen again with the feedback card now gone, leaving only the 'Save time every month' prompt.
...continuing: a score selected, the thank-you confirmation, and the home screen afterwards: the feedback card has done its job and doesn't reappear.

The same mechanism was designed for the self-serve Help forms. Once a request is sent, the “All done!” screen offers a light “Rate your experience with this form” link rather than an interruption; a rating opens a bottom sheet with an optional “what went wrong?” follow-up, then a short thank-you. Alongside the screens sits the component set: a star rating at three sizes, and bottom sheets for a star CES, a larger star CSAT and an emoji CES, so every team builds from the same parts.

From scores to action: prioritisation and continuous improvement

A Customer Effort Score is only useful if it changes what a team works on next. Two frameworks were designed to close that loop:

A prioritisation framework table showing seven example journeys (Pay bill, Move home, Order additional SIM, Initial onboarding, Broadband upgrade, Submit meter reading, Manage/setup DD) each scored by CES percentage, customer volume, business impact and a resulting priority rating from Low to Critical.
The prioritisation framework: CES score combined with customer volume and business impact to produce a single priority rating per journey, so “initial onboarding”, with medium volume but high impact and a high effort score, surfaces as Critical ahead of lower-stakes journeys.
  1. MeasureCES and analytics
  2. DiagnoseQuantitative and qualitative insights
  3. FixDesign UX improvements
  4. TestA/B or usability testing
  5. Re-measureCES again, to confirm the change

Then back to Measure

The ongoing cycle the framework is designed to run on: measure, diagnose, fix, test, re-measure. A CES score is where an investigation starts.

An admin-facing feedback dashboard concept was also scoped alongside the customer-facing mechanism, so the people acting on the data weren’t left combing through raw survey responses. The concept let the team search feedback by customer, by journey category, or by specific feedback request, with an aggregate CES calculated automatically as the mean score across all respondents for that journey, giving product teams one shared source of truth instead of each team keeping its own spreadsheet.

The target

The framework was built against a clear goal: closing the gap between current effort scores and a 25% improvement target, giving Utility Warehouse a concrete benchmark to design and prioritise against, and a rollout plan to get there in phases rather than instrumenting every journey at once, starting with the highest-priority journeys identified by the framework above, then expanding coverage once the survey mechanism, trigger rules and dashboard had proven themselves on real traffic.

← Back to all case studies