Reported gains in self-service use
Bulk-action engagement rose from approximately 34% to 57%. Weekly posting-goal adoption rose from 12% in week one to 34% in week eight.1
I helped Mobal evolve from an agency-built listing tool into self-service software, extend the platform into restaurant websites and introduce AI workflows customers could authorise.
Project snapshot
Problem
A services-led listing product had outgrown its original model. More capabilities were making routine tasks harder to find.
Solution
Establish Mobal as the source of business information, simplify self-service workflows, extend that foundation into restaurant websites and support explicitly authorised AI delegation.
Impact
Reported bulk-action engagement rose from approximately 34% to 57%. Weekly posting-goal adoption rose from 12% in week one to 34% in week eight.
- Role
- Product design · Discovery through delivery, with leadership participation
- Timeline
- January 2023–2026
- Contribution
- Product discovery, UX strategy, research, interaction design, prototyping, design systems and customer validation
- Partners
- Founders, product, engineering, Sales and Customer Success
What changed
Bulk actions gained traction.Engagement grew from approximately 34% to 57% among the same group of small-business customers.~34% → ~57%
The same group of businesses managing 1–10 locations, measured over four weeks before and four weeks after the changes. We used the same definitions for weekly active users and bulk actions. Other product changes happened during this period too.
More customers set posting goals.Adoption rose from 12% in week one to 34% in week eight.12% → 34%
Weekly posting-goal adoption at two measured points. The weeks in between are not shown.
Publishing still needed attention.Only 18% of people who clicked a reminder went on to create a post.18%
This result covers people who clicked a WhatsApp reminder. Follow-up research pointed to difficulties with ideas, visuals and preparing content.
Product results cover different groups and measurement periods.
The wider business context
Company growth
Revenue figures await verification of ARR versus turnover, the baseline and reporting period. Teams across the company contributed to the broader product transformation.
01 / The product had outgrown its original model
I joined Mobal in January 2023, taking responsibility for product design and participating in leadership. The company was moving from an agency-built tool for managing Google Business Profiles toward self-service SaaS.
I worked with the founders, product and engineering on the platform’s direction, then carried research, interaction design, visual design and implementation reviews through delivery. I was the second designer to join and later worked largely as the sole designer, with occasional support.
Businesses needed to maintain hours, holiday exceptions, posts, reviews, media and contact details across locations and publishing channels. My work spanned the self-service foundation, restaurant websites, design systems and AI delegation.
The experience still followed Google’s data model, and customers relied on Mobal’s team for routine work. I worked with the CEO, engineering, sales and Customer Success to investigate limitations affecting onboarding, operations and growth.
We chose to make Mobal the source of business information, with Google and other connected platforms as publishing destinations. That supported self-service onboarding and multi-location management. As we added capabilities, though, everyday tasks became harder to find.
Restaurant website case study
Keep routine tasks easy as the product grows
We needed to add capabilities while keeping routine tasks easy to find and complete.
02 / Making important actions discoverable
Bring common actions into profile management
Bulk operations sat one level down in an increasingly crowded action catalog. Using PostHog data, session recordings, customer feedback and recurring Customer Success conversations, I examined the most-used actions and those people overlooked despite their consequences.
I proposed showing Posts & Stories, Media, Special Hours and Business Hours in profile management, with less frequent operations under More bulk actions. We discussed the order and how many actions to show before settling on that grouping.
Product management proposed separate Post, Event and Offer entry points. I argued for one publishing entry point, with content-type selection inside the composer. People could start with the task they had in mind, then choose the format and destination.
People edited regular hours often, but could forget special hours. Incorrect holiday hours affected customer visits, so we made both actions accessible and added suggested holiday dates.
Measuring the improvement
Bulk-action engagement · Before / after
+23 percentage points in bulk-action usage
- Four weeks before
- Four weeks after
Surfacing important actions supported stronger engagement with bulk workflows.
Weekly active users in the same cohort of businesses managing 1–10 locations. Consistent activity and bulk-action definitions. Observational comparison; concurrent product changes are not isolated.
03 / Sequencing the redesign around agentic workflows
The profile editor had become a lengthy collection of business fields, integrations, schedules and publishing information. I explored three directions for simplifying the workspace.
Three directions explored before approval
A — A focused business home
An overview for businesses with roughly one to five locations, showing business status, relevant tasks and essential information.
B — Contextual row actions
Hover controls reduced navigation, but customers often wanted to inspect current information before editing. Less useful for multi-location work.
C — A workspace with more features
Product leadership’s direction exposed more features and broader navigation, partly supporting pricing-tier experiments. More features competed for attention.
We approved a direction combining clearer priorities, contextual actions and better visibility into background activity. Implementation was deferred until the agentic workflows were complete, allowing us to establish how customers would delegate work before changing the wider workspace.
Approved direction · Not yet shipped
Approved profile direction · Not yet shipped
Quick actions after profile setup
Approved direction, not yet shipped. Quick actions replace the setup checklist once initial configuration is complete.
Priority messaging within the profile workspace
Approved direction, not yet shipped. A contextual message surfaces information that needs attention above the workspace.
View the original profile designs in Figma.
Show the work Mobal does in the background
Company retention · Context
Reaching 100% NRR was not the same as sustaining it
100% NRR = recurring revenue retained from existing customers, including expansion and losses.
- Earlier platform
Starting context
- Mid-2025 peak
Above the 100% threshold
- Later level
Below the 100% threshold
Churn after six-month pilots kept retention under pressure and informed the focus on making ongoing value visible.
Selected company milestones, not a continuous time series. NRR fluctuated after its peak. These retained figures await reconciliation with the CV; they are not verified for publication. Company performance reflects work across teams; the approved, unshipped profile redesign is not presented as having improved NRR.
We suspected some customers could not see the work Mobal did while they were away. The platform maintained information and ran background operations, but customers had little visibility into those changes.
I explored priority messaging, onboarding checklists, contextual action cards and activity reporting to communicate what had changed or been handled. I questioned time-saving estimates we could not support. Customer-facing teams agreed to look for data and customer accounts we could use.
Making background activity visible · Exploration
Activity-view exploration
Unshipped exploration. Activity totals and time-saving estimates inside this mockup are illustrative, not validated outcomes.
04 / From navigating actions to delegating work
Let owners ask for the work they need
Owners wanted accurate information and completed tasks. We introduced an in-app AI agent, then extended it to WhatsApp. They could ask it to check holiday hours, audit business information, research competitors or draft posts.
During demos, users with less technical experience asked whether an action had happened, was underway or still needed approval. I designed the confirmations and status messages to distinguish what the agent knew, what it proposed changing and what it had completed.
Deep dive: changing holiday hours through WhatsApp
Changing holiday hours through WhatsApp
Retrieve the current state
The user asks about upcoming special hours. The agent returns existing schedules, including closures and modified opening hours.
Interpret two requested changes
The user asks to close January 6, 2027, and open December 26, 2026, from 18:00 to 20:00 in one message.
Clarify scope and request authorization
The agent identifies the affected business and shows proposed changes against existing values. A reply containing only the business name is not treated as authorization: the agent asks for explicit confirmation.
Execute and communicate completion
After confirmation, the agent reports the saved changes, restates the resulting hours and links to the affected profile. The captured evidence confirms the approval exchange and reported completion in Mobal, not publication across every external directory.
WhatsApp: from request to confirmed change
Retrieve the current state
Excerpt of the existing schedule: the two dates relevant to this request.
Request two changes
One natural-language request covers both dates.
Clarify the affected business
“Harbour” identifies the business; it does not authorize execution.
Require explicit authorization
The agent asks again. The user explicitly replies “Confirm”.
Report saved changes
The agent reports saved hours and provides a profile link; external publication is not independently verified.
Reconstructed from the supplied WhatsApp screenshots. Message order, dates and approval behavior are preserved; formatting is normalized. The retrieval panel is an excerpt. Phone number and internal profile links are omitted. These are reconstructions, not original screenshots.
Reviewing larger changes
A conversational summary works for a small number of changes. Larger batches, conflicting schedules or multiple locations could need a more detailed review. Saving in Mobal also does not guarantee immediate publication to external directories. We needed to distinguish a saved change from a published one and give users a way to recover when an update failed.
05 / Reminders brought people back, but few created a post
Customer Success encouraged businesses to publish, but reminders did not consistently lead to completed posts. In research, many restaurant customers described planning their work by the week.
We introduced weekly posting goals, surfaced progress in profile management and connected the experience to reminders.
Weekly posting goals · Adoption
+22 percentage points over eight weeks
- Week one
- Week eight
Weekly posting-goal adoption. Two measured endpoints; intermediate weeks are not shown.
WhatsApp reminders · From click to creation
A reminder brought users in. Preparing content remained the barrier.
- Reminder click rate
Click rate for WhatsApp reminders.
- Post creation after clicking
Of those who clicked, this share created a post.
Follow-up research pointed to ideas, visuals and producing content as the next problem to solve.
18% refers to people who clicked a reminder. The two stages describe different groups; card sizes do not represent group size.
Helping users decide what to post
I designed a flow for turning customer reviews into editable posts. Businesses could start with feedback they already had, then prepare the artwork and caption within the agentic workflow.
The direction was approved but had not shipped, so its effect on publishing conversion remains unmeasured.
Approved direction · Not yet shipped
Customer review → editable post · Approved, not shipped
Review artwork with styling controls
Review-to-post design: customer feedback becomes post artwork with styling and caption controls. Approved, not shipped.
Editable review-post composer
The composer carries review artwork into channel selection and editing, with layout controls and an optional AI-generation control. Approved, not shipped.
View the review-to-post designs in Figma.
06 / Results and what I took forward
Bulk-action engagement, posting-goal adoption and the reminder funnel showed progress at different points in the experience. They cover different groups and measurement periods and reflect collaboration across product, engineering, Customer Success and commercial teams.
The existing account records a company NRR peak of 102% in mid-2025 and difficulty sustaining performance above 100%. These company-level figures reflect work across teams and await reconciliation with the CV.
What customers said
In public G2 reviews, customers described managing multiple profiles, working across channels from one place and using AI-assisted review management. Their comments concern the wider platform; they do not establish the newer conversational agent’s impact.
G2 customer feedback
- Overall product rating
- Published reviews
G2, accessed 15 October 2026. Product-level rating across all review dates.
Easier to use
Mobal.io is extremely user-friendly
Simpler workflows
all from one place
Less manual work
Saves a lot of time by managing multiple profiles at the same time
Read the customer reviews on G2. We recorded the rating and review count on 15 October 2026. These accounts are qualitative product evidence, not a controlled before/after usability study.
Reflections
Shared patterns helped us reuse decisions across the growing platform. Read the Mobal design system case study.
We deferred the approved profile redesign until the agentic workflows were complete. That gave us time to understand how customers would delegate tasks before changing the workspace around them.
The WhatsApp work made me more precise about status. Retrieving information, recommending a change, getting approval, saving it and publishing it each need a clear message.
Footnotes
-
The existing case reports a same-cohort, four-week before/after comparison for bulk actions and a separate eight-week posting-goal adoption series. These are observational results across different groups and periods, not an isolated design effect. Metric definitions and source records remain subject to verification; detailed caveats follow below. ↩
