Outreach Dashboards

Reimagining the engagement tracking dashboards to serve a growing set of personas and use cases.

Team
1 Lead Designer, 1 Design ManagerCollaboration with PMs and engineers
Company & Role
Product Designer
Timeline
Nov '22 – Jan '23
Outreach Dashboards
Background / Innovaccer

In USA, a patient's data exists in an intelligent system of interconnected online portals.

Innovaccer builds the data and AI infrastructure that helps health systems (on any scale) unify that patient data and drive workflows to maintain the health, risks and satisfaction of their patient population.

Background / ‘Experience' & My role

I joined as a Product Designer under the ‘Experience’ vertical, Innovaccer's Patient Relationship Management module, working primarily on its campaign management tool, Outreach.

I supported and worked with a team of 1 Lead Designer and 1 Design Manager, collaborating regularly with other stakeholders (Product Managers, Engineering Managers and Engineers).

In Outreach, I worked to expand the product's use cases and features, catering to both customer requests and product vision.

This case study covers one of the first big projects I was handed: to completely re-imagine the dashboards for campaign management. It was exciting, and it let me test my design thinking in a real world application.

Background / Project kickoff

The dashboards had not been updated in a long time. Three things made the case for updating them.

  • 01

    Text and calls sit on an outdated PowerBI

    The only dashboard we have for text and calls is a very basic PowerBI one, outdated both in how it functions and in the metrics it carries.

  • 02

    No PowerBI at all for Email and Letters

    Both channels are read off a native dashboard the Outreach team built, which does not convey enough and holds no long term trend.

  • 03

    Pitched dashboards get tweaked

    The standard dashboards pitched to customers are tweaked to cater to each of them, which means items are missing from what we pitch in the first place.

Background / Previous snapshot
The existing PowerBI dashboard: a synopsis line, top-line counts, and a set of engagement charts filtered down the right-hand rail. This is a single and unique deployment.
Two tabs of the existing InConnect PowerBI dashboard side by side. The overview tab shows total patients, calls answered, engaged and not engaged counts and engagement success percentage across the top, with a synopsis panel and charts for engagement by gender, age, program trend and outreach by channel. The calls tab shows call-level KPIs across the top, a call synopsis, and charts for patient last call status, success by time of call, call duration density, volume of calls, success percentage by call attempts and engaged calls by gender and age.
Background / Root of the problem

Questioning the PRD, and diving deeper into the complete ecosystem, highlighted a problem that was never explicitly mentioned.

It was not only the dashboards that were outdated. The idea of who the dashboard was for, and what use they would make of it, was outdated too.

Restyling the dashboard would have made the existing problems harder to see.

We had to go back to the persona, the journey, and the questions.

User research / Re-establishing the user

Going back to the persona, the journey and the questions meant answering three things before anything could be drawn.

01Who is the user?

I started by refreshing the idea of the user we thought we were designing for. The sessions ran across 2 days of interviews with 6 users, covering Marketing / Clinical Managers, Marketing / Clinical Team Members and Customer Success Team Members.

02What is their journey?

Next, I mapped out the complete user journey, marking how the different personas interact with each other, and how each deployment works.

03What do they need?

We referred to the new personas and the user journey to draw Dashboard Stories for each persona, mapping the need to view the dashboard, where they draw their insights from, and what the action items are.

User research / Research snapshots

Research ran from an audit of every dashboard already in place, through the interviews, to the jobs and the stories they shook loose.

01 / 05 · Dashboard audit — Every dashboard in place at the time, broken down to the KPIs and fields each one captured.

A board mapping each persona — customer sponsor, service-line leader, marketing or clinician leader, campaign builder, CI director, CI product manager and CI engineer — to why they open the dashboard and the chain of questions they work through to answer it.
An audit board listing every dashboard already in place — the native text, text-assessment, email, email-assessment and letter campaign dashboards, and the CHI and CMICS Power BI program dashboards — with the KPIs and captured fields under each, annotated with open questions.
A user-interview board with a column per interviewee, each read down the same rows: about the persona, job to be done, job-related examples, how their job is measured, motivations, frustrations and challenges, skills, tools, collaborations and miscellaneous notes.
A journey map in four swimlanes — customer, CI / SE / CE, analytics and Outreach — tracing a dashboard from customer onboarding through the CI team, analytics deployment and review back to the customer, with a broader campaign journey beneath it and a row of action items users take away.
A jobs-to-be-done board: six numbered jobs — understanding KPIs, launching outreaches to drive revenue and close care gaps, studying dashboards for an overview, tracking drop-offs, setting goals against a benchmark, and taking action to close the gap — each with How-Might-We sticky notes and feasibility notes beside them.
A board mapping each persona — customer sponsor, service-line leader, marketing or clinician leader, campaign builder, CI director, CI product manager and CI engineer — to why they open the dashboard and the chain of questions they work through to answer it.
An audit board listing every dashboard already in place — the native text, text-assessment, email, email-assessment and letter campaign dashboards, and the CHI and CMICS Power BI program dashboards — with the KPIs and captured fields under each, annotated with open questions.
Ideation / KPI brainstorming

Working from the Dashboard Stories, I brainstormed the KPIs and metrics that answer the user's questions, and broadly classified them into 3 buckets. No visuals were decided yet: the KPIs were reviewed, trimmed and then re-iterated.

  1. 01

    Performance

    Conveys the usage and error rates of the product.

    Delivery failure reasons · Delivery health · Goals & outcomes · New patients

  2. 02

    Engagement

    Conveys how the patient population engaged with the campaigns.

    Click-through rates · Call pick rates · Total clicks on a link

  3. 03

    Demographical

    Conveys the reach of campaigns through the population.

    Total recipients · Age groups · Genders

Ideation / Lo-fi sketches

Now it was time to give some structure to the research. I explored visuals, looked at dashboards solving similar use-cases, and at what features PowerBI offered, which was one of the constraints we had to design within.

01 / 03 · Metric inventory — Every candidate KPI laid out and split into what is common to all channels and what belongs to one, tagged by whether it is a KPI, a by-day trend or supporting detail.

Four lo-fi prototypes of the dashboard, each shown across its overview, targets and journeys tabs — from blocked-out placeholder cards in the first, through tables and filter rails, to funnels, KPI rows and distribution charts in the later two, annotated with sticky notes and handwritten questions.
A metrics board titled Channels, split into common metrics — general, demographical and assessments — and channel-specific ones for email, voice, text and letter, with each candidate metric on a sticky tagged as a KPI, a by-day trend or supporting detail, annotated with open questions about labels, drop-offs and calibration.
An information-architecture board: the dashboard's user journey drawn as a flow into patient list, cohort creation and journey performance views, a dashboard exploration grid of grouped metric cards beneath it, and a wall of competitor dashboard screenshots annotated with sticky notes alongside.
Four lo-fi prototypes of the dashboard, each shown across its overview, targets and journeys tabs — from blocked-out placeholder cards in the first, through tables and filter rails, to funnels, KPI rows and distribution charts in the later two, annotated with sticky notes and handwritten questions.
A metrics board titled Channels, split into common metrics — general, demographical and assessments — and channel-specific ones for email, voice, text and letter, with each candidate metric on a sticky tagged as a KPI, a by-day trend or supporting detail, annotated with open questions about labels, drop-offs and calibration.
Final designs

The dashboards landed in two layers: a campaign-level view for the campaign in front of you, and unified dashboards reading across every campaign in the program, with the same metric buckets running through both.

01 / 05 · Campaign view — The campaign after a user launches one. The most important KPIs sit upfront, beside a list of the recipients who were in the cohort.

The unified population dashboard: a channel switcher for emails, SMS, phone calls and letters, KPI cards for recipients reached, delivered, open rate and average CTR, and charts for engagement by gender, age, region on a US map, and device, with the same filter rail down the right.
A campaign dashboard for an email campaign: recipients, delivered, open rate and click-through rate as stat cards down the left with a campaign-outcome donut beneath them, and a searchable, filterable recipient table beside them listing each recipient's status, status date and engagement score.
The trends view of an email campaign: a performance funnel from total to delivered, opened and clicked, a delivery-rate chart by day after launch with a hover card offering the recipient list, engagement over time, an insights panel calling out link preference and recipient opt-outs, a table of delivery-failure reasons and a drop-off donut.
The same trends layout for a call campaign: a performance funnel ending in call picked and engaged, delivery rate by day after launch, call answer duration against an engagement threshold, the insights panel, delivery-failure reasons and a drop-off donut split by call picked but not engaged, not picked, transferred and marked as spam.
The unified performance dashboard: eight KPI cards — campaigns launched, recipients reached, outreaches delivered, delivery rate, appointment completion, population engagement score, goals met and new patient acquisition — each against its three-month average, over charts of outreaches delivered by channel, reasons for failed outreaches, campaigns by channel and campaigns by creator, with a filter rail down the right.
The unified population dashboard: a channel switcher for emails, SMS, phone calls and letters, KPI cards for recipients reached, delivered, open rate and average CTR, and charts for engagement by gender, age, region on a US map, and device, with the same filter rail down the right.
A campaign dashboard for an email campaign: recipients, delivered, open rate and click-through rate as stat cards down the left with a campaign-outcome donut beneath them, and a searchable, filterable recipient table beside them listing each recipient's status, status date and engagement score.
Impact

The Mixpanel dashboards we tracked on were discontinued long ago, and with changing PMs, a solid tracker was never defined. However, here are some other patterns I observed.

015 more deployments, 2 tweaks

We deployed to 5 more customers after the dashboard work, and only 2 of them tweaked what they got: one to see engagement by demographics on the campaign-level dashboards, the other to add a few more engagement statuses to emails and voice calls.

02It held up in the room

Usability testing rendered good results, and it worked like a charm on demo calls. We had solved the transparency issues around drop-offs, while also letting users download a patient list for any filter and launch targeted campaigns off it, which were some of the key points raised in the user interviews.

03Untouched since handoff

The dashboards have not needed a change in the three years since handoff.