Mastercard’s view of consumer spending
SpendingPulse turns anonymized activity from Mastercard’s payment network into near real-time estimates of retail sales, by sector and region.
Economic IntelligenceResearch → AI synthesis → product strategy7 min readSpendingPulse
reimagined
How 30 interviews and 15 usability tests set a new direction for Mastercard’s economic-data product, built around the decisions its customers make.
- Problem
- SpendingPulse gave businesses powerful consumer-spending data, but it was built for exploring data, not acting on it. Customers exported it, rebuilt it in other tools and emailed Mastercard to ask what it meant. With a platform migration coming, Mastercard needed to know what to build next.
- My role
- Lead product designer & researcher. I ran all of the research myself, from the interviews and AI-assisted synthesis to the personas, jobs-to-be-done, journey map and usability tests, and presented the findings to Mastercard leadership. I worked with a product strategist, a UI designer and a project manager.
- What I did
- 30 interviews, synthesized with Claude and Granola and verified by hand, into 5 behavioral personas, 8 jobs-to-be-done and a 7-stage journey map. Then I designed prototype concepts and tested them in 15 usability sessions.
- Result
- A clear direction for the migration, shifting SpendingPulse from a reporting tool to a decision-support and monitoring platform, backed by research and a prototype that tested well with users.
- Timeline
- 14 weeks
- Interviews
- 30
- Usability tests
- 15
- Transcripts
- 595 pages
- Who uses it
- Executives to engineers
Tools
- Claude
- Granola
- Figma Make
- Figma
Methods
- Interviews
- AI-assisted synthesis
- Behavioral personas
- Jobs-to-be-done
- Journey mapping
- Prototyping
Context
The product and the stakes
What SpendingPulse is
Businesses planning around spending
Retailers, banks, travel and real estate companies use it to forecast, benchmark and plan. Some read a summary, others pipe the data into their own systems.
A migration, and a chance to rethink it
Mastercard was moving SpendingPulse to a new platform. Before rebuilding it, they needed to know how people really used it and what it should become.
At a glance
From 30 voices to one reframe
Fourteen weeks, narrowed to one direction
- Listen30interviews, external + internal
- Model5behavioral personas
- Frame8jobs-to-be-done
- Map7journey stages, 7 frictions
- Decide1product reframe
Stakeholder alignmentUser interviewsSynthesis → IA → prototypeUsability testing & next-gen vision
KickoffWk 4Wk 6Wk 14
Challenges
What made it hard
Designing for the migration and beyond
The same 14 weeks had to produce designs ready for a platform migration and a vision for the next generation of the product.
How I handled it
I split the work in two. Prototype screens the team could build for the migration, and six principles to guide what comes after it.
Executives to data engineers
Some users never log in and only want a summary. Others never see the interface because they pull data through an API.
How I handled it
I grouped people by how they use the data, not by job title. Five behavioral personas covered everyone, from the executive to the engineer.
595 pages, no time to waste
Thirty interviews produced hundreds of pages of transcripts, with an on-site leadership readout on the calendar.
How I handled it
I used AI to sort the transcripts into themes, then checked every theme against what people actually said before sharing it.
01 · Listen
30 interviews
What 30 people told us
I studied the whole workflow around SpendingPulse, not just clicks inside it. I looked at what decisions people owned, where data went after they found it, and how it reached the person who acted on it. Nobody wanted more data. They wanted to know what it meant.
19
External users, from C-suite, heads of cards & payments, FP&A, operations analysts, marketing & growth leaders
11
Internal Mastercard stakeholders across product, sales and economics
SectorsFinancial servicesRetailConsumer goodsTravelCommercial real estate
Need 01
Narrative context, the why
Users could see that spending moved. They couldn’t see why, or whether it mattered to them.
Designed → AI summaries and expert commentary beside the data
“The thing that is always murky is the why… Traffic right now is up. Is that because of new regulations? Is it a competitor who’s stumbling in the market? Sometimes those are far more buried answers than what I would hope.”
Need 02
Monitoring & alerts
People checked the same cuts every week to spot anomalies. When they found one, their only path was to email Mastercard.
Designed → Unusual-spending alerts on the dashboard
“We send emails out to Mastercard. If I see some sort of anomalies, I send an email asking them how to interpret that.”
Need 03
Forward-looking tools
Users wanted help moving from what happened to what’s next, like forecasting, scenarios, stress-testing plans.
Designed → Forecast overlays in analysis
“We use that information to try to predict the future… planning and forecasting and major headwinds that would come our way.”
Need 04
Structured exports
Data went into Excel, BI tools and decks. Exports needed clean labels and metadata, or people just took screenshots.
Designed → Saved views and labeled, structured exports
“I’ll export a CSV, or if it’s something on a dashboard and I just need something really quick, I’ll just do a screenshot, quite literally.”
Need 05
Clear definitions
Without clear category and geography definitions, users couldn’t reconcile SpendingPulse with their own dashboards.
Designed → Definitions one click from any chart
“Whenever I wanted to figure out what is the definition… I usually open up one of the presentation decks or PDFs from the sales process… that’s where I’ve been getting it.”
02 · Synthesize
AI in the research loop
AI did the sorting. I did the judging.
I built generative AI into my research workflow to move faster through 595 pages of transcripts, but nothing reached the team or Mastercard leadership until I had checked it against what people actually said.
- 01 · CaptureGranolaEvery interview recorded, transcribed and summarized as it happened.
- 02 · ClusterClaudeTranscripts organized and grouped into candidate themes and patterns across participants.
- 03 · VerifyMeEach theme traced back to source quotes. What didn’t hold up was cut.
- 04 · InterpretMeTime saved went into comparing audiences and tying findings to business decisions.
In595pages of transcripts
Out5behavioral personas
8jobs-to-be-done
7journey stages
Synthesis took about a week instead of three. Every finding traceable to what participants actually said.
03 · Model
Personas × jobs
Job titles don’t predict behavior
Some C-suite leaders logged in weekly and ran their own reports; others never touched a tool. Some heads of cards behaved like analysts. So instead of designing for titles, I modeled five behavioral personas and the eight jobs they hire SpendingPulse to do. One person can be several personas, depending on the task.
Persona 1
Direct Data Consumer
Explore, diagnose, benchmark.
Goes hands-on. Filters, slices, exports and builds their own analysis. Measures the product by speed, clarity and trust in the data.
- How they access data
- Directly in the UI, then exports
Persona 2
Insight Receiver
Fast clarity, no digging.
Wants curated takeaways and answers they can act on. Rarely logs in; depends on clarity and context, not tools.
- How they access data
- Narrative only, through decks, screenshots and email
Persona 3
Data Integrator
Clean, structured feeds.
Pipes Mastercard data into pipelines, data lakes and BI tools. Format, consistency and metadata matter far more than UI.
- How they access data
- APIs, feeds, BI tools and data lakes
Persona 4
Strategic Storyteller
Build narratives that influence.
Turns data into stories for leaders, boards and clients. Focused on accuracy, framing and change over time.
- How they access data
- Shared decks, exports, rebuilt charts
Persona 5
Monitoring & Alerts User
Detect change early, act quickly.
Tracks weekly and monthly shifts, anomalies and leading indicators, and needs to know fast what moved and why.
- How they access data
- UI or BI dashboards with SpendingPulse embedded
One real participant
Three personas, one person
Same title, layered behaviors.
A strategy analytics lead who acts as a Direct Data Consumer, a Data Integrator and a Monitoring user, sometimes in the same afternoon.
- How they access data
- UI weekly, CSV into Tableau, emails to Mastercard
- 1Forecasting, budgeting & planningPrimary
- 2Benchmarking & performance context
- 3Executive & stakeholder storytelling
- 4Spotting emerging opportunities & risks
- 5Evaluating campaigns, pricing & initiatives
- 6Cross-functional alignment
- 7Monitoring & alerts for meaningful change
- 8Validating internal KPIs against external reality
Spotlight · a weekly power user“I bring in the sales and the previous year’s sales, and the rest of the calculations I’ll do on my own.”
A strategy analytics lead who exports every week, rebuilds monthly and quarterly totals in Tableau, and emails Mastercard to ask what is driving the forecast. One person, three personas.
04 · Map
Current-state journey
Most of the work happens after the export
I mapped all five personas from the moment a question comes up to the decision it informs. SpendingPulse only owned the first three stages. The value, and the friction, lived in the four that came after.
- 01 · In SP
Trigger
Requests arrive reactive and under-specified, putting pressure on analysts.
- 02 · In SP
Access
Entry points vary by persona, fragmenting the experience.
- 03 · In SP
Explore
Comparing scenarios means repeated manual configuration.
- 04 · Outside
Export
Definitions, context and assumptions are lost once data leaves.
- 05 · Outside
Interpret
Users reconcile external and internal data by hand.
- 06 · Outside
Communicate
Insights aren’t presentation-ready; charts get rebuilt.
- 07 · Outside
Decide
Confidence depends on triangulating outside the platform.
Access 01Direct in the UIAnalysts who log in, filter and export
Access 02Through exportsRepackaged into Excel, decks and models
Access 03BI tools & data lakesIngested centrally; few ever see the UI
Access 04Narrative onlyScreenshots and briefings made by someone else
05 · Reframe
The strategic shift
From a reporting tool to a decision-support and monitoring platform.
This was the line I put in front of Mastercard leadership in the research readout. It meant serving two modes at once, so every persona gets value whether they log in or not.
Explore faster
Saved views, flexible cuts, in-platform aggregates and clean, labeled exports for Direct Data Consumers and Integrators.
Know what moved, and why
Alerts, anomaly detection, AI summaries and expert context for Insight Receivers, Storytellers and Monitoring users.
06 · Design
Insight → decision
Every screen traces back to a finding
The prototype kept the analytical depth power users rely on, and added faster paths to meaning for everyone else. I designed the concepts and flows, and the UI designer built the high-fidelity screens.
- 01 · ExportDownloads a CSV from SpendingPulse, again
- 02 · RebuildRecreates monthly and quarterly totals in Tableau
- 03 · AskEmails Mastercard to ask what’s driving the forecast
- 04 · WaitPresents to the CFO without the why
- 01 · OpenLands on a saved view on her dashboard
- 02 · AggregateToggles aggregated data in the platform
- 03 · UnderstandReads forecast drivers and the AI summary beside the chart
- 04 · PresentExports a labeled view, definitions attached
Personalized dashboard
Research insight
People had different priorities and recurring tasks, and many never wanted to configure an analysis just to see what changed.
Design decision
A home built around the user, with tailored KPI cards, recently saved views, sector trends, a key takeaway, and an alert when spending looks unusual.
Serves Insight Receivers · Monitoring users
Flexible analysis
Research insight
Power users needed granular cuts and repeat analysis. One rebuilt monthly and quarterly totals in Tableau every single week.
Design decision
Filter by sector, geography, time, channel and demographics; switch between chart, table and map; save views; aggregate in-platform; read an AI summary or listen to an audio one beside the forecast.
Serves Direct Data Consumers · Strategic Storytellers
Market insights hub
Research insight
Data meant more next to expert interpretation. The question users kept asking was why, not what.
Design decision
Economist commentary, chart of the week, forecasts and what to watch in one discoverable hub, also surfaced as related resources beside analysis.
Serves Insight Receivers · Strategic Storytellers
Definitions in reach
Research insight
A chief data officer and power user had never opened the definitions section. He dug through old sales decks instead.
Design decision
Plain-language sector and regional definitions, one click from any chart, so context survives the export.
Serves Data Integrators · everyone who exports
Role & interest settings
Research insight
Same title, different needs. Behavior, not role, predicted what someone wanted to see first.
Design decision
Users set role, interests and defaults, which tune what the dashboard and alerts surface for them.
Serves All five personas
Usability testing
10 prospects · 3 current users · 2 Mastercard stakeholders15 sessions confirmed the direction
I tested the clickable prototype in 15 moderated, hour-long sessions. The redesign tested well. People moved between the dashboard, insights and analysis with confidence, nearly everyone understood saved views right away, and many read the AI summary first to check their own interpretation.
“Here’s my top view, here’s getting more, and here’s down in the weeds. It seems intuitive to me.”Prospect, usability session
What testing surfaced
- KPIs needed contextPeople liked the numbers but couldn’t tell the timeframe or comparison behind them.
- The anomaly alert read as an errorIts warning styling made some people think Mastercard’s data was broken.
- One sector wasn’t enoughSome users needed to compare several sectors side by side.
What I recommended
- Label every KPIAdd the metric, comparison and timeframe to each card.
- Treat anomalies as insightsRestyle the alert and explain what was detected without leaving the page.
- Make customization everywhereExtend “Customize my KPIs” to every sector module, with multi-sector views.
07 · Impact
Validated with users
Tested, validated and ready to build
Outcome
Usability testing confirmed the direction. Across 15 sessions, the redesign tested well, and prospects saw it as a daily starting point.
It also showed business value. A Mastercard participant noted that the prototype answered questions their delivery team handles by hand today, and internal users saw new paths like the data feed offer as a strong upsell opportunity.
- 01Five behavioral personas that replace job titles as the way Mastercard understands its audienceStrategy
- 02Eight jobs-to-be-done, each backed by evidence and open questions, to prioritize the roadmapStrategy
- 03A current-state journey map showing where value leaks once data leaves the platformStrategy
- 04A tested prototype where every design decision traces back to a research findingProduct
- 05A next-gen vision built on six principles, from proactive alerts to AI-assisted explorationProduct
- 06A research readout that aligned Mastercard leadership on the decision-support directionLeadership
The next-gen vision, in six principles
P1PersonalizationShaped by role, interests and recurring tasks
P2Proactive intelligenceAlerts and anomaly detection that find you
P3AI-assisted explorationAsk questions of the data in plain language
P4Connected workflowsInto BI tools, reporting and data pipelines
P5Narrative intelligenceEvidence plus interpretation, ready to present
P6TransparencyDefinitions, method and provenance in reach
08 · Reflect
Looking back
Design for the decision, not the interface
The biggest insight was how much of the experience happened outside the product we were redesigning. I now map downstream workflows from day one.
Post-launch impact wasn’t in scope. Next time I’d set baselines early (time to insight, weekly exports, “what’s driving this?” emails) so results can be proven.
AI sped up my research and shaped the product. In both, trust was the work. Testing proved it, since people valued AI summaries most when they matched the exact view on screen.
What I took away
The value of a data product isn’t the information it provides. It’s how well that information helps someone make a better decision.