Sales Ecosystem 2.0
Research Synthesis

What is Sales Ecosystem 2.0?

A redesigned sales toolkit for Talabat, foodpanda & Yemeksepeti agents — and why we tested it before rollout.

Sales Ecosystem 2.0 is the next version of the sales platform that field, tele, and inside-sales agents use every day to manage leads, move vendors through the pipeline, and communicate with prospects — built around three core areas: Dashboard, Pipeline, and Communication Hub (WhatsApp, Email, Calls, and AI-assisted calling). This research was commissioned to usability-test the redesign with real agents and managers before committing to a global rollout — checking whether the new layout, terminology, and AI concepts are actually understood in practice, and surfacing blockers and market-specific gaps while they are still cheap to fix.

What did we learn from the results?

What we believed would feel natural to agents turned out to work very differently in practice.

Usability validation before a global rollout

60 hours across 7 markets — real agents, real friction, before a global rollout.

Study Setup

Moderated Usability Testing

Think-aloud protocol · Mar 26 – Apr 8, 2026

Markets by Entity

🇦🇪UAE

🇪🇬EGY

🇧🇭BHR

🇲🇾MYS

🇵🇰PAK

🇵🇭PHL

🇹🇷TUR

Research Objectives

01

Comprehension & discoverability of Dashboard, Pipeline, and Communication Hub

02

Identify blockers, friction points, and terminology confusion per market

03

Validate AI calling and objection-handling concept across agent roles

04

Understand role-specific needs — agents, team leads, managers, Onboarding agents

05

Surface market-specific configuration requirements before rollout

Marvin

Session recordings, highlight reels & insight tagging

140 clips tagged

Gemini AI

AI-generated session notes, one per interview

14 session notes

Figma

Interactive prototype tested across all 7 markets

1 prototype · 15 screens

Who sat in the room with us

Real agents, team leads, managers, and specialists — the people who will use this tool every day.

01

Hany Abdelmalak

Sr. Sales Acquisition

🇦🇪 Talabat UAE Mar 26, 2026

02

Ahmed Saqr

Sr. Sales Acquisition Lead

🇪🇬 Talabat Egypt Mar 28, 2026

03

Lee Yee Mei + Victor Lim

Field Sales + Team Lead

🇲🇾 Foodpanda Malaysia Mar 29, 2026

04

Zeeshan Zahid

Assistant Manager, Telesales

🇵🇰 Foodpanda Pakistan Mar 30, 2026

05

Joshua Fawzi Ackroyd

Sr. Sales Executive

🇧🇭 Talabat Bahrain Mar 30, 2026

06

Aisyah Ummul Mukminin

Coordinator, Telesales

🇲🇾 Foodpanda Malaysia Apr 1, 2026

07

Abdulrahman Mohamed

Sr. Sales Acquisition Executive

🇪🇬 Talabat Egypt Apr 1, 2026

08

David Cabal

Manager, Inside Sales

🇵🇭 Foodpanda Philippines Apr 2, 2026

09

Muhammad Ali Qureshi

Territory Sales Manager

🇵🇰 Foodpanda Pakistan Apr 2, 2026

10

Havva Hazal Demir

Sales Executive, Acquisition & Incubation

🇹🇷 Yemeksepeti Turkey Apr 2, 2026

11

Sudenaz Uzunkaya

Sales Specialist, Acquisition & Incubation

🇹🇷 Yemeksepeti Turkey Apr 2, 2026

12

Juan Raphael Miguel

Specialist, Inside Sales

🇵🇭 Foodpanda Philippines Apr 6, 2026

13

Ng Sze Yah (Jamie)

Manager, Field Sales

🇲🇾 Foodpanda Malaysia Apr 8, 2026

14

Syafiqah Nazri

Manager, Field Sales

🇲🇾 Foodpanda Malaysia Apr 8, 2026

Sessions in action · Live usability interviews

Dashboard

David Cabal · 🇵🇭 Foodpanda PH

Manager, Inside Sales · 4.5 / 5

Dashboard

Abdulrahman Mohamed · 🇪🇬 Talabat Egypt

Acquisition Executive · 5 / 5

Dashboard

Ahmed Saqr · 🇪🇬 Talabat Egypt

Sr. Sales Acquisition Lead · 5 / 5

Dashboard

Zeeshan Zahid · 🇵🇰 Foodpanda PK

Asst. Manager, Telesales

Dashboard

Sudenaz Uzunkaya · 🇹🇷 Yemeksepeti TR

Sales Specialist, Acquisition

Comm Hub

Joshua Fawzi · 🇧🇭 Talabat Bahrain

Sr. Sales Executive

Moderator

Farzam Anjum · Delivery Hero

Product Design · Research lead

What Did We Test?

A Figma prototype of Sales Ecosystem 2.0 — a redesigned Salesforce-based tool for sales acquisition agents. Each of the 14 sessions covered 7 areas in a consistent task-based structure. Participants were given realistic scenarios and asked to navigate freely.

Findings — Mapped to the Screen

Every finding, where it happened

14 sessions across 7 markets. Each finding is pinned to the exact prototype screen where the observation occurred, with the evidence, the verbatim, and the recommended action.

Severity system

Every observation rated so the team knows what to fix before launch, what to monitor, and what to double down on.

4

Blocker

The issue prevents the user from accomplishing the task

3

Major

It causes frustration and/or delay

2

Minor

It has a minor effect on task performance

1

Suggestion

It's a suggestion from the participant

0

No Interference

It's what we expected the participant to do

−1

Delight

Positive feedback from the participant

Screen Area 1 of 4

Dashboard

Blocker Major Delight

01

Delight D5

Recommendations strip and hot leads — immediately understood and acted on

11 of 14 sessions "Justifies adoption alone" All markets

The Recommendations strip surfaces urgent priorities — Hot Leads, Docs to Review, Rejected Cases, Onboarding Failed, Zombie Vendors — as inline action cards. 11 of 14 agents said this feature alone justifies daily adoption. Inside sales managing 200+ accounts found it especially powerful: the system surfaces exactly where to start instead of requiring agents to manually scan the pipeline.

"This is the first thing I'd look at every morning. 20 hot leads — okay, I know exactly where to start. "

3

4

1

2

1

Outreach tracker — "3 / 15 outreaches · Behind pace"

When asked to interpret their daily progress from the dashboard, 9 of 14 agents paused at the Outreach tracker and asked what "Outreach" meant. Once briefly explained, they immediately understood the concept, the "3 / 15" counter, and the progress graph on the right — all without further prompting. The data structure is intuitive; the label is the only barrier. Team leads also noted this view is individual only and asked for a team-wide version showing each agent's pace. Major

2

Hot Leads — scoring criteria not visible

When asked what makes a lead "Hot," 9 of 14 agents could not identify the logic behind the classification. Agents acted on the category without questioning it in the moment — but in debrief, all nine said they would want to know why a lead qualifies as Hot before trusting the prioritisation long-term. A recurring suggestion across markets: surface Hot, Warm, and Cold lead tiers so agents can calibrate their outreach strategy rather than treating all Hot Leads as equal. Suggestion

3

Recommendations strip — "50 urgent" across 5 categories

When asked where they would start their working day if this dashboard were live, agents pointed directly to the Recommendations strip without hesitation. The 5 categories — Hot Leads, Docs to Review, Rejected Cases, Onboarding Failed, Zombie Vendors — were understood immediately as a prioritised action queue. No explanation of the categories was needed in any session. No Interference

4

Hot Leads CTA — "Reach out now" inline action

When asked to action their hot leads directly from the dashboard without navigating away, 12 of 14 agents successfully tapped "Reach out now" on the first attempt — no guidance given. The inline CTA kept agents on the dashboard throughout the task. 20 Hot Leads, 18 Docs to Review, 12 Rejected Cases — all reachable in a single tap from a single row. Delight

3 Design Decision Protect this strip above the fold on all screen sizes — it is the product's primary adoption hook. Ensure the strip is the first thing agents see on dashboard entry, every time.

2 Design Decision Surface the Hot Leads scoring criteria — add a tooltip or sub-label on the "Hot Leads" header explaining what qualifies a lead as Hot. Agents need to trust the classification before acting on it at scale. V1.1: introduce Hot / Warm / Cold lead tiers so agents can segment their outreach strategy rather than treating all Hot Leads as a flat, undifferentiated list.

Participant Recordings · Marvin

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition · 10 yrs tenure

Delight

Hany confirmed the new dashboard would eliminate his personal Google Sheets tracker entirely — giving him visibility into who's been contacted, follow-up priorities, and new leads from a single view.

"It will give you also visibility to us... to let us know who we just reached out to, who we need to follow up with."

View in Marvin →

Muhammad Hafizuawan · 🇲🇾 Foodpanda MY Supervisor, Telesales

Delight

Hafizuawan explained his team's prioritisation logic: highest priority goes to vendors almost at activation ("hot leads") and fresh incoming registrations — both now surfaced automatically by the dashboard, replacing manual scanning.

"High priority is given to 'hot leads,' which are only a few steps from going live, as these are crucial for meeting activation targets."

View in Marvin →

Finding 03

Estimated payout is always visible — privacy and compliance risk

5 of 14 sessions PDPA (Personal Data Protection Act) risk · Malaysia Shared office environments

Estimated monthly payout (e.g. AED 55K–70K) is displayed in large, always-on text on the main dashboard with no way to hide it. In shared offices across UAE, Malaysia, and Egypt, this exposes agent earnings to anyone passing by. Malaysia raised PDPA concerns (Personal Data Protection Act — displaying an employee's earnings without a privacy control may violate Malaysia's data protection law). Turkey noted the figure rarely matches actuals — visible to peers, it erodes trust.

"I don't want my colleagues to see my commission. This screen would be open all day — that's a problem." — Malaysia, Field Sales

3

4

1

2

1

KPI Breakdown — 5 fixed metrics, same for all roles

When asked to review their performance on the dashboard, agents noted that the 5 fixed KPIs did not reflect their actual role. Telesales agents needed calls and meetings front and centre — metrics like self-activated and quality score were irrelevant to their daily targets. Managers needed team-level totals and per-agent breakdowns to run their team effectively. Outbound sales reps needed territory and visit-based metrics to track their pipeline. Every persona needed a different default set — one aligned to their goals, not a one-size-fits-all view. Blocker

2

"Stalled Deals (Past EDC)" — terminology too technical

When asked to interpret the Pipeline Health panel on the dashboard, almost every agent paused at "Stalled Deals (Past EDC)" and asked what it meant. EDC (Expected Deal Close date) is an internal CRM term with no visible definition. Once briefly explained, agents grasped the concept immediately — the data is useful, the label is the only barrier. Plain language alternatives like "Overdue Deals" or "Deals Past Due Date" were universally preferred across all markets. Blocker

3

EST. PAYOUT: AED 55K – 70K — always-on, large text

When agents were asked to explore the dashboard for the first time, agents in Malaysia, UAE, and Egypt spontaneously flagged the always-visible payout figure as a privacy concern — no task prompt required. There is no way to mask the figure, and agents had no workaround even when aware of the PDPA (Personal Data Protection Act) risk. The concern surfaced on first screen view, every time. Major

4

Grade filters (A+ / A / B / C) on Pipeline Snapshot

When asked to prioritise which leads to contact first, agents who discovered the A+ / A / B / C grade filter on the Pipeline Snapshot applied it immediately and correctly — zero explanation needed. A single tap narrowed the pipeline to their highest-quality leads. The filter was intuitive enough that no agent who found it required any clarification. Delight

3 Design Decision Add a toggle — mask payout as AED •••• by default. Tap to reveal. Persist per session so agents don't re-toggle after every login. This is standard in banking apps and fully resolves the privacy and PDPA (Personal Data Protection Act) risk in one change.

1 Design Decision Ship role-aware KPI views from launch — Telesales agents see calls and meetings front and centre; Managers see team-level totals and per-agent breakdowns; Outbound sales reps see territory and visit-based metrics. Phase 2: allow each entity to configure a market-specific KPI set so no market's metrics are misrepresented.

2 Design Decision Rename to 'Overdue Deals' across all dashboard surfaces. Add a permanent sub-label on first view: 'deals past their expected close date'. Eliminate all internal CRM abbreviations (EDC, QC, SSU) from agent-facing copy — plain language only. One copy change, zero engineering cost.

Participant Recordings · Marvin

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

Major

Victor raised PDPA (Personal Data Protection Act) and data sensitivity concerns about the payout being visible by default on the main dashboard — in a shared office environment, this exposes sensitive financial data to colleagues.

"It could be no PDPA or it's a bit sensitive in terms of showing that rough payout to someone else openly."

View in Marvin →

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

Major

Victor proposed showing an estimated range rather than an exact figure — preserving the motivational intent of the incentive tracker while mitigating privacy risk for agents who share screens in the office.

"An estimated estimate in a payout would be a better option than the actual amount that you are getting."

View in Marvin →

Finding 04

Dashboard views and labels don't match agent roles or natural language

7 of 14 sessions Field, tele-sales, SSU, managers 5 markets affected

Agents were divided on the Leaderboard — some valued seeing how teammates were performing and tracking their own standing, while others found the competitive framing unnecessary and deprioritised it entirely. Team leads had a distinct need: they wanted to identify underperforming members, not just view a ranked list. The current individual-rank view serves personal motivation but does not support any management accountability use case. Separately, across every task involving the pipeline, agents naturally said "restaurant" rather than "vendor" — every column header requires a silent mental translation. One pattern that worked: agents who found the Vendors Needing Follow-up section acted on it immediately and correctly, without any instruction.

"I need to see my team's numbers, not just mine. I'm a team lead — I don't care about my own go-live count, I care about the whole team." — UAE, Team Lead

1

2

3

1

Team Leaderboard — individual rankings only

Agents were split on the Leaderboard. Some found it useful for seeing how teammates were performing and tracking their own standing; others felt the competitive angle was unnecessary and ranked it as least important to their daily work. Team leads had a different need entirely — they wanted to identify underperforming members, not just see a ranked list. The current view serves individual motivation but misses the accountability use case managers actually need. Minor

2

"Vendor name" column — agents prefer "Restaurant name"

Throughout every session task — from filtering the pipeline to actioning follow-ups — 10 of 14 participants naturally said "restaurant" rather than "vendor." Every time they read the column header, a momentary translation step occurs. Across a 200+ account workload, this friction compounds with every interaction. Major

3

Vendors Needing Follow-up — 7 total, with status and actions

When asked to identify which vendors needed immediate follow-up from the dashboard, agents who spotted the Vendors Needing Follow-up section acted on it without any instruction. The "3d overdue" and "Due Today" badges were read correctly on first view, and the inline "Follow up →" CTA was tapped without hesitation. A strong interaction pattern that should be extended to the manager view. Minor

1 Design Decision Add a Team view to the Leaderboard for manager logins — show per-agent output totals, not just self-rank. Individual rank is fine for agents; team leads need to identify underperformers at a glance. Make the team view the default for any manager-role login.

2 Design Decision Rename "Vendor" → "Restaurant" across all column headers, list labels, and UI copy where agents interact directly. Agents universally think and speak in restaurants — every "vendor" label they read is a momentary translation. This is a copy-only change with zero engineering cost and immediate comprehension benefit across all 7 markets.

Participant Recordings · Marvin

Ahmed Saqr · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Lead

Major

Ahmed's main KPIs do not include GMV — a metric prominently featured in the Overall Achievement section. He questioned its relevance and noted it belongs to Account Management, not his field acquisition role.

"My KPIs don't have this situation this part of GMV. Because I not my main target. This target for AMS."

View in Marvin →

Muhammad Hafizuawan · 🇲🇾 Foodpanda MY Supervisor, Telesales

Major

The Estimated Payout was confusing as a KPI metric — agents at Foodpanda MY don't track payout as a performance target, and the label didn't communicate whether it referred to agent earnings or vendor revenue.

"We're not really sure on this. I mean, we don't have any specific target on estimated payout. Is it like to agents, or..."

View in Marvin →

Ahmed Saqr · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Lead

Enhancement

Ahmed suggested renaming "Vendor Name" to "Restaurant Name" in the pipeline activity breakdown — "Vendor Name" implies the restaurant owner's personal name rather than the business name, causing confusion when scanning lead lists.

"The vendor name. This means the vendor owner's name — not the restaurant name."

View in Marvin →

Screen Area 2 of 4

Pipeline

Blocker Major Delight

02

Finding 01 & 02

Pipeline label confusion and hidden filters

12 of 14 sessions All 7 markets 2 blockers combined

When asked to filter for vendors who had reached an agreement, 11 of 14 agents struggled with the "Qualified" stage name in the sidebar — and changing the label alone is unlikely to resolve the confusion at this point. Separately, 9 of 14 agents could not find the pipeline filter system at all — scanning the top action bar first (Salesforce muscle memory), finding only search and "+ Add Lead," and concluding filters did not exist. Both issues block agents before they can begin any pipeline task.

"Qualified for me means they've already passed the quality check — that they're actually ready. But then why would I be doing outreach on them?" — Egypt, Field Sales

1

2

1

"Qualified" in the sidebar stage filter

When users were asked to filter for vendors who had reached an agreement, no one understood the "Qualified" stage name. 11 of 14 users had difficulty understanding it — and it seems like changing stage names alone wouldn't help make them more efficient at this point. Blocker

2

Pipeline filters — not found by 9 of 14 agents

When asked to filter the pipeline by stage or grade, 9 of 14 agents went directly to the top action bar — the first place Salesforce muscle memory sends them. They found only search and "+ Add Lead," concluded that filtering was not a feature of the app, and stopped looking. The filter system exists and works; it simply lives where no Salesforce-trained agent thinks to look first. Blocker

1 Design Decision Rename "Qualified" → "Verbal Agreement" across all three surfaces — sidebar filter, vendor row badge, and header stage badge. Zero engineering effort, one text change eliminates the most widespread confusion across all 7 markets. Validate with 2–3 agents per entity type before shipping.

2 Design Decision Add a Filter button (funnel icon + label) in the top action bar next to search. The sidebar stays as the filter panel — it just needs a visible entry point where agents' Salesforce muscle memory already looks. One button added to one location resolves the blocker entirely.

Participant Recordings · Marvin

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition · 10 yrs tenure

Blocker

All pipeline stages were immediately clear — "Qualified" was the sole exception. Hany suggested renaming it to "Verbally Agreed," a term already in daily use across his team.

"All... it's very clear for me... only qualified. It's a little bit confusing for me... maybe verbally agreed?"

View in Marvin →

Ahmed Saqr · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Lead

Blocker

Ahmed conflated "Qualified" with sending a contract — the two concepts are not adjacent in his workflow. He recommended renaming it "Verbal Agreement" and removing it as a distinct stage for field acquisition.

"I don't understand what the qualified qualified means because it. If I send the code or the proposal it means I send a contract."

View in Marvin →

Lee Yee Mei · 🇲🇾 Foodpanda MY Field Sales

Blocker

Lee Yee immediately mapped "Qualified" to QC approval — a completely different stage in her workflow — revealing how deeply the term conflicts with field sales mental models in Malaysia.

"Qualifies, which means QC... QC already passed."

View in Marvin →

Aisyah + Hafizuawan · 🇲🇾 Foodpanda MY Coordinator, Telesales + Supervisor

Blocker

Both agents were uncertain. Understanding came only after the moderator explained it replaced "Verbal Agreement" — confirming the label requires immediate renaming, not user onboarding.

"Qualified for qualified link send... we're not really sure about this."

View in Marvin →

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition

Blocker

Hany could not locate the pipeline filters without moderator guidance. Filters are hidden in a scrollable left sidebar — a placement that directly conflicts with Salesforce muscle memory expecting them at the top of the page.

"No, where is the filter? Is there a different view?"

View in Marvin →

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition

Delight

Once Hany discovered the grade-based filters (A/B/C) after being guided to them, his reaction was immediate and enthusiastic — the value is there; discoverability is the only barrier.

"Okay... okay! That's really good. I love that. I think that's so good... it's really helpful."

View in Marvin →

Finding 05

Pipeline action clarity — ownership and icon recognition

6 of 14 sessions Egypt · Pakistan · Multiple markets Ownership + icon clarity

Two actions in the pipeline caused consistent confusion across sessions. When agents encountered "Menu Rejected" on a vendor row, agents in Egypt and Pakistan could not determine whether re-engagement was their responsibility or Operations' — the status signals a failure but gives no ownership or next step. Separately, when asked how they would share an onboarding update with a vendor, 6 of 14 agents could not understand the activity icons at first — spending time scanning before eventually finding the right action with some guidance. The icons are not self-explanatory for agents who are less tech-savvy.

"I didn't know what any of these icons meant — I had to guess. If it just said the word I would have clicked it straight away." — Pakistan, Field Sales

1

2

1

"Menu Rejected" sub-status — ownership unclear

When asked what they would do after seeing "Menu Rejected" appear on a vendor row, agents in Egypt and Pakistan immediately asked whether a notification had been sent to the vendor and whether re-engagement was their responsibility or Operations'. The status communicates that something failed — it does not communicate who acts next, or whether any automated follow-up has already been triggered. Major

2

"Log an activity" — icon-only actions not recognised

When asked how they would share an onboarding update with a vendor, 6 of 14 agents could not interpret the activity icons on first view. Agents spent time scanning the screen before finding the right action — and only after guidance. The icons alone are not enough for agents who are less tech-savvy. Recommendation: pair icons with visible text labels, or replace icon-only CTAs with explicit labelled buttons. Major

1 Design Decision For rejection sub-statuses like "Menu Rejected," add an explicit ownership chip — Sales or Ops — alongside the badge, with a primary CTA ("Re-engage vendor" or "Escalate to Ops"). The next action must never be invisible.

2 Design Decision Replace icon-only activity actions with labelled buttons or add visible text beneath each icon. Agents who are not tech-savvy should not need to guess what an action does — especially for a core task like sharing an onboarding update with a vendor.

Participant Recordings · Marvin

Abdulrahman Mohamed · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Executive

Major

Abdulrahman flagged that "Documents" appears after "Contracts" in the pipeline activity breakdown. In Egypt's actual workflow, documents must be collected before a contract is sent — the current order is reversed.

"What's the meaning here about documents? Oh... it could be a typo on our end; documents would appear before the contract."

View in Marvin →

Ahmed Saqr · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Lead

Major

"Link Sent" is a new concept with no equivalent in Ahmed's current workflow. He couldn't determine whether it referred to a portal link or agreement link — the stage name gives no contextual signal.

"Can you explain link sent. What do you understand from it? No, I don't know, I don't get it — the portal link or the agreement link?"

View in Marvin →

Aisyah Ummul Mukminin · 🇲🇾 Foodpanda MY Coordinator, Telesales

Enhancement

Aisyah suggested adding "close, lose, or create a new opportunity" actions directly in the new pipeline view — without this, agents must switch back to the legacy system to perform these common end-of-cycle actions.

"Currently, in our current ecosystem pipeline — we do not have the opportunity to close, lose, or create a new opportunity directly."

View in Marvin →

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

Enhancement

Victor asked whether pipeline data could be exported to Excel or PDF — team leads need to extract data to cross-check field agents' work and share progress reports offline or in stakeholder reviews.

"Is it possible for us to export this pipeline? Into PDF or Excel? Yeah, sometimes we may want to export these leads into Excel for certain usage."

View in Marvin →

Pipeline Delights

Pipeline baseline tasks — high task success across the board

14 of 14 sessions All 7 markets Task success

Despite the labelling and discoverability issues found elsewhere in the pipeline, agents had very high task success on the core pipeline actions — adding a lead, signing up a vendor, and calling a vendor — all without instruction. These results indicate that the fundamental pipeline structure and primary CTAs are well-placed and intuitive.

1

2

3

1

"+ Add Lead" button — 14 of 14 agents successful

When asked to add a new lead from the pipeline, all 14 agents immediately identified and clicked the "+ Add Lead" button without guidance. Zero hesitation across all markets — the placement and label are working. Delight

2

"Sign Up" button — 13 of 14 agents understood its purpose

When asked to initiate the vendor sign-up flow, 13 of 14 agents correctly identified the "Sign Up" button and completed the flow without confusion. Next step: test the complete end-to-end SSU journey with agents to validate the full flow beyond the entry point. Delight

3

Calling a vendor from the pipeline — 14 of 14 agents successful

When asked to call a vendor directly from the pipeline, all 14 agents found and used the call action without instruction. Inline calling from the pipeline list was immediately understood across all markets and agent types. Delight

1 Design Decision Protect these placements — "+ Add Lead," "Sign Up," and the inline call action are working as intended across all markets. Do not move or redesign these CTAs without re-testing. Prioritise completing the end-to-end SSU flow test as a next research step.

Screen Area 3 of 4

Communication Hub

Major Delight

03

Finding 06

AI Auto-Pilot — agents understand it but don't trust it yet

9 of 14 sessions All markets Trust + accuracy concerns

When asked to simulate an autonomous AI outreach call on their behalf, 9 of 14 agents understood the concept and saw some appeal — particularly for high-volume workloads. However, the majority raised two consistent blockers before they would consider using it: vendors would perceive the caller as an AI and drop the call, and if the AI communicated incorrect information the damage would be unrecoverable. Agents across markets said they don't feel the need for this feature right now, and they don't yet have the trust to use it.

"If the vendor finds out it's AI they will hang up immediately. And what if it tells them the wrong price? I can't fix that." — Egypt, Field Sales

1

2

3

1

AI Auto-Pilot — concept understood, adoption blocked by vendor perception

When asked to set up an autonomous outreach call, 9 of 14 agents understood the feature immediately. However, the majority raised the same concern unprompted: vendors would realise they were speaking to an AI and drop the call. Agents felt this would damage the relationship before a human even had the chance to intervene. Major

2

AI accuracy risk — wrong information to vendor is unrecoverable

When asked whether they would use Auto-Pilot for real outreach, agents across markets raised a second blocker: the risk of the AI communicating incorrect pricing, terms, or information to a vendor. Unlike a human mistake that can be corrected in the same call, agents felt an AI error would permanently damage vendor trust with no way to recover it in the moment. Major

3

Feature not needed right now — agents don't feel the gap

When asked whether they would use AI Auto-Pilot in their daily workflow, agents across markets said they didn't feel the need for it at this stage. The trust threshold required to let an AI call on their behalf — especially with vendors they are actively trying to onboard — has not been reached. The concept is understood; the readiness is not there yet. Major

1 Design Decision Do not ship AI Auto-Pilot as a default or prominent feature in V1. Agent trust has not been established and vendor perception risk is real. Consider a disclosure mechanism — informing vendors upfront that they may be contacted by an AI assistant — before any autonomous outreach is triggered.

2 Design Decision Address the AI accuracy risk before re-testing. Agents need confidence that the AI will not communicate incorrect pricing, terms, or commitments. Consider a fact-lock mechanism that constrains what the AI can state — surfacing only verified, pre-approved information for each vendor context.

3 Design Decision Re-test AI Auto-Pilot after trust-building features are in place — transparent AI disclosure, accuracy guardrails, and a track record from the AI co-pilot feature. Agents who are already comfortable with AI assistance in calls will be the early adopters; start there rather than pushing broad adoption.

Participant Recordings · Marvin

Muhammad Hafizuawan · 🇲🇾 Foodpanda MY Supervisor, Telesales

Major

Hafizuawan raised a critical question about AI language support — whether the call AI can capture and process conversations in Bahasa Melayu, not just English, for the feature to be genuinely usable in Malaysia.

"Will it capture the local language of our vendors as well, or will it only capture the English language?"

View in Marvin →

Delight D1

Communication Hub — the most praised feature across all 14 sessions

14 of 14 sessions Avg. 5/5 rating All markets

WhatsApp, Email, Calls, and AI in one tab row per vendor — described as "brilliant," "the thing I've always wanted," and "this alone would make me use the system every day." Agents currently split their day across Salesforce, WhatsApp Web, Gmail, and Excel. A single unified view per vendor eliminates all of that context-switching in one step.

"Right now I have five tabs open — WhatsApp Web, Salesforce, Gmail, Excel... this would be just one." — Bahrain, Field Sales

1

2

3

4

1

Channel tabs — WhatsApp / Email / Calls / AI Assistant in one row

When asked to reach out to a vendor directly from the pipeline, agents found WhatsApp, Email, Calls, and AI in a single row per vendor and tapped the right channel immediately — zero instruction given. Every agent across all markets recognised the layout and acted without hesitation. The moment the channel row appeared, agents described the elimination of their biggest daily friction: switching between Salesforce, WhatsApp Web, and Gmail. Delight

2

AI Call + Call Now — two distinct calling modes, side by side

When asked to initiate a call with a vendor, agents correctly identified and distinguished between "AI Call" and "Call Now" without confusion. UAE and Egypt agents asked for a one-line descriptor before the call starts — not because they were uncertain, but because they wanted confirmation of which mode they were entering before committing to a live dial. Delight

3

Conversation filter tabs — Hot Leads / Document Collection / Post-Contract

When asked to locate a vendor who was mid-document collection from within the Communication Hub, inside sales agents managing 200+ accounts found the Document Collection filter tab, selected it, and reached the right conversation immediately — no manual search or scroll required. The pre-filtered views replaced what would otherwise be a multi-step filter task done multiple times per day. Delight

4

Quick-reply template tags — Document Collection, Contract Ready, QC Rejection, Go Live Nudge

When asked to send a standard onboarding message to a vendor, agents discovered the quick-reply template tab and completed the task in a single tap. Agents across markets estimated this feature would save 20–30 minutes per day compared to their current practice of typing manually or copy-pasting from personal message banks saved in WhatsApp or Notes. Delight

Design Decision This is the product's primary differentiator — protect it. Keep the unified inbox as the default entry point. Add brief tooltips differentiating "AI Call" vs "Call Now" before launch. V1.1: add Viber and Facebook Messenger for Philippines.

Participant Recordings · Marvin

Abdulrahman Mohamed · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Executive

5 / 5

Abdulrahman rated the Communication Hub 5/5 with no improvement suggestions — the consolidation of email, WhatsApp, and calls into one panel was immediately compelling with zero friction.

"Maybe five... I have no adding value here; it's just amazing!"

View in Marvin →

Ahmed Saqr · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Lead

4.5 / 5

Ahmed called the WhatsApp + email integration a "fantastic option" — it directly replaces his current manual workflow of switching between personal WhatsApp and Gmail for vendor communication.

"WhatsApp and email — this is a fantastic option because it aligns with my primary method of contacting vendors."

View in Marvin →

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

5 / 5

Victor rated Communication Hub 5/5 unprompted, calling the AI-driven mechanics "totally different" from anything currently available — and explicitly noted it will be helpful for the whole team's daily workflow.

"There's a totally different new mechanics that will help us a lot. Yeah, I'll say five."

View in Marvin →

Muhammad Hafizuawan · 🇲🇾 Foodpanda MY Supervisor, Telesales

Delight

Hafizuawan praised the message templates and conversation type segregation in the hub — agents can select a template for "document collection" or "follow-up" and the message is pre-drafted, reducing manual composition time.

"It has its own template. So we want to contact this vendor to collect documents... we just click these options, and then the message is there."

View in Marvin →

Muhammad Hafizuawan · 🇲🇾 Foodpanda MY Supervisor, Telesales

Delight

Hafizuawan called real-time AI objection handling "really awesome" — detecting vendor objections mid-call and surfacing counter-suggestions for the agent without interrupting the conversation flow.

"It shows that we are in call with Eleanor... it also detects questions from the vendor, and then they also recommended counter additions. Yeah, yeah. So I think it's... it's really awesome."

View in Marvin →

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

Delight

Victor's reaction to the AI call assistant was immediate — "very great, very awesome" — recognising its potential to help agents handle objections with confidence and access relevant vendor context during a live call.

"There's an objection detected... Okay. So there's an AI to help you... very great, very awesome."

View in Marvin →

Victor Lim · 🇲🇾 Foodpanda MY Team Lead, Field Sales

Enhancement

Victor raised the need for multiple contacts per vendor — many Malaysian restaurants are family-owned, with co-owners and legal contacts who must all be reachable from the same vendor record in the system.

"If as a sales agent, I want to contact other contacts of the vendor, how can I see all the relevant contacts, like co-owner, you know, legal person, etc.?"

View in Marvin →

Delight D4

Post-call AI intelligence — auto-summary and one-click next actions

10 of 14 sessions Eliminates manual CRM entry All markets

The moment a call ends, the AI generates a structured summary — key discussion points, vendor interest level, agreed next step — with one-click buttons to execute each action without navigating away. Agents described this as eliminating the most tedious part of their day. "I forget what I promised by the time I'm done with the next call" came up verbatim in multiple sessions. This resolves it completely.

"No more forgetting what I said. No more writing notes. After 30 calls a day this is the thing that kills me — now it's just done." — Egypt, Tele-Sales

1

2

3

1

Call Completed ✓ — "Summary logged to timeline" auto-confirmation

When asked what they normally do after completing a call, agents described 5–10 minutes of manual note-writing in Salesforce. After the call simulation ended, they were shown the auto-generated summary already logged to the vendor timeline. 10 of 14 agents reacted immediately with recognition of the time saved — without any prompting. The confirmation screen alone resolved the single most cited post-call pain point in the study. Delight

2

Key Discussion Points — 5 AI-generated bullets

When asked to review the AI-generated call summary, agents scanned the 5 structured bullets in seconds and confirmed accuracy without needing to cross-reference any notes. The brevity was specifically praised across sessions — agents said a longer summary would slow them down. "I just confirm it's right and move on" was the consistent response. Delight

3

Recommended Next Actions — AI-suggested workflow

When asked what they would do next after reviewing the call summary, agents checked the Recommended Next Actions panel first — then confirmed it had already identified the correct step. The AI suggestions (Schedule / Send contract / Create task / Advance stage) matched what agents intended to do in every session where this scenario was tested. "I don't have to think about what comes next — it already knows." Delight

Participant Recordings · Marvin

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition

Delight

Hany immediately grasped the value of the post-call AI summary — it would capture the commission rate agreed and the specific terms discussed, eliminating the risk of forgetting what was promised during a vendor call.

"I think it will record what I have discussed with them over the call... I think it will give me the five points that I have discussed."

View in Marvin →

Abdulrahman Mohamed · 🇪🇬 Talabat Egypt Sr. Sales Acquisition Executive

Delight

After seeing the post-call screen with AI-generated discussion highlights and one-click follow-up prompts, Abdulrahman described it as "perfect" — structured, actionable, and immediately useful without any learning curve.

"That information we discussed, the code, it's highlighted from AI... Also, we need to schedule another follow-up."

View in Marvin →

Hany Abdelmalak · 🇦🇪 Talabat UAE Sr. Sales Acquisition

Enhancement

Hany noted that substantive pitches conducted over the phone should be classifiable as "meetings" — currently the system logs every phone interaction identically as a "call," losing an important distinction in his activity history.

"How can I differentiate if I have a meeting over a call? Like, sometimes we pitch over a call; it will always show as the call log, not as a meeting logged."

View in Marvin →

V1.1 Backlog · All screens

Minor Enhancements & Market Requests

Minor Suggestion

04

Severity	Pain Point / Need	Design Decision
2 · Minor	Philippines needs Viber and Facebook Messenger — 2 sessions. WhatsApp is not the dominant channel; agents use Viber and FB Messenger as primary vendor contact channels.	Add Viber and Facebook Messenger as communication channels scoped to the PH entity.
2 · Minor	Malaysia needs multi-contact per vendor — 2 sessions. Family-owned businesses often have 2–3 decision-makers (owner, spouse, ops manager) who all need to be reachable.	Support multiple contacts per vendor record in the communication panel.
2 · Minor	Share / SSU link icon misread — 2 sessions (PK, BH). Current paper-plane icon is interpreted as "forward to colleague" or "share on WhatsApp" rather than "send self-sign-up link."	Replace with a labelled button "Send SSU Link" or a chain-link icon to make the action self-describing.
2 · Minor	Automation needed for adoption — 2 sessions. If agents must manually log every next activity, adoption will fail. System should automate next-step suggestions and status updates.	Auto-escalate hot leads after 48h no-contact; auto-suggest next steps post-call; auto-update pipeline stage when vendor completes document submission.
1 · Suggestion	Date filter missing from leads — Pakistan. Leads go cold within 1–2 days; agents need to sort by assignment date to prioritise freshest leads.	Add lead assignment date as a sortable column; default pipeline to ascending date for PK entity.
1 · Suggestion	SSU link can only be sent via one channel at a time. Agents want simultaneous WhatsApp + email send to ensure delivery.	Enable multi-channel simultaneous SSU link delivery in the Share modal.
1 · Suggestion	Onboarding deadline not visible from pipeline list — Malaysia. Agents miss deadlines because the date is only visible after clicking into a vendor record.	Surface onboarding deadline as a visible column or badge in the pipeline list view.
1 · Suggestion	Vendor phone number and area / zone missing from hot leads list — Malaysia field sales. Needed to plan route-based outreach.	Add phone number and geographic area/zone to hot leads list columns.
1 · Suggestion	AI assistant not accessible during WhatsApp / email text conversations — only available during active calls.	Extend AI assistance to in-conversation text chat panel (WhatsApp and Email tabs).
1 · Suggestion	Communication category tabs are fixed — agents want custom categories (new leads, warm leads, document collection, activated vendors).	Enable customisable communication category tabs per agent preference.

Supporting Evidence · Share SSU Link Modal Share Onboarding Link — Email or WhatsApp, not both simultaneously The modal correctly shows Email and WhatsApp as send options — but only one channel at a time. Agents in Malaysia and Pakistan asked for simultaneous multi-channel delivery.

1

2

1

Send via — Email tab or WhatsApp tab (mutually exclusive)

When asked to send the onboarding link to a vendor, agents in Pakistan and Malaysia immediately asked whether they could send to both channels simultaneously. The Email and WhatsApp tabs are mutually exclusive — only one can be selected. Agents who don't know which channel their vendor monitors said they would try both anyway, sending sequentially rather than in one action — a workaround that adds unnecessary steps to every link send. Suggestion

2

Share icon in modal header (paper-plane)

When asked to locate the option to send the onboarding link from the pipeline list view, several agents misread the paper-plane icon as "forward to colleague" or "share on WhatsApp." The modal title correctly identifies the action once opened — but the row-level icon has no text label and was misidentified without that surrounding context. New agents encountering this icon for the first time are unlikely to tap something they cannot name. Minor

Dashboard

Average rating: 4.1 / 5. The concept of one unified cockpit is well-received, but KPIs are too generic and payout visibility creates sensitivity.

Pipeline

Average rating: 4.5 / 5. The pipeline view is well-structured, but stage terminology and filter placement are blockers. Market-specific stages are entirely missing.

Communication Hub

Average rating: 4.6 / 5 — the standout feature of the entire prototype. Near-unanimous 5/5 across all markets with specific channel and customisation requests.

Onboarding Flow

SSU link sharing and vendor impersonation were well understood. Document collection process and onboarding deadline visibility need attention.

AI Call Feature

Real-time AI coaching during human-led calls is highly valued. Autonomous AI calling (AI initiates the call) was rejected by at least one participant and needs careful positioning.

Market Insights

Each entity has distinct workflow requirements. A single pipeline configuration will not serve all markets without customisation.

🇦🇪🇪🇬🇧🇭

UAE · Egypt · Bahrain

• More Salesforce-familiar — smoother conceptual onboarding than other markets
• Egypt: Document collection comes BEFORE contract — current pipeline order is reversed
• Bahrain: "Rejected" stage is critical after QC; "Quality Check" stage should be removed from pipeline
• Commission fields must include all fees: tablet cost, monthly subscription, bank fees
• UAE: Calls exceeding ~20 min should auto-escalate to "Meeting" classification
• Shared: Rename "Qualified" → "Verbal Agreement" (all three markets)

🇵🇰

PK

• Leads go cold in 1–2 days — date filter and date-sorted pipeline are critical, not a nice-to-have
• Multi-step quote approval (Regional Manager / Head of Sales approval required before quote is sent) is not reflected in current pipeline stages
• Missing stages: "Quote Sent," "Quote Pending," "Onboarding Failed"
• Regional KPIs: clawbacks for zero-order vendors, QC rejection rates, lead utilisation
• Vendor upload takes 45–50 minutes — biggest operational burden, must be reduced
• Separate Google Sheet approval is major redundancy — Salesforce should be single source of truth

🇲🇾

MY

• PDPA (Personal Data Protection Act) concern — estimated payout visible to colleagues creates legal risk
• Personal vs. company WhatsApp number is a workflow blocker for field sales outside office hours
• Multiple contacts per vendor needed — many vendors are family businesses with co-owners and legal contacts
• Field sales need vendor geolocation and phone number visible in pipeline list for physical visits
• Onboarding deadline must be surfaced without requiring deep account navigation
• Team lead needs dedicated dashboard: submission performance, stalled leads by agent, individual tracking

🇵🇭

PH

• WhatsApp is NOT widely used by vendors — Viber and Facebook Messenger are the primary channels
• Managers need agent-level pipeline filtering — click an agent's name to see their specific accounts
• Per-vendor notes are critical for managers tracking complex accounts (e.g. "corporate bank account missing")
• Inside sales managing 200+ accounts — Hot Leads / prioritisation is the most valued feature
• Custom email templates per Business Development agent needed

🇹🇷

TR

• SSO/VPN access issues throughout both Turkey sessions — prototype couldn't be tested directly
• KYB status visibility is critical — QC cannot start until KYB clears, and this isn't currently trackable
• Estimated payout is sensitive due to Turkey's complex tax system — estimate vs. actual divergence would demotivate agents
• Pipeline stages don't match YS workflow — "Processing Vendor" stage doesn't exist; restaurants go directly to menu processing after QC approval
• WhatsApp Business API 24-hour reply window restriction is a dealbreaker for their workflow
• Team recommends 3–4 weeks of multi-person parallel testing before full activation
• Rejected cases are the #1 priority — agents need to see and action QC-rejected restaurants quickly
• "It's a huge change and it can create a huge mess. We should be able to test it before activation."

Prioritised Recommendations

Ordered by urgency and impact. Immediate items are pre-launch blockers. High priority items are pre-launch or V1.1. Medium items target V1.1–V1.2.

Supporting Evidence

140 tagged clips from 5 participants across Talabat UAE, Talabat Egypt, and Foodpanda MY — exported from Marvin and mapped to findings by severity. Click any clip to open it directly in Marvin.

Marvin Workspace · Sales Ecosystem 2.0

Project Files

14 sessions · all recordings tagged

Each session imported, transcribed, and clipped. Tags mapped to findings by severity before analysis.

Analysis Board

Clips grouped by finding, severity & theme

Each card is a verbatim quote mapped to a finding. Blockers, majors, and delights colour-coded across all markets.

AI Insight Report

Auto-generated synthesis across all sessions

Marvin's deep research report cross-referenced with manual analysis to validate patterns and coverage.