GROWTH DECISION STUDY / 02 — B2B REAL ESTATE MARKETPLACE
The high-intent conversations no one saw disappear.
I cross-referenced two databases and found hundreds of customers with real purchase intent who never reached a broker's inbox.
Interface shown in Spanish — built for a Mexican marketplace.
30-SECOND EXECUTIVE SUMMARY
01 · THE PROBLEM
Hundreds of interested customers were never handled by a human broker.
02 · THE FINDING
The loss started in one specific week, and no one at the company noticed.
03 · THE PROPOSAL
A prioritized rescue queue where a person decides every contact.
ACT I · WHAT WAS BEING LOST
01 / CONTEXT
The challenge wasn't building a tool. It was noticing what was being lost.
Total traffic grew month over month, but advisors' calendars in key markets were emptying with no visible explanation.
STEP 01
Early signal
Overall traffic up (+28%), but brokers kept complaining about empty weeks.
STEP 02
A simple question
How many people who write in with real purchase intent never reach an advisor?
STEP 03
Cross-referencing data
Cross-reference the automated chat sessions with CRM records before touching the interface.
STEP 04
Clear direction
Pause spending on new acquisition until the leak in lead delivery is fixed.
Prioritize recovery
02 / SIGNAL
The drop wasn't gradual. It had a date.
Charting conversion week by week over 16 consecutive weeks turned the anomaly from an opinion into a measurable event.
Weekly rate of conversations delivered to a broker
16 consecutive weeks analyzed
Over a 16-week window, about 51% of chats were delivered successfully before the break identified in week 7. After the break, delivery fell to about 13%, a 38-point difference.
Real anonymized data, grouped by week.
- Normal flow
- Unnoticed leak (−38 points)
03 / RESEARCH
Before building, I worked through four questions.
Instead of assuming causes, I designed four direct analytical checks to rule out the team's assumptions.
01 · WHEN
Was there a specific breaking point?
Or was it a gradual fade in market interest?
One specific week (Week 7). A discrete event.
Confirmed
02 · WHY
Was the chatbot update to blame?
Engineering assumed a week-12 change had broken the flow.
The break happened 5 weeks before that update.
Ruled out
03 · WHO
Were they browsers with no budget, or real buyers?
Detailed audit of 120 archived, unassigned sessions.
Customers with square footage, a budget, and a corporate email.
Validated
04 · HOW MUCH
How much money did the leak represent?
Financial model based on historical real estate commission tickets.
An estimate above MXN 300,000.
Quantified
04 / INSIGHT
The lead didn't drop off. No one picked it up.
The user eagerly completed every requirement, the system promised contact within 24 hours, and the message was archived unassigned.
A real, silent loss. So how much was it worth? ↓
05 / SIZING
A loss no one measures has no owner.
To get leadership to act right away, I turned the lost records into a model of conversion and recoverable value.
01 · TOTAL UNIVERSE
643
Conversations with contact details and explicit intent, left unassigned.
02 · POTENTIAL APPOINTMENTS
~190
Estimated visits if contacted within 2 hours of triage.
03 · PROJECTED CLOSES
~19
Lease agreements signed at the historical ratio.
04 · VALUE AT RISK
~$347k
MXN in estimated direct commissions at the average ticket.
* Assumptions based on prior historical rates and average commissions reported by market. An analytical estimate for decision-making.
06 / DECISION
Don't rescue everything. Rescue what's worth the most first.
Calling all 643 contacts at once would have swamped the sales team. I ranked them by value.
THE 643 CONTACTS, RANKED BY VALUE
- START HERE
PRIORITY 1
Rescue today≥ 70 pts
High impact
A senior advisor calls within 2 hours.
PRIORITY 2
Needs review40–69 pts
Medium impact
A coordinator reviews the record the same day.
PRIORITY 3
Low priority< 40 pts
Passive nurturing
They get market report emails. No cold calls.
HOW IT'S SCORED
5 signals, each worth up to its points depending on strength
- +30 ptsStated budgetExplicit rent mentioned in the conversation
- +20 ptsSpecific spaceSize or features
- +15 ptsSpec sheetsDownloads or repeat views
- +20 ptsClosing questionsAvailability or contracts
- +15 ptsRecencyUnder 14 days old
EXAMPLE · ANDREA SALAS
Total score from the 5 signals
- Budget 30/30
- Space 20/20
- Spec sheets 12/15
- Closing questions 20/20
- Recency 10/15
92 / 100
● High · Rescue todayACT II · WHAT I BUILT
07 / PRODUCT
From an invisible loss to a work queue.
Instead of automating cold messages that hurt the brand, the system helps draft in context and leaves the decision and the sending to the human advisor.
Review
Sees who they were, what they were looking for, and what the assistant showed them.
Adjust
Edits the suggested message before using it.
Log
Copies the message, sends it themselves, and logs the outcome.
Non-negotiable rule: no message goes out without a person approving it.
08 / JUDGMENT
The system knows when not to act.
A good automation product is defined more by what it chooses to refuse than by what it runs blindly.
| SITUATION | DETECTED CONDITION | SYSTEM ACTION |
|---|---|---|
| High intent, complete data | Score ≥ 70 with a valid phone and explicit budget | PRIORITY Adds it to the on-duty advisor's direct queue with a 2-hour window. |
| Incomplete data | No size or no clear budget in the session | REVIEW Asks the advisor to complete the record before reaching out, to protect the relationship. |
| Contact older than 60 days | More than 2 months with no activity | NURTURE Suggests a quarterly email report. No cold calls. |
| No valid phone | Fake number, missing digits, or bounced email | QUARANTINE Blocks the task so the sales team doesn't waste time. |
| Privacy and governance | Throughout the initial view | MASKED Hides numbers (+52 55 ••••0000) until the advisor formally accepts the case. |
09 / MEASUREMENT
The next step wasn't scaling. It was validating.
Leadership validated the prototype. I set three operating goals to test it with a pilot cohort before any rollout at scale.
GOAL 01 · RESPONSE
≥ 25%
Positive response rate for booking an appointment with the broker.
GOAL 02 · SPEED
< 90 min
Average time to first contact after the lead is assigned.
GOAL 03 · CARE
< 5%
Maximum rate of opt-outs or complaints about unwanted contact.
WHAT WE MEASURE AND VALIDATE
Validation with a pilot cohort
Goal 01 · Response ≥ 25%
Positive response rate for booking an appointment with the broker. A signal of real intent to buy or lease.
Goal 02 · Speed < 90 min
Average time to first contact after the lead is assigned. Cuts critical latency at the top of the funnel.
Goal 03 · Care < 5%
Maximum rate of opt-outs or complaints about unwanted contact. Protects the brand with the pilot cohort.
CLOSING · LEARNINGS & CONTINUITY
10 / LEARNINGS
What I learned by making the invisible visible.
MY DIRECT ROLE
I investigated the gaps between the chat conversations and the CRM inboxes. I found the exact date of the break, which had gone unnoticed for over a month.
I turned the missed handoffs into a quantified financial impact for senior leadership and prototyped the operational rescue tool, putting the advisors' judgment first.
A loss no one sees isn't small. It's a loss with no owner.
WHAT I TAKE AWAY
01 · Detect
Before building new features, audit where current users with real intent are leaking out.
02 · Quantify
Turn silent failures into clear numbers and transparent assumptions to align leadership.
03 · Prioritize
Not every contact deserves an immediate human call. Rank by value so you don't overwhelm the team.
04 · Validate
Simple tools with human control beat blind automation in high-value consultative sales.
CONFIDENTIALITY NOTE: This case was anonymized to protect the business's identity. Company names, internal systems, people, and real contact details are omitted. Conversations and profiles are illustrative recreations based on real patterns. Financial figures are estimates based on documented assumptions, not audited financial statements.
NEXT CONVERSATION
What's holding back your product's growth?
If you're dealing with conversion, activation, or experience friction, we can start by understanding what's happening and what's worth improving.
Tell me about your problem