Freeze sales hiring during the quality rebuild
Alternative considered
Keep hiring while building in parallel
Why this one
Onboarding vendors into a still-broken system would have compounded the churn problem.
Wedding services · Program management
40% Vendor Churn. The Fix Was a Sales Freeze, Not More Salespeople.
“Fix retention before scaling acquisition. New vendors into a broken system just creates more churn to replace.”
Program management at India’s largest wedding marketplace: the decisions, the stakeholder conflict, and the near-launch incident that almost derailed peak season.
- Role
- Business Development Manager
- Team
- 40-person cross-functional team
- Timeline
- 2019 – 2020
- Company
- WedMeGood
87%
Vendor churn reduction
40% down to 5%
94%
ARR growth
In 12 months
40
Person cross-functional team led
01
Background
Rapid growth had outrun manual process. Vendor quality was degrading faster than sales could replace what was leaving.
Manual vendor verification
85% of time went to repetitive checks, costing roughly $10K a month in operational overhead.
Duplicate vendor listings
Degraded the user experience directly: a 30% rise in customer complaints traced back to it.
No quality scoring system
Decisions were subjective and inconsistent, and high-value vendors were leaving as a result: 40% churn.
02
The problem to solve first
Freeze sales, fix quality first. Two other directions got ruled out.
Ruled out: hire 20 more salespeople
Scales the leak, it doesn’t patch it. More acquisition into a broken system just replaces the same churn.
Ruled out: a custom-built platform
Six months of engineering time would have meant missing peak wedding season entirely.
03
What I considered instead
Build on Salesforce, or request a 6-month custom build? Salesforce configured in 4 weeks versus 6 months for custom. Some advanced features weren’t possible, but the platform shipped 5 months earlier.
Weighted decision matrix
Salesforce versus a custom build
| Criterion | Weight | Salesforce (chosen) | Custom build |
|---|---|---|---|
| Speed to launch | ×3 | 3 | 1 |
| Fits the team’s actual resources | ×3 | 3 | 2 |
| Feature completeness | ×1 | 2 | 3 |
| Long-term scalability | ×1 | 2 | 3 |
| Weighted total | 22 | 15 | |
Automated scoring with manual override
Alternative considered
Fully automated, no human intervention
Why this one
Weddings are high-stakes: a false positive (rejecting a good vendor) cost more than a false negative. Human judgment stayed in the loop for edge cases.
04
Stakeholder alignment
Three people had real, legitimate reasons to resist this. Each got reframed, not overruled. I held Driver throughout: the one person accountable for moving the decision forward.
ConsultedVP of Business Growth
OpposedAdvocate
Strongly opposed to the sales hiring freeze; saw it as an attack on his team.
“Team’s bonuses depend on new vendor numbers. You’re asking me to tell them to stop working.”
Reframed the metric: 60% of his team’s work was re-acquiring vendors who had already churned. Proposed a temporary shift from “new vendors” to “net vendor growth,” and shared credit for the retention gains.
Became the biggest advocate, presenting the approach to the C-suite as a sales efficiency initiative. His team’s earnings rose 20% from retention bonuses.
ApproverCEO
SkepticalApproved
Skeptical; needed growth numbers for an investor update.
“We have a board meeting in Q3. I need to show vendor growth, not explain why we paused.”
Built two projections side by side: continuing the current path (15K net new vendors, 40% churn) versus fix-then-scale (8K in Q2, but 30K by year-end at 95% retention). The second was the better investor story.
Approved the pause, and used the retention story as a competitive-moat narrative in the investor presentation.
ConsultedEngineering Lead
Wanted custom buildDeprioritized it
Wanted to build a custom platform from scratch.
“Salesforce is a hack. Give me 6 months and I’ll build something scalable.”
Agreed in principle, but proposed validating the business logic on Salesforce first: if it worked, a scalable rebuild made sense; if not, only 4 weeks were lost, not 6 months.
The Salesforce solution worked well enough that the custom rebuild was deprioritized; engineering time went to revenue features instead.
Tagged using DACI (Atlassian): Driver, Approver, Consulted, Informed.
05
What went wrong
Two weeks before peak-season launch, real vendors started getting flagged as duplicates.
- 2 weeks out
Duplicates flagged real vendors
Detection logic used name + phone number as the unique ID; it didn’t account for multi-city vendors sharing one contact.
- Within 4 hrs
CEO escalation
200+ premium vendors were temporarily suspended before anyone caught it.
- That weekend
Manual rollback
Reverted flagged accounts to manual review, added location-aware logic, personally called the top 50 affected vendors.
- Fixed
Location-aware logic shipped
Tested against real power users before the next launch.
The vendors who broke the system were often the best customers with the most complex setups: multi-city vendors sharing one phone number never showed up in the sample data used to test the logic.
06
What I’d do differently
Three real mistakes, and what they cost.
Waited too long to involve sales in scoring design
Impact
The first version rejected vendors that sales had personally recruited, creating political tension.
What I’d do instead
Include sales reps in defining scoring criteria from day one, not after the system is built.
Didn’t document the reasoning behind scoring thresholds
Impact
When I left, the team struggled to adjust thresholds for new vendor categories.
What I’d do instead
Keep a decision log explaining the “why” behind each threshold.
Over-optimized the Excel analytics layer for speed
Impact
Spreadsheets broke when data formats changed, requiring weekend firefighting twice.
What I’d do instead
Spend the extra week on data validation and error handling from the start.
07
How this connects to outcomes
Two metrics that used to be conflated: verification speed and acquisition throughput are different things, and both improved.
Stepwise attribution
How churn actually dropped, 40% to 5%
40%
Starting churn
−14pt
Quality scoring framework
−9pt
Automated verification
−7pt
Duplicate-detection fix
−5pt
Sales-freeze breathing room
5%
Ending churn
Verification time per vendor
4 hours 10 minutes96% faster
Vendor acquisitions per rep
6–8 / day 12–14 / daynearly 2x
Duplicate listings
15% of catalog 2% of catalog87% cleaner
Revenue
$1.8M ARR $3.5M ARR94% growth
Get in touch
Let’s talk about what you’re building.
Untangling operational friction, or deciding what to fix before you scale? I’d like to hear what you’re working on.