Take-home brief

Reducing Early Churn Without Slowing Gross Adds

Enter password to view.

Reducing Early Churn Without Slowing Gross Adds

Kevin Middleton
May 2026
Section 1 · Problem Framing

Customer quality, not customer volume.

The right customer at the right tier, every time.

The Problem

Leadership wants to reduce early churn and customer support costs, without materially slowing gross adds.

Strategic Context

Three voices from the onsite shaped how I thought about this.

Recommendation

Target: 6-15% reduction in 90-day early churn within two quarters, validated by component-level A/B testing, without measurable impact on gross adds.

Why this won't slow gross adds. The hypothesis is that the right friction at the top (coverage transparency, tiered identity verification, plan education that sets honest expectations on throttling) filters customers who would have churned in the first 30 days anyway. Higher quality customers, not fewer customers. The answer isn't slower onboarding; it's right-sized onboarding for who's coming through the door.

Customer Segments

Four segments to design for, ordered by risk profile.

Fixed bundlers adding Mobile Lowest risk

Already verified, billing history on file. Highest LTV. The current flow probably feels like overkill to them.

Goal: frictionless. Reward the existing relationship.

Returning customers Variable risk

Identity and credit history are signals the current flow throws away. They left for a reason, came back for a reason.

Goal: acknowledge the return, calibrate based on history.

Referred new customers Partial trust

Don't currently exist as a segment because Mobile has no referral program of its own. A referral comes pre-qualified by trust.

Goal: smoother path earned by the referral context.

Cold new customers acquired by promo Highest risk

No existing relationship, no trust signal. Highest fraud risk, highest churn risk, lowest unit economics.

Goal: full verification. Each step prevents a downstream cost.

Stated Assumptions

Without internal data, I'm working from industry benchmarks. MVNO and prepaid wireless monthly churn runs 3-5%. The first 90 days carry disproportionate risk; the first 7-14 days are the most predictive of long-term retention.

Four churn drivers that frequently show up:

Insight

Cricket Wireless cut early churn 37% by tightening consistency between sales promise, delivered service, and first bill.

Section 2 · Options & Tradeoffs

Three approaches that fit together.

The funnel is the foundation, the referral program increases lead quality/LTV, and unified support reduces confusion and churn.

01

Adaptive componentized funnel

Foundational

Build onboarding as a library of modular components (coverage check, plan education, light or full eKYC, credit decision, deposit logic, activation, post-activation engagement) and assemble different paths per segment based on a real-time risk score. Fixed bundlers see four steps. Cold customers see seven. Same components, different path.

This aligns directly with the instinct on differentiated fraud flows. ID scanning isn't on or off. It's one tier in a graduated identity system. Light eKYC for referred customers, full eKYC for cold promo signups, skip entirely for verified Fixed bundlers.

Pros
  • Maximum compliance flexibility
  • Supports today's bundle base AND tomorrow's standalone play
  • Every component is independently A/B testable
Cons
  • Larger initial build
  • Requires modular front-end architecture
  • Decisioning service is a new platform component
Risks & Mitigations
  • Decisioning is a single point of failure: fall back to "treat as cold" if down
  • Flow sprawl: strict component ownership and shared analytics
Track Record · I've Built This Before

At HVAC.com I owned a 7-step funnel. The whole thing was instrumented for A/B testing: I could reorder steps, update the copy within them, and remove steps entirely, then measure the conversion impact of each change.

Modular Funnel: Same Components, Variable Steps

Four customer segments use the same components, but follow different journeys. Click a customer segment to see how the funnel adapts.

Adaptive Funnel by Segment Interactive

Click a customer segment to see how the funnel adapts.

01
Coverage Check
02
Plan Education
03
Identity / eKYC
04
Credit Decision
05
Deposit Logic
06
Activation
07
Post-Activation
Full depth
Light depth
Conditional
Skipped
02

Retention-tied referral program

Acquisition + Retention

Add a Mobile referral program alongside promo codes. Both referrer and new customer earn $10 at activation and $20 more at 90 days, with the larger credit gated on staying through the promo cliff. This extends Optimum's existing Fixed-side referral program ($50, 60-day vesting), with smaller credits sized to Mobile and two vesting milestones instead of one.

Pros
  • Better-quality leads (referrers self-select for fit)
  • Stickier than promo codes
Cons
  • Slower top-of-funnel in months 1-2 while seeding
  • Introduces a new fraud surface
  • Requires Fixed-side technical alignment
Risks & Mitigations
  • Referral abuse: identity-tied accounts plus 90-day vesting
  • Cannibalizing promo signups: run both in parallel, compare the LTV delta
03

Unified, context-aware support

Quick Wins

Today the mobile homepage chatbot and the Allie support bot are two different experiences, and a user mid-onboarding can land on either. Unify both into a single chat that gets users where they need to go, regardless of where they entered from.

Pros
  • Smaller scope, faster to ship
  • Addresses the support-cost half of the brief directly
  • Existing chat infrastructure to build on
Cons
  • Requires content coordination with the content team
  • Chat platform may have routing constraints
Risks & Mitigations
  • Routing complexity: start simple with page and intent signals, expand only when measurable problems demand it

Why all three, in what order

Each approach hits a different churn driver. The funnel hits onboarding friction and bill-shock prevention. The referral program hits customer quality at the top. Unified support hits support friction and ticket volume per new customer.

Sequence the work to deliver value while the bigger build is in flight. Unified chat ships first as quick wins and instrumentation. The modular funnel builds in parallel. The referral program layers on once the funnel can receive segment-aware traffic. The Execution tab walks through the wave-by-wave plan.

Section 3 · Metrics & Validation

What we measure, what we test, what we validate first.

Primary Metrics

Track five metrics to see if the changes are working. Cut each by segment so we can tell when one kind of customer is getting better while another may be getting worse.

Headline Metric

90-Day Churn Rate

6-15%
Reduction within two quarters, cut by segment. The top-line measure of whether the program is working.
Activation Completion Rate

Cut by funnel variant. Did the customer finish the path they were assigned?

Support Contact Rate

Per new customer in the first 30 days. Tracks support load and serves as an onboarding-experience proxy.

Gross Adds

The protected constraint from the brief. Watch for any meaningful negative delta.

Cost-to-Serve

Per onboarded customer, online versus in-store. Tracks the economics of each channel as volume shifts.

Secondary & Diagnostic Metrics

Used to diagnose moves in primary metrics:

Validation Approach

Modular architecture pays off here: we can run separate experiments on coverage check placement, eKYC step ordering, deposit messaging, and welcome emails without touching adjacent components. Existing flow as control, new adaptive flow as treatment, evaluated separately by segment.

The first and easiest test: make the existing coverage checker (optimum.com/mobile/network) a required Step 1 of onboarding (auto-confirmed for Fixed bundlers since we already have their address). The example here shows a multi-zip coverage checker flow: home, work, and other, so that you can easily check your most-visited locations in one shot. Mint Mobile makes coverage check a hard requirement at signup. Coverage gaps are a leading silent-churn driver. Customers who exit at this step aren't lost gross adds, they're 30-day churns we surfaced early and didn't pay support, SIM, and activation costs to discover later.

The metric to watch: 90-day churn delta between the new flow (with required coverage check) and the existing flow (no coverage step in onboarding). Coverage-related exits get their own bucket, not counted as funnel abandonment.

The target: 6-15% reduction in 90-day churn. That's the conservative end of a 6-20% range from industry case studies. We'll adjust once we see internal numbers.

Coverage Check
Onboarding · Step 1
Where do you need service?
Home
Required Step 1 of signup. Filters customers who would have churned at week 2 over coverage, before we paid activation costs.

Seven components, seven test surfaces

Each card below shows what one component actually looks like, plus the experiment we'd run on it.

Coverage Check
Step 1
Where do you need service?
Home
Test: required Step 1 vs no coverage step at signup.
Metric: 90-day churn delta + pre-commitment exits.
Choose a Plan
Step 2
What fits your usage?
1 GB$15
5 GB$25
Unlimited$20/line · 3+ lines
Unlimited Max$26.67/line · 3+ lines
Speeds may slow on heavy use.
Test: honest throttling copy vs marketing-only.
Metric: bill-shock tickets in days 30-45.
Verify Identity
Step 3
Tier: Full eKYC
SSN
DOB
Upload Government ID
Test: light vs full vs ID scan, by risk tier.
Metric: fraud rate × completion rate.
Credit Decision
Step 4
Real-time decisioning
Reviewing...
Approved
Tier 2 · Standard deposit
Test: model thresholds and decline messaging.
Metric: deposit acceptance, churn by tier.
Refundable Deposit
Step 5
Based on credit tier
$0
Refundable
Returned after 6 on-time bills
Soft pull, no impact on credit
Test: amount, refund timing, soft vs hard pull copy.
Metric: completion-after-deposit, retention.
Activation
Step 6
Choose your SIM
eSIM (instant)
Physical SIM (ship)
Port
Test: SIM vs eSIM default, port-in ordering.
Metric: time-to-activation, abandon rate.
Welcome Aboard
Step 7
Your first bill preview
Today (promo)$30.00
After promo$45.00
Promo ends Aug 24
Test: bill preview timing, day-0 vs day-30 nudge.
Metric: 30-day retention, contact rate.
Section 4 · Execution & Sequencing

Three waves so we deliver value while the bigger build is in flight.

Quick wins in weeks 1-6. Modular funnel through month 4. Referral program by month 6. Each wave produces signal that informs the next.

Three Waves

Wave 1 Weeks 1-6

Quick wins & instrumentation

  • Unify chat into a single experience that gets users where they need to go, regardless of where they are on the site.
  • Surface the existing coverage checker (optimum.com/mobile/network) as part of onboarding. A/B against the current flow where it sits on a separate page.
  • Add two new messages to Optimum's existing bill SMS: an activation welcome previewing the promo cliff, and a warning before it hits. No more surprise bills.
  • Stand up a 90-day churn dashboard cut by segment.
Bill expectation SMS sequence 1 of 4
Optimum Mobile Text Message · SMS
Welcome to Optimum Mobile! Line ending in 5309 is active. First bill of $30 due May 16. Heads up: promo ends Aug 24, monthly becomes $45.
View your Optimum Mobile bill ending in 5309 at optimum.com/billview. Total amount due of $30 will be automatically paid on May 16.
Thanks for your recent payment of $30. It will be applied to your Optimum Mobile account ending in 5309.
Heads up: your promo period ends in 7 days. Your next bill will be $45 (vs $30 previously). Auto-pay is on.
Outcome: Quick wins shipped before engineering capacity is absorbed by the bigger build.
Wave 2 Months 2-4

Build the modular funnel

  • Define and own components: coverage, plan, light eKYC, full eKYC, credit, deposit, activation, engagement.
  • Build the risk-tier decisioning service with Credit & Collections (tier definitions, fall-back behavior, audit logging).
Modular funnel build order · by segment
1
Fixed bundlers
Lowest risk. Validates the architecture with the simplest case.
2
Returning customers
Known identity and credit. Light extension of Fixed.
3
Referred customers Wave 3
Trust signal + dual-vesting lower the risk. Ships with the referral program.
4
Cold customers (full eKYC) Later
Highest risk, biggest upside.
Outcome: Segment paths built in priority order (Fixed, Returning, Referred, Cold). Decisioning service powers the adaptive funnel.
Wave 3 Months 4-6

Referral program

  • Extend Optimum's existing Fixed-side referral program at refer.optimum.com ($50 each, vested at 60 days). Scope the Mobile path.
  • Pilot with bundled customers referring friends or family to add Mobile. Lowest-risk audience for a new program.
  • Dual-incentive vesting: small credit at activation, larger credit at 90 days, both parties.
Referral SMS sequence 1 of 3
Optimum Mobile Text Message · SMS
Your Optimum Mobile invite to Maya is sent. You both earn $10 at activation, $20 more once Maya hits 90 days on network.
Maya just activated Optimum Mobile! $10 referral credit applied to your next bill. $20 more after 90 days on network.
Maya hit 90 days on Optimum Mobile. $20 referral credit applied to your next bill. Thanks for the referral.
Outcome: Referral becomes a measurable acquisition channel with built-in retention incentive.

Dependencies

Five teams own pieces of this. Each one is essential to at least one wave.

Engineering

Modular front-end architecture. Decisioning service. Component-level analytics infrastructure.

Credit & Collections

eKYC vendor selection. Deposit policy by tier. Risk-tier definitions and audit logging for the model.

Legal

Honest messaging today (TCPA). Required disclosures when declining credit (FCRA). Readiness for the upcoming FCC Revoke-All rule.

Content (Mobile site)

Chat routing logic and copy. Coverage-check standalone landing page. Plan-comparison clarity on throttle behavior.

Network Engineering

Device compatibility messaging during the 5G SA migration. Coverage data accuracy at zip-code level.

Risks

Modular complexity
Strict component ownership. Each component has a single PM owner. Shared analytics so cross-component effects are visible.
Referral fraud
Identity-tied accounts plus 90-day vesting. Both parties have to be real and stick around to collect.
Bundle vs. standalone bet
Build for both scenarios. Same components serve bundle and returning customers today, standalone signups tomorrow.
Regulatory changes
Getting compliance wrong means fines and forced product changes. Store consent rules (what customers see, when, how they revoke) as settings, not code. When rules shift (like the upcoming Revoke-All rule), we update settings instead of rebuilding features.
Where I'd Dig In

First-week priorities:

  • 90-day churn data, cut by segment. So we have a real baseline instead of an industry one.
  • Which compliance rules are locked, and where we have flexibility. Some are fixed (the exact language we must use when declining credit), others are policy choices (how and when we ask for consent). Tells me how aggressive the risk-tier system can be.
  • Which channels are in scope (online, retail, B2B), and what it costs to onboard a customer in each. Different channels need different onboarding flows, and the cost difference between online and in-store onboarding tells us where to push customers.
Section 5 · AI Usage Disclosure

How I Use AI

AI is a research accelerator and a drafting partner. I'd use it the same way on Optimum's product work: speed up discovery, stress-test framing, surface counterarguments, and produce drafts I review, edit, and own.

01

Post-interview note capture

I did: Brain-dumped fresh after the onsite, set the structure, edited for tone and emphasis.
AI: Structured 4 sessions of raw debrief into organized notes by session.
02

Problem space exploration

I did: Generated the ideas from interview cues, set the categorization criteria, decided which threads to develop.
AI: Categorized brainstormed ideas by usefulness for the exercise, sorted by problem stage and segment.
03

Competitive and industry research

I did: Identified the gaps that needed research, ranked which findings to use, committed to specific numbers in the brief based on what I could defend.
AI: Verified claims about Optimum's pricing position, the T-Mobile 5G SA migration, eKYC industry standards, FCC and TCPA regulatory context, and churn benchmarks across multiple sources.
04

Drafting the brief

I did: Defined the recommendation, picked the "three approaches that fit together" framing, made the call on what stays and what goes. Every claim, example, and number is mine.
AI: Assisted with framing against my outline and pushed back on my framing when I asked for it.
05

Building this interactive document

I did: Decided to ship an interactive version alongside the doc, designed the funnel widget concept, set the brand styling rules, reviewed every tab for tone and accuracy before sign-off.
AI: Wrote the HTML, CSS, and JavaScript for this site (including the funnel widget) against my design specifications. Iterated tab by tab with me reviewing each.