cx-channel-strategy

Use to decide which support channels to offer, add, or retire, and to audit channel mix against accessibility obligations and cost to serve. Trigger for "should we add chat", "which channels should we

SKILL.md

GitHub

Channel strategy

The usual failure mode is adding channels because customers ask for them, without a migration plan, a cost model, or a reason the existing channels failed. You end up with five ways to reach support, the same backlog, and higher cost — because each new channel is staffed, instrumented, and trained separately while demand never moved.

Channel strategy is not "be everywhere." It is matching contact routes to customer need, accessibility obligation, and what you can afford to run well.

Start with demand, not preference surveys

Question Why it matters
Where do contacts actually arrive today? Volume and driver mix by channel
Where do customers go when channel A fails? Leakage, not deflection
Which drivers are channel-specific vs universal? Phone for complex; chat for quick
What is cost per contact by channel? Async is cheap; voice is not
Who cannot use your current channels? Accessibility gap, not "nice to have"

A "customers want WhatsApp" survey without why they want it often means chat wait times are bad or the help centre failed — not that WhatsApp is the right fix.

Cost to serve by channel

Use marginal cost per contact — variable agent time and vendor per-contact fees — not fully loaded cost that includes fixed overhead. Savings are realised only when capacity or headcount changes.

Channel Typical cost driver What "cheap" hides
Self-service / KB Content maintenance, search tooling Contact-after-view failures
Async (email, messaging) Handle time × concurrency Slow resolution, repeat contacts
Live chat Concurrent sessions + SLA pressure Leakage to voice when queues spike
Voice Agent time + telephony + occupancy Highest cost; often legally required
Social / DM Brand risk + manual triage Public escalation, no identity

Rule: If you cannot staff a channel to your SLA without stealing from another, you do not have a channel — you have a queue that teaches customers to switch.

Accessibility and reachability obligations

This is not legal advice. It is an operations checklist for whether support is actually reachable by people with access needs.

Need Channel implications
Vision Screen-reader-compatible help centre; no image-only instructions
Hearing Text/chat/email; relay services for voice if offered
Speech Text alternatives; do not force voice for identity
Motor No drag-only widgets; keyboard-navigable contact forms
Cognitive Plain language, predictable paths, callback instead of hold
Language Human or quality MT in primary contact languages

Voice-only dead ends — "call us to reset your password" — fail multiple needs at once. If you offer voice, you still need an async path for people who cannot use it.

Hours matter as much as channels. 24/7 chat with 9–5 agents is a channel on paper and a bot-or-nothing experience in practice.

When to add a channel

Add only when all of these are true:

  1. Documented failure of existing routes for a specific driver segment — not anecdote.
  2. Migration plan — how volume moves, what gets de-emphasised, timeline.
  3. Staffing and SLA model — headcount or vendor line item, not "the team will absorb it."
  4. Instrumentation — same customer identity across channels so leakage is visible.
  5. Content parity — macros, KB, and bot flows updated before launch, not after.

When NOT to open a channel

Situation Why not
Existing channel under-served Fixes queue; adding routes splits demand
"Competitors have it" Their cost base and customer mix differ
Bot containment looks low Leakage may be to another channel you already pay for
No identity resolution You cannot measure cross-channel repeats
BPO contract minimums New volume may not reduce spend
Launch without suppression rules Outbound + inbound on same issue doubles contacts

Retiring a channel is valid when volume is low, cost is high, and an accessible alternative exists — with a communicated migration period, not a overnight redirect.

Channel-role matrix

Assign each channel a primary job. Overlap is expensive; gaps are worse.

Role Best fit Poor fit
Deflect simple, high-volume KB, bot, async templates Voice
Complex troubleshooting Voice, screen-share, callback Twitter DM
Account-sensitive Authenticated chat/email Public social
Status / incident Status page, banner, macro Individual ticket replies
Sales-adjacent Chat with routing rules General support inbox

One channel should own proactive outbound for a given trigger; otherwise customers get email and push and SMS about the same shipment.

Traps

Channel switching as success. Chat "containment" rises while phone volume rises. Measure total contacts per customer per issue, not per-channel metrics.

Async as infinite buffer. Email without a backlog cap becomes a warehouse of delayed dissatisfaction. Worse CSAT than chat with a honest wait time.

Social as support inbox. Public channels attract performative escalation. Route to private channels fast or pay twice — public reply plus ticket.

Free tier chat widgets. Unstaffed chat trains customers that support is unavailable; they phone angry.

Present results to the user

  1. Current channel mix — volume, top drivers, and cost per contact by channel (with cost basis stated).
  2. Leakage map — where customers go when a channel fails; quantify cross-channel repeats on the same issue.
  3. Accessibility gaps — who cannot complete support journeys today; name the barrier (CAPTCHA, voice-only, hours, language).
  4. Recommendation — add, fix, merge, or retire; each tied to a driver segment and a staffing or content dependency.
  5. What not to do — channels or launches that would increase cost without migration, called out explicitly.
  6. Measurement plan — identity resolution, SLA by channel, and total-contact metric alongside channel KPIs.