Product decisions (triple-checked)
Framing, ladder order, and stacking re-verified against product UI, CS Confluence, and failure-code research.
Coach v2 — DO / DON’T (locked 2026-09-29)
Source of truth: COACH_V2_REMODEL + WHY_NO_ALPHA_UPSELL. Same UX for every account — no provider split.
DO
- One coach for every account: exclusive phones × ~2,000 msgs/phone/day · pace · window · duration
- Exclusivity is HARD — shared ACTIVE phones block Next (Replace / Pause required) · ACTIVE-only busy · Activate/Unpause re-check
- Ladder: Replace → Buy → sub → brand → (Full: raise daily only if trust headroom) → Slow last · show ≈ msgs/day → ~days · suggest a wider window
- Whales → Buy to cap, then Slow / longer duration / different brand (no migrate)
- Kill switch · never propose blocked numbers as unused · new camps only (no historical orange flood)
- Eng break-test proto · limits ref below · every other warn stays soft (Soft Next)
DON’T
- No Alpha / migrate / higher-capacity CTA anywhere — customer or Present talk
- No hard block on list size · no scheduler rewrite as v1 · no TPR/templates/FU UI in First
- No keyword in First · don’t lead with Slow · don’t offer raise-daily when planned TPM is already 102
- Don’t count paused as coach-busy · don’t panic-paint old list
- Don’t redesign from speculative interviews · no provider names or two-physics story in chrome
- Don’t count paused as coach-busy · don’t panic-paint old list
Are we ready? — meeting asks vs this proto
People’s questions from the room, marked against docs + live Pages. Ready for First design review / eng break-test on coach behavior. Not “production shipped.”
| Meeting ask | In doc? | In proto? | Status |
|---|---|---|---|
| What are the per-phone ceilings? | GENERAL + CAPACITY_DECISION | Yes — table below | Covered |
| Exclusive phones hard (Next blocked) · other warns soft | Yes | Yes (Create) | Covered |
| Same rules for every account (no route split) | Yes | Yes (Create) | Covered |
| ACTIVE-only · blocked ≠ unused · kill switch | Yes | Yes (Create tools) | Covered |
| Whale → Slow / duration / different brand (no Alpha) | Yes | Yes (Slow / days / brand CTA) | Covered |
| Code + tests for eng/QA | — | Tabs 8–9 | Covered |
| New camps only (no historical orange) | Yes | Intent / Present | Ship checklist (FE enforce at build) |
| ACTIVE flag ticket · Activate/Unpause wire | Yes | Designed | Eng wire (not a product debate) |
| Ask Commio (configurable DID caps?) | Noted | — | Ops (parallel; don’t block coach) |
| Map Vonage etc. → routes | Noted | Route model only | Ops |
Limits reference (eng) — same for every account
Internal numbers used by coach math. Customer chrome never names the provider. Campaigns get ~2,000 msgs per exclusive phone per day (campaign budget — not 3k × 0.9). Identical-body is not in First coach.
| Ceiling | Value | Rule in coach | In First coach? |
|---|---|---|---|
| Campaign day budget | 2,000 msgs / exclusive phone / day | phones × 2k = daily volume · shown as ≈ msgs/day → ~days | Yes → Buy / duration |
| Per-phone pace | ~15 / min (~900 / hr) | Planned TPM above phones × 15 → Slow | Yes → Slow gate |
| Delivery window | Default 8 h | msgs/day = min(phones × 2k, TPM × 60 × window) · <10 h + >1 day → wider-window tip | Yes → tip |
| Exclusivity | 0 shared ACTIVE phones | HARD — Next blocked until Replace / Pause / Buy | Yes |
| Identical body / hour | ~200 | — | No — out of First coach |
| Account phone cap | ~49 messaging phones (≈ 98k msgs/day) | Buy up to cap · then Slow / more days / different brand | Yes → Buy stop |
| Brand daily (trust tier) | Tier from trust score | Raise daily only if planned TPM < 102; else different brand / longer duration | Full (needs BE history) |
Sources: COACH_V2_REMODEL.md · WHY_NO_ALPHA_UPSELL.md · COACH_V2_STRESS_TEST.md (prefer 2k/phone campaign budget). Backend rails still differ internally — never shown in chrome.
Your calls (applied)
- All accounts / all providers? — One panel, one formula. Exclusive phones × ~2,000 msgs/phone/day · send pace · delivery window · duration. Same CTAs for everyone — no route toggle, no Alpha / migrate / higher-capacity CTA (
WHY_NO_ALPHA_UPSELL). - Coverage — shared ACTIVE exclusivity is hard for all accounts. Day + rate fleet math (2k/phone/day, ~15/min/phone) runs for all. No hard block on list size — only pace, duration, phones, window.
- Two kinds of Buy — Product min (1 · or 3 @ 2,500) and shared → exclusive are product / exclusivity proposals, not “you won’t deliver because you have fewer phones.” Fleet Buy / Slow (day + rate) now applies to every account, up to the ~49 phone cap.
- Next gate — hard only when selected phones are shared with another ACTIVE campaign. Every other warn is soft (Soft Next); Soft Activate shape unchanged.
- Delivery: First version (standard create + Activate) → Full version (list · Unpause · follow-up adapted · …).
- Follow-up adapted UI — Full version. First version still counts follow-up camps in the ACTIVE pool.
- Shared = ACTIVE only — paused leave the pool; refresh on create and before Activate / Unpause. Pool includes follow-up camps.
- No keep-shared — exclusivity is a hard rule; fix via Replace / Pause / Buy. (Hotfix ships to everyone, then learn from CS reaction.)
- Recommend all that’s needed — ladder order: Replace → Buy → seats → brand → templates tip → Slow (if rate-hot) → whales: more days / different brand. Show ≈ msgs/day → ~days; suggest a wider window when <10 h. Never buy phones just for planned TPM speed. Raise-daily only when planned TPM < 102 (trust headroom).
- Empty wallet — not in this panel; Buy opens the normal Buy/wallet flow.
- Sub TPR — Ask primary only.
- After Replace, no Add-sub upsell — unused-first; stay quiet once phones are fixed.
- Campaign names on chips — skip for v1.
- Brand-hit — toggle stand-in until BE history exists. At planned TPM 102 → no raise-daily; different brand or a longer run instead.
- Templates — soft tip only when under recommended (anchors 2.5k→13 … 100k→50 × phones; cap 50). Hide when looking good — no all-clear tip card. Same formula for follow-up (study = standard; one rule).
Product calls — all closed
Q1–Q5 settled. No open product picks left for First coach behavior.
| # | Decision | What ships |
|---|---|---|
| Q1 | C — FU day-short → Slow when it helps · else Pause other / accept longer duration (no Alpha) | Follow-up + phones locked + list too big for day · Slow only if it actually helps · else Pause other or longer end date (≈ msgs/day → ~days) |
| Q2 | A — Keep “Limits may drop texts” + Buy for brand-daily under product min | Same customer chrome. Docs note: this is a product-min rule, not delivery physics |
| Q3 | Exclusivity HARD · other warns soft | Shared ACTIVE phones disable Next · never hard-block list size |
| Q4 | Keep Buy before Slow on mid lists | Up to need <40 |
| Q5 | FE hides coach at 0 phones | Impl gate |
Still outside product (ops/eng): Ask Commio on DID caps · Vonage route map · ACTIVE flag ticket · Activate/Unpause wire · new-camp-only list paint. Not open questions — parallel work.
Closed earlier: templates cap 50 · Buy before Slow ≥2.5k · Buy up to ~49 · kill switch · blocked ≠ unused · no Alpha / migrate / higher-capacity CTA · 2k/phone campaign budget · raise-daily trust-gated.
Name + title + status words
Product: AI Campaign Coach — quiet grey caption above the panel question (locked).
Panel question (problem-first): “Will phone or daily limits drop texts?” + i (scope lives only here — checks vs doesn’t).
Statuses (create v1): Looks fine · Limits may drop texts · Can’t start (list).
Looks fine body: one summary + compact check facts with numbers. No second “OK” line. No footer repeating the tooltip.
Problem body: one summary + stacked actions. Scope stays in the i. Status colors stay.
Out of sends today — cut from create
Does it happen? Yes at runtime (brand daily hit → DAILY_LIMIT_REACHED; SC-9575 = pause + resume tomorrow).
Create coach? No. We have no brand-daily remaining at create — can’t honestly show “cap empty today.”
TPR? Not for same-day empty (CS + carrier tier takes days). Offer TPR on history hits only. Schedule tomorrow = runtime / later meter.
Templates = quality tip, not a limit fix
Separate yellow tip — problems only (under recommended). Hide when templates already meet the soft aim. Same contact×phones anchors for follow-up (study was standard campaigns; no separate FU rule — keep one formula). CTA opens Message templates picker.
Sub-account (verified Confluence permissions)
Can: buy messaging phones · pick/replace phones · templates · slow send · create campaigns.
Cannot: add/manage sub seats · access Sub accounts page.
TPR: not self-serve on sub → Ask primary only.
Different brand: Ask primary · chat support.
Brand daily (create) → trust-gated raise · different brand · longer run
Create path = hit daily in last 30 days. Planned TPM < 102 → Talk to support · raise daily and/or different brand. Already 102 → different brand or run longer — never Alpha. Subs never raise T-Mobile daily.
Never offer TPR for shared-phone / DID overload alone — TPR does not add numbers.
Sub-account
See permissions box above. Do not blanket “Contact primary” for Buy phones — subs can buy when messaging is active.
Fix ladder (numbers first)
- Replace shared with unused numbers — remove all shared from this campaign; select unused you already own. If unused < phones needed, also offer Buy (same click path: replace first, then buy shortfall).
- Buy phone numbers — also strips any remaining shared. After buy, if no shared left → just “Add phone numbers” / Looks fine.
- Add free sub account when free seats left (Elite: first 2); else Add sub account paid (3–5); at max 5 → different brand
- Different brand / EIN — after sub max, or for T-Mobile brand daily
- Add message templates — quality tip only
- Slow the send — only when no unused phones and no sub seats left (not next to Use unused / Add sub)
Schedule tomorrow = runtime / later live meter — not create v1.
Stack in order. Button under each. Prefer adding capacity over slowing.
Sub-account vs 2nd brand (verified)
Product CTA is a button: “Add sub account” / “Add free sub account” — not a text callout.
Subs help DIDs + camps; AT&T TPM scales. T-Mobile brand daily is shared across all subs of a brand — CS Confluence explicitly: don’t sell subs for T-Mobile; use separate EIN. PlanOptions: different brand for SMS campaign limits.
Templates — include, but don’t confuse with phones
More templates help message variety / delivery rates (product copy already says so; min by list size). They do not raise per-phone hourly or T-Mobile daily the way more numbers do. Show when under the existing minimum — before slow, after capacity upselle.
Scenario demo
Toggle multiple issues on/off so actions stack.
What this check covers vs what it doesn’t
| In scope (this check) | Out of scope (other drops) |
|---|---|
| Shared ACTIVE phones (hard) · phones × 2k/day · duration | Spam flags on number/content (607) — templates help variety |
| Account send speed (maxTpm) · slow send · delivery window | Dead/unreachable numbers (351 — most common code) |
| T-Mobile brand daily (history) | Opt-outs, DNC, recipient blocks |
| Sub seats / buy phones / different brand | Flooding one contact (4103/4104), Safety Block, non-US |
“Looks fine” = no phone/daily-limit issue for this campaign — not “every text will deliver” (spam, flooding, etc. are out of scope).
Create campaign (product layout)
AI Campaign Coach — switch scope below to demo First vs Full.
Pause / end / re-activate to refresh which numbers are busy. PAUSED & ENDED do not count.
Same rules for every account: exclusive phones × ~2,000 msgs/phone/day (campaign budget). No provider / route split.
ACTIVE pool includes standard and follow-up camps (even in First — they load busy phones). Flip Weekend follow-up to ACTIVE to demo collision. Shared ACTIVE phones disable Next.
Full: follow-up adapted · templates tip · brand tip · Unpause · list on tab 4. Next hard-blocks only on shared ACTIVE phones; every other warn is soft.
Details
Message
Campaigns list
Full version only. Same statuses as create: Looks fine / Limits may drop texts / Can’t start — includes follow-up rows. Not in First version (use tab 3 · Create → First version).
| Campaign | Type | Status | Limits | Contacts |
|---|---|---|---|---|
| Spring blast | Standard | Active | Limits may drop texts | 12,400 |
| Weekend follow-up | Follow-up | Active | Limits may drop texts | 890 |
| March leads | Standard | Draft | Can’t start | 3,200 |
| Old promo | Standard | Completed | — | 5,100 |
Column = phone/daily limits only — not spam or flooding. Completed stays neutral. Follow-up can warn when locked phones overlap another ACTIVE camp.
Present AI Campaign Coach
Visual slides for the room. Live UI: tab 3 · Create → switch First version / Full version on the page · list = tab 4.
First version = ship fast (phones / shared). Full version = tips + list + follow-up. All = demo recipes. First demos: 1–3 · 6. Full also: 4–5 · 7–9.
First version — ship fast
Solve the core problem first: don’t let people start standard campaigns that will drop texts on shared / too-few phones. Shared ACTIVE phones block Next · every other warn is soft.
Pre-carrier drops
- Shared phones across ACTIVE camps
- List / speed bigger than the fleet
- Create is silent today
AI Campaign Coach · standard create
- Hard Next block on shared ACTIVE phones · other warns soft (Next allowed)
- ACTIVE busy phones (incl. follow-up camps in the pool)
- Replace unused → Buy if short → Slow last · shows ≈ msgs/day → ~days
- Soft Activate gate (same panel)
- Whales: Buy to cap, then Slow / longer duration / different brand (see below)
Follow later in Full
- Templates quality tip
- Brand-history CTAs (needs BE)
- Campaigns list column
- Unpause · follow-up adapted UI
Same rules for everyone: exclusive phones × 2k/day · pace · window · duration
One panel, one formula, every account. We don’t teach two provider physics in chrome — the coach speaks phones · day pace · window · trust.
Hard rule
- No phone on two ACTIVE campaigns (incl. follow-up)
- Next is blocked until Replace / Pause / Buy
~2,000 msgs / phone / day
- Campaign budget — not 3k × 0.9
- Shows ≈ msgs/day → ~days
Slow · wider window
- Buy before Slow (up to ~49 phones)
- Window < 10 h → suggest a wider one
Only if it can actually rise
- Raise daily only when planned TPM < 102
- At 102 → different brand or run longer
Two kinds of Buy: product min / exclusivity are proposals, not “you won’t deliver because you have fewer phones.” Fleet Buy (day + rate) now applies to everyone, up to the account cap. No hard block on list size — only pace, duration, phones, window.
Buy · Slow · duration — whales stay in the same product
~2,000 campaign texts/day per exclusive phone. Account phone cap ≈ 49 → ~98k same-day texts on one account. Past that, list size is not blocked — it just takes more days.
Lists that fit under the cap
- Exclusive phones + enough fleet / Slow → traffic fits
- Examples: 12k–100k lists → Buy / Slow / a few days
Slow · longer duration · different brand
- Buy to the cap, then slow the send and show “~Y days”
- Different brand / EIN when that is the real lever
- No migrate / no provider pitch in the coach
Does this solve the core ask?
- Stops most dumb starts when people follow advice
- Fast to ship · reuses Buy / Slow / phone picker
- Exclusivity is hard; pace / window advice is soft (they can ignore)
- Scheduler can still stack ACTIVE camps
- No per-number runtime pacer yet
Other solutions we checked (no easier complete fix)
If someone asks whether something simpler fully stops drops: we checked — there isn’t. Say these two clearly:
Real complete fix for per-phone day walls
- Scheduler paces each phone · shares capacity across ACTIVE camps
- Big BE / scheduler work — not a small FE coach
- Follow-on after Coach (or parallel if eng capacity)
Runtime catch
- Pause when limit failures fire mid-send
- Customers often dislike surprise pauses
- Fits Tracker / Insights later — not Coach create
Hard block is only for shared ACTIVE phones (exclusivity hotfix, ships to everyone). List size is never hard-blocked; every other warn is Soft Next.
First version — locked cut
- Name: AI Campaign Coach · quiet grey caption
- Exclusive phones hard (Next blocked) · other warns soft · ACTIVE-only busy · standard create + Activate
- ACTIVE pool includes follow-up camps (they load busy phones for standard create)
- Same rules for all accounts · ≈ msgs/day → ~days · no Alpha / migrate CTAs
- Keyword / TF still out
Live demo: tab 3 · Create → First version, Recipes 1–3 · 6 (skip 4–5 templates/brand — Full).
Full version — complete AI Campaign Coach
Everything in First version, plus templates tip · brand tip · list · Unpause · follow-up adapted · drip caveat.
- Campaigns list statuses (standard + follow-up rows)
- Unpause — same check vs ACTIVE pool
- Follow-up create — adapted (Pause other · Slow when it helps · longer end date · no phone Replace/Buy)
- Templates quality tip (not a phone-limit fix)
- Drip = same panel + day/Slow caveat
- Brand-daily history → TPR / different brand (when BE signal exists)
- Keyword create · toll-free specials
- Customer mid-send limit notify (needs live edit)
- Runtime per-number pacer (separate big project)
- Standard and follow-up camps when ACTIVE
- Paused / ended don’t count as busy
- Shared ACTIVE phones — hard block (exclusive rule)
Campaign types — full scope
Full ladder
Same panel + caveat
Adapted · phones locked
Deferred
Transparency ≠ changing phones on follow-up — show risk; don’t offer Replace/Buy that swap locked numbers.
Full version — locked target
- First version + templates tip + brand tip + list + Unpause + follow-up adapted + drip caveat
- Exclusive phones hard · every other warn Soft Next
- Live demo: tab 3 · Create → Full version · Recipes 1–9 · list on tab 4
All — demo recipes
First: Create → First version, Recipes 1–3 · 6. Full: Create → Full version, also 4–5 · 7–9. Coach proposal is always on — only switch First / Full. No provider names in the room — say “from-number limits” or “brand daily.”
Recipe 1 — Shared phones main story
Expect: Can’t start · Next disabled · Use numbers you already own (required)Shows: ACTIVE busy · Next blocked · Replace unused · (Buy if still short after). Say: paused camps don’t count as busy; every campaign needs its own phones.
Recipe 2 — Need more phones
Expect: Limits may drop texts · Add more numbers (Buy)Shows: list/speed needs more dedicated phones than you have · Buy in the ladder. Optional: click Slow only if it appears after Buy path.
Recipe 3 — Looks fine
Expect: Looks fine · exclusive phones × 2k/day · ≈ msgs/day → ~daysShows: quiet all-clear with ≈ msgs/day → ~days (First — no templates tip). Hover i once for scope.
Recipe 4 — Templates tip (quality) Full
Expect: yellow Delivery quality tip (may also show limits if other issues on)Shows: tip is separate from “Limits may drop texts.” Not a hard gate.
Recipe 5 — Brand-daily history Full
Expect: Limits may drop · Raise daily (TPM < 102) or different brand · at 102: different brand / run longerShows: trust-gated raise-daily / different brand. Say: never offer raise-daily when TPM is already 102, and never for shared phones. Never Alpha.
Recipe 6 — Next gate (hard vs soft)
Expect: shared phones → Next disabled · other warns → Next enabledShows: exclusivity is the only hard gate; everything else is soft.
Recipe 7 — Drip caveat (talk only · same settings as recipe 1)
Expect: same panel as standard · say the caveatDon’t demo a separate drip UI in this proto. Product keyword: Add message / drip steps on a standard campaign.
Recipe 8 — Follow-up adapted (shared phones)
Expect: Can’t start · Next disabled · Pause other camp (required) · Slow only when it helps · no Replace/BuyShows: same busy check · phones locked chips · Pause other (required — Next blocked). Slow when it helps, else longer end date (≈ msgs/day → ~days). Never Alpha. Say: transparency, not “change the follow-up numbers.”
Recipe 9 — Standard draft vs ACTIVE follow-up
Expect: Can’t start · Next disabled · Use numbers you already own (standard ladder)Shows: ACTIVE pool includes follow-up camps · standard create still gets full Replace/Buy ladder.
Uncovered — say it, don’t demo as “broken coach”
- Other systems · out of *i* scope
- No live remaining meter at create
- Separate create flow · defer
- Different min · Buy may be hidden · defer specials
- Same panel OK for busy/Replace · Slow/day math imperfect · caveat
Three AI surfaces
Template Guardian · AI Campaign Coach · Campaign Tracker. Same family, different moments. Later Tracker ideas are mapped below with dependencies — not this coach ship.
Content · before save
- Checks message / template risk
- Already shipping (separate product)
Setup · before / at start
- Create side panel · Soft Activate
- Unpause · list statuses · follow-up adapted = Full version
- Shared ACTIVE phones block Next · other warns are soft
Runtime · while sending
- Watches delivery / opt-outs / errors
- Safety Block can pause on bad rates
- Different job from Coach phone/daily setup
AI Campaign Coach — First → Full
- Create panel · exclusive phones hard · other warns soft
- Soft Activate
- ACTIVE pool includes follow-up camps
- Replace → Buy → Slow on standard
- Unpause re-check
- Campaigns list statuses
- Follow-up adapted create
- Templates tip · drip caveat · brand tip
Exclusivity = hard (Next blocked). Everything else = soft warn, they can still proceed.
Later · improve Tracker (ideas + dependencies)
Not in Coach create ship. Map only — so we don’t confuse “notify” with Coach.
Insights-style + fix copy / links
- Today: since Aug 2024, an internal email goes to #dev-monitoring when brand-daily / send-limit failures are classified — not the customer
- Idea: customer sees the same story + Coach-style fixes (raise daily · different brand · resume tomorrow)
- Dependency: useful fix buttons need live campaign edit (or clear pause/resume path). Without that, notify mostly annoys — they’d redo the campaign
- Fits as a Tracker / Insights bundle, not Coach create
“This number on this campaign hit a wall”
- Message rows already store campaign + from-number + raw error code
- But from-number route often does not classify those codes into “limit reached” (so today’s limit email path rarely fires)
- Dependency: map provider error codes → limit type, then notify; plus live edit before Buy/Replace mid-send helps
Optional watch while camps are live
- Coach v2: block at create (hard); re-check at Activate / Unpause (Soft Activate shape)
- Later: mid-send watch if legacy ACTIVE camps still share phones
- Not required if people follow dedicated-phone advice
Tracker today
- Pauses on low delivery / high opt-out / high block-error rates
- Don’t paste Coach “Buy phones” on those pauses
Alignment cheat sheet
- Guardian = content before save
- Coach = First (create + Activate) → Full (+ Unpause + list + follow-up adapted)
- Tracker = runtime Safety Block (+ later limit notify after live edit)
Coach flows
Which checks run, what status you get, and which CTAs fire — same rules for every account, by create type. Click any box for detail, or pick a scenario and open it live in Create.
Pick a box or scenario
Grey boxes don’t apply for this create type. Blue highlight = selected. “Open in Create” jumps to tab 3 with matching knobs.
Code for developers
All coach math + presentation sources used by this prototype. Copy or download. Regenerated at deploy via capacity_design/scripts/embed_dev_bundle.py.
Loading…
If this stays empty, open from the deployed Pages build (includes dev_bundle.js) or run python3 capacity_design/scripts/embed_dev_bundle.py.
Tests for eng & QA
Unit / matrix / simulator tests that lock the ladder. Same download/copy pattern. Also: blocked numbers must never appear in unused; kill switch hides the coach panel.
Run locally
Loading…