Timed Product Case Coach

BridgeDesk decision memo and oral-defense card

Fictional practice demo — coached artifact revision. This public demo was compressed and its economics repaired after the initial run and subsequent evaluator-material access. It is not a fresh blind evaluation. The exact initial bytes are preserved in qa/timed-coach-initial-blind-demo.md. Both versions use only the supplied synthetic case evidence. Neither establishes an actual 90-minute execution or literal rendered two-page fit.

Decision memo

Approve the small-account self-service activation pilot for next quarter. Use the six engineer-weeks ($18,000) to clarify required setup fields through the existing checklist and validation. Help operators complete a useful, verified workflow; capture value only through additional paid account-months after costs. Confidence is moderate in this learning choice and low in its financial return. The VP Product is accountable; product and engineering own scope, data owns measurement, and finance validates contribution. Authorize the bounded pilot now, with continuation decided after the final participant reaches day 28.

Why this customer and mechanism

Small activation is 80/200 = 40%; midmarket is 105/150 = 70% (Exhibit 2). Five of eight small-account interviews describe setup-field uncertainty (Exhibit 3). This suggests a removable obstacle, not representative prevalence or a proven cause. Experienced operators may encounter extra guidance; make it dismissible.

Assisted midmarket onboarding is the strongest alternative: four of six interviews wanted workflow-mapping help, and midmarket's lost annualized month-12 subscription run rate is $486,000 versus small's $240,000 (Exhibits 1/3). These are run-rate differences, not cumulative lost revenue. Onboarding and retention cohorts cannot be joined. Challenge the assumption that lower activation automatically warrants priority: self-service leads because its obstacle is bounded and its permitted learning population is larger, while assisted could produce greater paid-use lift.

Economics: capacity and actual test footprint

Use

12-month incremental contribution dollars per funded option, relative to the current approach, after one-time delivery and the $18,000 fixed investment. Let L be incremental paid months per treated account, net of churn/cannibalization; m be subscription margin after ordinary hosting, payment and recurring support; c_s be extra self-service delivery/support cost per treated account; and d_m assisted delivery cost. Lift, margins and self-service cost are unknown. Use supplied current assisted cost $180 as a planning input; actual pilot cost can differ.

The test conservatively counts both arms within the 100/30 pilot limits: 50 treatment/50 control small accounts, or 15/15 midmarket. Only self-service runs. Full-capacity rows are comparison scenarios, not the selected pilot's forecast. Using $180, assisted variable delivery totals $5,400 at capacity or $2,700 in the test; self-service remains 100c_s or 50c_s.

Break-even for N treated accounts is L_s ≥ (18,000/N + c_s)/(100m_s) or L_m ≥ (18,000/N + d_m)/(300m_m), requiring positive margins. At

assumed 80% margins, c_s=$0 and d_m=$180, recovery needs 2.25/3.25 paid months at capacity, but

4.5/5.75 at the actual balanced-test sizes. Assuming c_s=$30 raises self-service to 2.625/4.875 months. These are sensitivities, not measured margins or outcomes; recovery over twelve months is not cash payback timing.

Assisted wins at capacity when 9,000m_mL_m − 30d_m > 10,000m_sL_s − 100c_s; compare the test rows separately for pilot recovery. Under the illustrative equal-margin/zero-self-cost assumptions, either footprint favors assisted when L_m > (10/9)L_s + 0.75. Paid-month lift is the decisive unknown. This pilot buys learning; profitable deployment is not established.

Scope, measurement and decision

Reuse verified field validation and checklist; clearer guidance is an extension. Human scheduling already exists for the alternative. Reliable assignment/outcome reporting, AI planning, import and reusable templates remain unverified. Check instrumentation before scope freeze; narrow guidance if measurement cannot fit six engineer-weeks. No AI planner or import platform is funded.

Hypothesis: clearer required fields increase first verified workflows by reducing setup uncertainty. Randomize eligible small accounts entering setup before exposure. Exclude internal/test accounts beforehand; deduplicate by account ID. Primary numerator: assigned accounts with a valid, persisted first workflow verified within 28 days of assignment. Denominator: all eligible assigned accounts in that arm, including abandonment/cancellation. Join authoritative workflow records to assignment logs and audit validity. Compare absolute treatment-minus-control percentage points, counts and an uncertainty interval, segmented by entry week. Historical 40% is context; concurrent control is the baseline. This gross behavior measure differs from net financial paid months.

Proposed rule: continue only to bounded follow-up if activation improves at least 20 points, incremental support cost is at most $30 per treated account, and no serious workflow incident occurs. Fifty accounts per arm cannot guarantee precision or statistical power; a threshold crossing with plausible harm remains inconclusive. Modify unclear positive results or cost overruns; stop for no improvement, unreliable measurement or serious harm, restoring the current experience. Data and customer success review integrity/costs weekly.

Review paid use and cancellation at 90 days; require credible net paid-month and contribution evidence at six/twelve months before scale. More completed workflows may increase support demand; track that burden. Reverse toward assisted if setup guidance fails, service costs become prohibitive, or credible assisted lift crosses the economic boundary. Target the day-28 decision by quarter-end only if enrollment finishes early enough; otherwise state the later date.

Private oral-defense card

30-second recommendation

Fund the quarter's one pilot to help small operators reach their first useful workflow through clearer setup guidance. The 40% activation baseline and directional setup uncertainty make a bounded learning opportunity using current capabilities. Randomize 100 accounts into 50/50 arms; a proposed 20-point lift earns only further bounded learning, with quality, cost and uncertainty checks. Compare paid-month contribution before scale. Assisted midmarket wins if its larger causal lift compensates for limited capacity and service cost.

Three critical assumptions

Fields are a removable obstacle. Five of eight interviews support the hypothesis, not prevalence. Inspect failed fields and task observations before scope freeze.
2.

Verified workflows create retained paid use. The supplied cohorts cannot test that link. Follow randomized assignment through net paid months, cancellations and costs.
3.

Scope and measurement fit six engineer-weeks. Validate instrumentation and estimates. Reduce guidance scope before weakening measurement or increasing the allocation.

Strongest rejected alternative and main risk

Assisted workflow mapping serves a higher-price segment with more lost subscription run rate and may resolve organizational complexity. It is credible; its causal effect is unknown. Its balanced 15-treated pilot has $2,700 planned delivery cost plus $18,000 fixed cost, versus unknown self-service delivery cost plus the same fixed cost. The main risk is improving a setup metric without useful retained work. Verify persisted outcomes, watch support demand and cancel/rollback for harm; activation alone cannot become a revenue forecast.

Five likely reviewer questions

Why small rather than midmarket? A bounded obstacle and broader learning sample favor small initially. Midmarket's observed economics make it the serious alternative, not evidence that assisted onboarding causes improvement.
2.

Which costs are unknown? Subscription margins after ordinary delivery costs, self-service incremental cost and actual pilot-assisted delivery cost. $180 is the supplied current assisted cost; 80% margin and self-cost examples are labeled sensitivities.
3.

Can this sample prove success? It can generate a constrained learning signal. Report actual counts and uncertainty; borderline or potentially harmful effects are inconclusive. No power or observed effect is claimed.
4.

What recovers the investment? Twelve-month incremental subscription contribution. With illustrative 80% margins and zero self cost, the selected 50-treated test needs 4.5 additional paid months each, versus 2.25 at 100 treated. Actual costs and effects may invalidate either case; neither gives cash payback timing.
5.

What changes the decision? No valid-workflow lift, quality harm or excessive service burden stops the pilot. Credible assisted paid-month lift above the common comparison boundary changes the priority. Mature retention cannot be proved at quarter-end.

Evidence notes

Observed counts, prices, capacities, capabilities and $18,000/$180 inputs come from Exhibits 1–3. Lost run rate is (starting − month-12 retained) × monthly price × 12: small (400−200)×100×12=$240,000; midmarket (450−315)×300×12=$486,000. No expansion, refunds or pass-through revenue is modeled. Diagnosis is inference; thresholds, enrollment interpretation and financial sensitivities are proposed assumptions. No new company fact, measured effect or external evidence was introduced by this revision.