Twenty-one capstone scopes modeled on real GB410 projects, read through the five SOLVE questions. Four recurring families of scope; SOLVE applied to every project; where students predictably go wrong, letter by letter, illustrated with composite team paths; and the simulation experiences in GB210 that would let them practice those moments first.
Worked examples
Six walkthroughs
These follow the shape of the teaching note: what students do at each question, what they should discover, where they predictably stall, and the coaching question that moves them. Each ends with a loop worth telling in class and a note on how the project would translate into a GB210 simulation. Numbers follow the source scopes but have been rounded or altered as part of the disguise; treat them as illustrative. The team paths described are composites of how capable teams tend to approach such projects, not records of any actual team.
The setup. Kestrel deploys robotic systems to manufacturers and maintains them. Parts and spare components are tracked through work orders and maintenance-tool reports; requisitions travel over Slack. A new ERP is going live. The finance team asks for a recommended inventory management system that tracks parts across statuses, integrates with a new ERP and its maintenance tool, supports barcode or mobile scanning, and comes with process documentation, requirements, a comparison of at least three vendors, a trial plan, and a roadmap.
S
Surface the real problem What's the ask behind the ask?
The ask is a system. The problem is that nobody can say where a given part is, so installs and repairs wait on parts that exist somewhere, purchases get duplicated, and Kestrel cannot prove what it returned from a decommissioned site. Students should restate the problem in those terms, attach a bottom line (repair and install turnaround, parts written off or billed back, technician hours spent hunting), and get the CFO to agree that a system is one candidate answer, not the question. The baseline is the current mess, measured: how many hours a week, how many dollars of unaccounted parts last quarter.
The predictable failure is accepting "compare three vendors" as the problem statement and spending the semester on a feature matrix.
Coaching question: If the perfect system were installed tomorrow and nobody logged anything, what would change?
O
Observe the behavior Whose behavior has to change?
The behavior is a technician or integrator logging a part at each handoff: when it arrives, when it is requisitioned, when it is installed, when it comes back. Today that happens late, inconsistently, or over Slack. Any solution, software or not, lives or dies on whether people in a hurry at a customer site take thirty seconds to scan or type. That makes field technicians the customer, and the CFO a sponsor.
L
Learn the drivers What drives that behavior today?
A journey map of a part through the system lifecycle. Imagine a team that does this well: six process maps covering acquisition from integrators (stocked and build-to-order), preventive and break-fix maintenance, in-place upgrades, and decommissioning, built from interviews. The blockers such a team finds are the right ones: systems logged into the maintenance tool too late, undocumented recurring parts, Slack-based requisition, dependency on one person, no verification step at decommission. SOLVE would add quantities to each blocker (how many parts, how many hours, how many dollars) so the branches can be ranked.
Coaching question: Which of these six maps carries the most money? Go deep there and leave the rest as sketches.
V
Vet the shakiest claim What's the cheapest way to find out we're wrong?
Serious options: (a) buy and integrate a dedicated inventory system; (b) enforce standardized touchpoints and a requisition form inside the tools Kestrel already has; (c) do (b) now and (a) when volume justifies it. For option (b), which is where that team lands: the rules must be codifiable as a checklist, technicians must comply in the field, and the existing tools must hold the data well enough to answer "where is part X." Shakiest: field compliance.
The cheapest test is an arrival-intake checklist at one integrator site for two weeks, with compliance counted and the first "where is it" question answered from the log instead of from the one person who knows. That test would also produce the number the CFO needs to decide whether to buy a system at all.
E
Earn the yes What should Kestrel do Monday morning?
Either a vendor recommendation that names the process changes it depends on, or the process changes alone with a stated trigger for when a system becomes worth buying. In both cases the argument is heavy on adoption: who owns each touchpoint, what happens when a tech skips a step, and what the first month of monitoring looks like. Value quantified in technician hours and parts dollars, with a range.
The loop the team lives but does not narrate. The scope asked for vendor rankings. The team discovers during mapping that the process is not ready for any vendor and re-scopes to five practical fixes. That is S reopened by L, which is the method working. But its final presentation never says "we were asked for X, here is why we are recommending Y first, and here is what we would need to see before X." A one-slide project log would have turned a silent deviation into the most credible moment of the presentation, and would have protected the team from a sponsor who still expected three vendors.
As a GB210 simulation. Opening prompt from the Director of Finance: "we need a vendor comparison for an inventory system by week three." Client personas: the finance director (wants the comparison), an operations lead who, only when asked how a requisition actually happens, says "we Slack one guy," and a field technician who describes a repair visit where the part was somewhere but nobody knew where. Designed discovery: the integration requirement is premature because half the parts never enter any system. Students should leave with the scope renegotiated in writing with the finance persona, a quantified baseline (hours, parts dollars) pulled from the operations persona, and a designed two-week intake-checklist pilot they would propose to the sponsor.
The setup. A student-run residential snow removal service at about ten universities, serving several hundred homeowners, with a mobile app and an admin dashboard built by the CTO. The founders list six asks: an in-app job workflow for shovelers, a dispatch notification approach, remote onboarding and training, a ratings system, a process to reconfirm that a shoveler still intends to work 24–48 hours ahead, and routing with backups at scale. Data offered: snowfall and refund history, pricing, before-and-after photos, shoveler testimonials, management feedback.
S
Surface the real problem
Six asks, one bottom line: jobs completed on time per storm without a founder stepping in, and refunds avoided. The background says so directly ("time-intensive operational challenges"). Students should get the founders to state the baseline (what share of jobs need a manual intervention in a storm; the refund rate and its causes) and to rank the six. The commitment-confirmation ask is the one that most plausibly predicts the rest, and the ratings and onboarding asks are downstream of it.
Coaching question: In the last big storm, how many jobs needed a human to fix something, and what went wrong in each?
O
Observe the behavior
A shoveler confirms 48 hours ahead and then shows up and finishes; a customer receives one clear notice and does not call. Confirmation reply rate and no-show rate are the leading indicators; refunds and founder hours per storm lag. Customers of this project include the student shovelers themselves, which is where adoption gets heavy: every design asks a college student to do something at 1:40 in the morning.
L
Learn the drivers
A journey of one storm: forecast watched, routes built, shovelers notified, confirmations (or silence), travel, work, photos, customer notification, refund requests. The refund-by-weather data and the management feedback locate the blockers; interviews with shovelers explain why confirmation gets ignored (notification at 1:40 a.m., unclear expectations, no consequence). The predictable failure is designing screens for all six asks; the coaching move is to follow one storm end to end on one campus.
V
Vet the shakiest claim
"We believe a 48-hour reconfirmation prompt with a visible consequence will raise show-up rate. We'll know when the reply rate exceeds X% and no-shows fall on the piloted campus relative to the others." Must-be-trues: shovelers see and answer the prompt, a backup can be assigned in time, the consequence is enforceable. Shakiest: that students reply at all. Cheapest test: a plain text-message reconfirmation on one campus for two storms, no app change, with the founders' manual process as the control. A team might propose an A/B test on recruiting channels, which shows the habit; the version above tests the behavior the scope actually asked about.
E
Earn the yes
Workflow designs for the confirmation-and-dispatch path, the pilot results, and a rollout argument that makes the new path easier than ignoring it (one tap, a reason to reply, a consequence). Value in founder hours per storm and refunds avoided. The other four asks get a prioritized backlog with the reasons.
The loop a team makes. Picture a final presentation that shows an initial scope (the six asks) and a revised one (recruitment insights, struggling-campus support, website and SEO changes, app UX polish, notification timing). That is a re-scope, and some of it is well observed: the 1:40 a.m. email is a genuine finding. But a revision that is not tied to a metric or signed off as a restated problem reads as drift rather than method. SOLVE's house rule, document your loops, would have made the same work look like disciplined narrowing.
As a GB210 simulation. Opening prompt: the six asks, from a founder who wants screens designed. Client personas: the founder, a CTO who wants a routing algorithm, and a customer-service lead who knows the no-show and refund problem by heart. Designed discovery: the refund data correlates with no-shows, not with snowfall, and the dispatch email went out at 1:40 a.m. Students should leave with the six asks ranked and agreed with the founder, one storm mapped end to end from the service lead's account, and a designed text-message reconfirmation pilot for one campus. A good simulation for teaching that six asks are a ranking exercise, not a to-do list.
The setup. A direct-to-consumer seafood subscription with a loyal base that skews old: roughly two-thirds of subscribers are over 55 and a large share over 65. Leadership wants 25-to-35-year-olds and suspects some mix of price, portion, lifestyle fit, and awareness, but says plainly that it does not know whether the problem is product fit or market fit. The deliverables asked for are an insight report, product concepts, a go-to-market plan, and a directional financial model. Data on offer: subscriber analytics, preference surveys, competitive analysis, sourcing criteria.
S
Surface the real problem What's the ask behind the ask?
The ask is clean, but there is still a restatement to earn. "Attract, convert, and retain 25–35-year-olds" is three behaviors and a values constraint. Students should pin the bottom line (new subscriber growth and lifetime value in that cohort), the baseline (how many current subscribers are under 35, and how long they last compared with the 55-plus base), and the constraint (no compromise on sourcing). They should also ask the question the scope dodges: is this cohort worth it? If a 30-year-old subscriber has half the lifetime value of a 65-year-old, the answer changes shape.
Coaching question: If this worked perfectly, how many under-35 subscribers would Tidewater have in twelve months, and how many does it have today?
O
Observe the behavior Whose behavior has to change, measured how?
The target behavior is a 25-to-35-year-old who cooks at home starting a subscription and still being subscribed at month four. Two leading indicators make it manageable: the trial-to-second-box rate (does the first box get cooked?) and the cost to acquire a trial. Students should resist stating the behavior as "younger people become aware of Tidewater." Awareness is not something you can watch someone do; placing a second order is.
The predictable failure here is building personas from generational trend reports. The scope's own framing ("limited budgets, misalignment with cooking habits, low awareness") is a list of hypotheses, and the data that can distinguish them is in Tidewater's subscriber analytics and churn records, not in a PwC study.
L
Learn the drivers What drives that behavior today?
This is a journey-map project. Students map how a 30-year-old buys fish now (grocery counter, frozen aisle, meal kit, restaurant, not at all) and mark boosters and blockers at each step: price per meal against a meal kit, portion and freezer space in a one-bedroom apartment, cooking confidence with whole fillets, and whether the sustainability story is even heard. The subscriber data then quantifies what it can: do the few under-35 subscribers churn faster, buy smaller boxes, skip more months? Ten interviews with non-subscribers in the cohort, and five with young subscribers who stayed, will do more than any survey.
What they should discover is that the blockers are not equally weighted, and that the first answer (price) is usually the one young people give and not the one that predicts purchase. The designed discovery in a simulation would be that portion and cooking confidence dominate once price is controlled.
Coaching question: Of the six barriers on your map, which one have you measured, and which ones are you assuming?
V
Vet the shakiest claim What's the cheapest way to find out we're wrong?
Serious options: (a) a smaller, cheaper box for one- and two-person households; (b) the current box with a recipe-and-video layer that removes the cooking barrier; (c) a lower-commitment entry (one-off trial box, gift box) with the same product. For option (a): "We believe a smaller box at a lower price will get 25–35-year-olds to start a subscription. We'll know we're right when a trial offer converts at X% of the rate the standard box converts for older buyers." Must-be-trues: a smaller box is profitable per unit, fulfillment can pack it, and the segment actually responds to price rather than to confidence. Shakiest: the last one.
The cheapest test is a landing page with the smaller box at its price and a targeted ad spend of a few hundred dollars, with clicks and pre-orders counted. A second test, a recipe bundle on the existing box offered to the same audience, distinguishes price from confidence. Both fit in three weeks and neither requires sourcing a new product.
Coaching question: If your first experiment takes a month, it isn't an experiment. What can you learn by Friday?
E
Earn the yes What should Tidewater do Monday morning?
A conclusion-first recommendation: the one offer to launch, the evidence from the test, the per-box profitability, and a twelve-month subscriber and lifetime-value estimate with a range. The adoption argument is for operations and sourcing, who must support a new SKU, and for whoever owns marketing spend, who is being asked to reallocate toward a cohort with lower near-term value. The measurement plan closes the loop opened in O: trial-to-second-box rate tracked weekly.
A loop worth telling. The team launches the small-box landing page and conversion is poor. Interviews with the people who clicked but did not buy reveal that they wanted the box but did not know what to do with a whole coho fillet. The team returns to L, re-weights the cooking-confidence blocker, and re-runs the test with a recipe card and a two-minute video. That second result, not the first, is the evidence the recommendation rests on. Both belong in the project log.
As a GB210 simulation. Opening prompt from the co-owner: "young people aren't buying and we think it's price; tell us what to sell them." Client personas: the co-owner (holds the values constraint and the guess about price), an operations lead (holds subscriber analytics by age band and churn, shares them only when asked what data exists), and a lapsed 29-year-old trial subscriber who can describe what happened to her first box but cannot be shown a new one. Designed discovery: the under-35 subscribers who exist churn at the second box, and the lapsed subscriber never cooked the second fillet. Students should leave with a restated problem the owner agrees to, a target behavior with a leading indicator, and a written design for the landing-page test they would run in GB410.
Why aren't they buying · grab bag
Quietcue Health
Chosen because it is three projects in one scope, and because the illustrative team's narrowing move is instructive in both directions.
The setup. A discreet sensor that helps patients stay on their medications, billed to Medicare through about a few dozen pharmacy clients. Seniors resist enrollment ("defensive, defiance, and denial") and outreach depends on cold calls. The founder cites a very large market opportunity and asks for three things: how seniors make health purchase decisions and which channels work; which new markets (tribal communities, federally qualified health centers, caregiver networks, consumer) to enter and how to differentiate; and what else the sensor could be used for and who would pay. Deliverables span personas, outreach strategies, cost-benefit by market, and new-application feasibility.
S
Surface the real problem
The grab bag has one bottom line under it: enrolled, monitored patients per month, because each one bills. Baseline: a few dozen pharmacies, a conversion rate from cold call to enrollment the sponsor can produce, and a persistence rate after that. The S-work is negotiating the three asks down to the one that moves enrollments fastest, and writing down what was set aside and why. Growth in new markets, new applications, and better senior marketing are three different hypotheses about why enrollments are low; students should name which one they are testing.
Coaching question: Of your three asks, which one, if it worked, adds the most enrolled patients in the next year?
O
Observe the behavior
Two behaviors: a senior says yes to enrollment and keeps the sensor in use past ninety days; a pharmacist makes the introduction instead of leaving it to a cold call from a stranger. The second is probably the leading indicator for the first, and it is a behavior of a partner, not a customer. Students who write "increase senior adoption" have named a bottom line, not a behavior.
L
Learn the drivers
A funnel and a journey: pharmacist referral, outreach call, consent, install, persistence. The sponsor's adoption data should show where it leaks. Interviews with pharmacists and with a few enrolled and refused seniors (or their adult children) put the reasons on the map: who called, what was said, whether a trusted person vouched for it. The "defensive, defiance, denial" line in the scope is a symptom; the map should show what triggers it.
V
Vet the shakiest claim
Suppose the team chooses the market-expansion ask and concludes that tribal pharmacies are the best fit, on the grounds of chronic-disease prevalence, localized decision-making, federally protected reimbursement, and no change to the business model. As a claim: "We believe entering tribal pharmacies will grow enrolled patients faster than deepening existing pharmacies. We'll know we're right when a 90-day pilot in named pharmacies enrolls patients at a higher rate per pharmacy than the current base." Must-be-trues: the pharmacies sign; the seniors there resist less (or the pharmacist relationship overcomes it); billing works as described. The team's evidence covers the first and third with market statistics; the shakiest condition, that the senior adoption problem is smaller in the new channel, is asserted from "community trust," not tested. The cheapest test is the outreach pilot the team proposes, with enrollments counted rather than meetings.
Coaching question: You have shown the market is large and the billing works. What evidence do you have that seniors there say yes more often?
E
Earn the yes
A recommendation with the pilot, a quantified enrollment estimate with a range, a measurement plan (enrollments per pharmacy per month, 90-day persistence), and an explicit statement of what was descoped. Adoption burden is moderate and sits with the sponsor's outreach staff and partner pharmacists.
A loop worth telling. That team's executive summary is strong SOLVE practice: one problem statement, three options compared on named criteria, one recommendation, a 90-day proposal, an email template. Where SOLVE would have pushed: the behavior that made the original outreach hard was swapped for a channel that sounded easier, without evidence that the behavior changes. A single week of calls to two of the named pharmacies would have either confirmed the claim or sent the team back to L with better information.
As a GB210 simulation. Opening prompt: the grab bag itself, delivered by a founder who leads with the market size. Client personas: the founder, an outreach staffer who knows the call-to-enrollment rate and what seniors actually say, a partner pharmacist who explains why she does not make the introduction herself, and possibly an adult daughter of a patient who can describe what finally made her mother agree. Designed discovery: the funnel leaks at the first call, not at awareness, and the pharmacist's referral is the missing behavior. Students should leave with a descoping memo the founder has accepted, the funnel quantified, and a designed 90-day outreach pilot with enrollments, not meetings, as the endpoint.
The setup. About ten fast-casual locations across the state, a tenth opening soon. Leadership has discussed a commissary kitchen for years and never had time to evaluate it. The scope asks for a structured feasibility assessment: centralized versus hybrid, what to centralize, co-located versus separate, logistics, quality and risk, startup and operating cost, savings, break-even, sensitivity to store count, and implications for real estate. Data: P&Ls, product mix, recipe costing, equipment costs, rent, labor grids, bulk-pricing savings.
S
Surface the real problem
The scope is a concept looking for a verdict, and the S-work is to name what the verdict is for. Three candidate bottom lines hide in the background paragraph: cost per plate (food and labor), consistency across stores, and the ability to open stores faster and in smaller footprints. They are not the same project. A commissary that saves 2% of food cost but lets the next five stores open with half the kitchen is a real-estate strategy, not a cost program. Students should get the owner to rank them and to state the baseline per-store labor and food cost from the P&Ls.
Coaching question: If the commissary works, which line on the store P&L changes first, and by how much would it have to change to be worth the trouble?
O
Observe the behavior
Feasibility projects feel behavior-free, and they are not. Store kitchens must stop making the centralized items; store managers must accept product they did not make; prep staff hours must actually come out of the schedule rather than drifting into other tasks; a driver must deliver on a schedule that holds in a February. The leading indicator is prep hours per store actually removed. The lagging one is food and labor cost per plate.
L
Learn the drivers
An issue tree, built from recipe costing: for each prep process, labor minutes per batch, ingredient cost, waste, and how often it is made, across nine stores. That ranks what is worth centralizing (high labor, high volume, travels well) versus what must stay in store (finished to order, quality-sensitive). The savings side is labor consolidation plus bulk purchasing; the cost side is facility, equipment, transport, packaging, and the quality risk the owner cares about most. The predictable failure is modeling the whole menu; the coaching move is to pick the five processes that carry most of the prep labor.
Coaching question: Which five prep items account for most of the labor minutes across all nine stores?
V
Vet the shakiest claim
Options: (a) a separate central facility; (b) a hub store with extra capacity serving its neighbors; (c) no commissary, but a bulk-purchasing cooperative and standardized recipes. For option (b), usually the leading one at nine stores: savings survive transport cost, quality holds through a day of refrigerated travel, and managers accept it. Shakiest: quality-plus-acceptance, because it is the one the owner will not trade away.
Analytical evidence: the model with sensitivity on store count and transport cost, asked as "does the answer flip?" (at how many stores does a separate facility beat a hub store; at what diesel price does the hub stop paying). Behavioral evidence, cheap: make one high-volume prep item in one store and deliver it to three others for two weeks, track cost and ask managers and staff blind whether they can tell. Outcome evidence comes only after a decision, so the recommendation should build it in.
E
Earn the yes
A conditional recommendation: pursue the hub model when the store count crosses N or when the next lease is signed in a given region, with the sensitivity table showing why, a pilot design, and an adoption plan for store managers who will feel the change as a loss of control. Value stated as cost per plate and stores-openable-per-year, with ranges.
A loop worth telling. The two-week pilot shows the savings hold but one item arrives noticeably worse. Instead of killing the concept, the team returns to L and splits the process: centralize the labor-heavy step, finish the quality-sensitive step in store. The hybrid that results is the recommendation, and it only exists because the test failed in the cheap version first.
As a GB210 simulation. Opening prompt from the owner: "we've talked about a commissary for years; tell us if we should do it." Client personas: the owner (cares about quality above all, has not ranked cost against expansion), a finance lead who holds store P&Ls and recipe costing, and a store manager who does not want to become a reheating station. Designed discovery in the data: a handful of prep items carry most of the labor minutes, and transport cost, not facility cost, is the sensitive variable. Students should leave with the bottom line ranked by the owner, a sensitivity model built on the data, and a designed one-item, three-store pilot as the first step in their recommendation.
The setup. A member-owned, not-for-profit life insurer serving small towns, with an aspirational vision of closing the rural health and longevity gap. The COO asks for a non-insurance product roadmap: one to three concepts and a top recommendation, a five-year financial case, an execution plan, a launch marketing plan, and a change-management plan for employees and agents. Data: member research and an anonymized 5,000-member sample.
S
Surface the real problem
Two bottom lines compete in the scope, and the team has to make the COO choose the weighting: health and longevity impact on members (the vision) and economic workability (the constraint). "Economically workable" for a member-owned nonprofit might mean break-even, not profit; that single clarification changes which concepts survive. Baseline: members currently have insurance products plus a handful of benefits, including a wellness app that is underused.
Coaching question: If this product breaks even and measurably improves members' health, is that a success? If it profits and nobody's health changes?
O
Observe the behavior
The member behavior is sustained use of a health product (monthly active use, sessions completed), not purchase. The second behavior is the one students miss: a commission-paid agent brings the product up in a life-insurance conversation. A team on this path might note on its change-management slide that agents are not incentivized to sell the kit, and that a small share of agents produce most sales. That is an O-question insight sitting in an appendix.
L
Learn the drivers
A tree first: which part of the health gap (chronic disease management, provider access, isolation, physical activity) is largest for this member base, using the 5,000-member sample and public rural health data. Then a journey for the leading gap: what a 58-year-old member in a town of 3,000 does today about blood pressure or exercise, and what stops her. Three plausible concepts (traveling nurses, remote patient monitoring, a home fitness kit with an improved app) map to different branches; the tree is what should have ranked them, rather than a scoring matrix built before the drivers were known.
V
Vet the shakiest claim
For the recommended Home Kit plus app: "We believe a kit priced around $60 paired with senior-focused workout videos will get members exercising regularly. We'll know when pilot participants log 15-plus activities a month at a rate of X%." Must-be-trues: members buy at $60, they use it past the first month, agents sell it, the supplier price holds. Suppose the team assumes a 3% adoption rate to reach break-even (a few thousand kits a year against the member base) and a five-year profit in the low five figures on similar fixed cost. Shakiest: sustained use, then agent selling. The team proposes the right cheap test, an email opt-in to members 45–65 with free kits to volunteers and usage tracked, and stops at proposing it.
Coaching question: Your whole model rests on 3%. Where did that number come from, and what would you do differently if it were 1%?
E
Earn the yes
A recommendation that states value in both currencies: a health-behavior metric (members active per month) and the economics (break-even at N kits). The adoption argument has two audiences, agents and members, and a sweepstakes idea is a reasonable member-side mechanism. The measurement plan is the pilot itself, specified so the COO can run it in a quarter.
A loop worth telling. Suppose the email pilot returns strong sign-ups and weak month-two usage. The team returns to L and finds that the members who kept going were the ones in the sweepstakes or with a buddy. The product becomes a program (kit plus monthly challenge plus community), and the five-year model gets rebuilt around participation rather than kit sales. The financials get smaller and the mission case gets much stronger, which for this sponsor is the right direction.
As a GB210 simulation. Opening prompt from the COO: the vision statement and "bring us a product." Client personas: the COO (will not volunteer what "economically workable" means until asked), a product or actuarial lead who holds the member sample and the wellness app's usage numbers, a sales manager who can say what agents are paid on and what they will and will not bring up in a policy conversation, and a member in her late fifties who downloaded the app once and can say why she stopped. Designed discovery: a mission bottom line and a profit bottom line that point at different concepts, forcing the S-question, plus the agent-incentive problem hiding in the sales manager's answers. Students should leave with the COO's weighting of the two bottom lines, both behaviors named, and a designed member opt-in pilot.
Based on: recent GB410 project request forms (disguised), the SOLVE framework pages, and the instructor teaching note. Team paths are composites.