The research question

What useful thing can these skills produce for someone else? The answer does not have to be a trade. Building reproducible experiments develops skills in Python, data quality, evaluation, automation, documentation, and technical communication. Those skills can support educational content, consulting projects, and software tools.

This final edition creates a bounded offer and checks its arithmetic. It does not promise customers, revenue, or investment returns. The exercise is about selling a defined deliverable that you can demonstrate, not selling confidence in a market prediction.

Time: approximately 40 minutes. You need: your earlier lab artifacts, a text editor, and Python for an optional arithmetic exercise. No customer outreach or publishing happens automatically.

Step 1 — Pick one problem and one buyer

Choose a problem that you can verify through conversation. For example: a training team wants a beginner workshop on reproducible Python experiments; a researcher needs a CSV validation tool and readable error report; or a technical publisher needs help turning a working script into a complete tutorial.

Avoid an offer as broad as “I will build an AI system that makes money.” A specific offer is easier to evaluate: I will deliver a two-hour workshop with a synthetic dataset, three runnable labs, an instructor guide, and a troubleshooting sheet.

Write down who receives the deliverable, what they can do afterward, and what is explicitly outside scope. Brokerage execution, investment recommendations, and guaranteed trading performance are outside the offer in this series.

Step 2 — Build a small proof package

Use the artifacts you already made: a data contract, a cost-aware ledger, a frozen evaluation, an agent validator, a job-state demonstration, and a reproducible report. Select two that directly address the chosen buyer's problem. Keep the synthetic-data labels and limitations intact.

Package them with a README that includes the supported Python version, exact run command, expected output, and troubleshooting steps. Add a short screen recording only if it helps explain a workflow; the runnable files remain the evidence.

For consulting, define acceptance criteria before starting: the tool rejects duplicate identifiers, produces an intelligible error report, and passes the agreed fixture checks. For a workshop, define learner outcomes and the required starting knowledge. For a paid tutorial, say exactly what the reader receives.

Step 3 — Check the economics without pretending it is a forecast

The following optional lab uses invented planning amounts. It does not calculate a market price, a recommended rate, taxes, payment fees, or a promise of sales. Save it as lab_08.py.

from decimal import Decimal, ROUND_CEILING

D = Decimal
# Invented planning assumptions, not a forecast or recommended selling price.
preparation_hours = D("6")
internal_hour_value = D("25")
fixed_tools_cost = D("20")
example_price_per_seat = D("40")
variable_cost_per_seat = D("5")
fixed_planning_cost = preparation_hours * internal_hour_value + fixed_tools_cost
contribution_per_seat = example_price_per_seat - variable_cost_per_seat
assert contribution_per_seat > 0
break_even = (fixed_planning_cost / contribution_per_seat).to_integral_value(
    rounding=ROUND_CEILING
)
print(f"fixed_planning_cost={fixed_planning_cost:.2f}")
print(f"contribution_per_seat={contribution_per_seat:.2f}")
print(f"planning_break_even_seats={break_even}")
for seats in (0, 3, 5, 8):
    contribution = contribution_per_seat * seats - fixed_planning_cost
    print(f"seats={seats} planning_surplus={contribution:.2f}")
print("scenario_only excludes_taxes_fees_marketing_and_refunds")

Expected output:

fixed_planning_cost=170.00
contribution_per_seat=35.00
planning_break_even_seats=5
seats=0 planning_surplus=-170.00
seats=3 planning_surplus=-65.00
seats=5 planning_surplus=5.00
seats=8 planning_surplus=110.00
scenario_only excludes_taxes_fees_marketing_and_refunds

Five seats cover only the specified planning costs. If you add taxes, payment processing, marketing, support, revisions, or refunds, the result changes. If nobody buys, preparation still consumed time. Treat the calculation as a question generator: which costs or constraints did you forget?

Step 4 — Write a small offer page

Use this structure in a draft: the customer's problem; the deliverable; the starting requirements; what is included; what is excluded; delivery timing; acceptance criteria; price and commercial terms you have actually chosen; and a sample artifact.

Use Publish Haven's free article tools to explain your process or reader memberships for an ongoing educational publication, if that fits your offer. Do not describe a one-off workshop seat as a platform feature that does not exist. If an offer needs separate booking, contracting, or fulfillment tools, state that clearly and choose them deliberately.

A free edition can teach one complete idea. A paid educational series should offer a clear additional benefit, such as deeper walkthroughs, exercises, or organized reference material. Keep essential limitations visible to both free and paid readers.

Step 5 — Validate demand before expanding

Ask prospective users what they currently do, where they get stuck, and what a successful deliverable would change. Collect permission before adding anyone to a mailing list. Do not fabricate testimonials or present a teaching fixture as a customer result.

Start with a bounded pilot if there is real interest. Track the work required to deliver it, what the customer actually used, and what needed revision. Those observations are evidence for improving the service; likes on a promotional post are not the same thing as paying demand.

Troubleshooting

If the offer needs a paragraph of vague adjectives to sound useful, narrow the deliverable. If a customer expects a profitable trading system, clarify the educational or engineering scope before accepting the work. If the arithmetic looks attractive only when your own time is valued at zero, make that assumption visible. Do not hide unpaid preparation behind a revenue total.

Evidence and limits

A reproducible portfolio demonstrates some technical and communication capability. It does not guarantee competence for every production system or guarantee employment, sales, or income. Scope, support obligations, data permissions, commercial terms, and any applicable legal requirements depend on the actual engagement.

Completion check: You have one specific offer, two reproducible sample artifacts, acceptance criteria, a cost scenario, and a list of assumptions still requiring validation.

Your next experiment

Return to the journal's original question: what counts as learning? Publish a new experiment that changes one declared assumption, keeps a comparison, and reports what happened honestly. Or turn one lab into a lesson and observe where real learners get stuck. Both paths build on evidence rather than a promise.

Return to the beginning of the research series.

Primary reference

  • Python Decimal documentation documents decimal arithmetic and explicit rounding. The offer framework and planning scenario are original educational examples, not observed business results or pricing guidance.