Project description

Due on Canvas by 11:59 pm on Friday, October 23. Each team then gives a three-minute talk about it in session B of meeting 10, on Wednesday, October 28.

This is the first written piece of your team’s proposal. It replaces the one-page pitch announced in meetings 1 and 3.

The assignment in brief

What	A description of your project, up to three pages, and a three-minute talk about it
Who	One submission per team
Weight	5 % of the course grade. Every member receives the same grade
Due	On Canvas by 11:59 pm on Friday, October 23. Canvas asks for your NetID
Length	Up to three pages for the six sections. The references and the AI disclosure go on a separate page that does not count
File	One PDF, named `Team-A_project-description.pdf` with your team’s letter
Talk	Three minutes in session B of meeting 10, on October 28, followed by one question from the room
Examples	Two worked examples by hypothetical teams. Solar farms and home values asks a question a town would act on (PDF). Boycotts and store visits at Target asks one a company would pay for (PDF)

What the document must contain

The first line of page 1 gives your team letter and a working title. The second line lists every member’s full name. A name on that line means the member has read the document and agrees with it.

Then use the six headings below, numbered and worded as they are here, in this order. Each one gives a rough length and the session-B sheet to start from. If your team kept no copy of a sheet it handed in, ask me for it.

1. Research question

About five to eight lines. Start from your meeting 3 question sheet.

State the question in one sentence ending with a question mark. It is the first sentence of the section.
Name its shape from meeting 3: evaluation, preference and pricing, prediction and risk, diagnosis, feasibility, measurement, or valuation.
Name who would act on the answer or pay for it, by role. “The pricing manager of a regional grocery chain” is specific enough. “Policymakers” is too vague.
State the answer you expect, with its direction and a rough size, and the result that would show you were wrong.

2. Why it matters

About half a page. Start from the line Who acts on it, or pays for it and from filter 4 on your meeting 3 question sheet.

Name the decision that would change with the answer, who makes it, and what they would do if the answer came out each way.
Summarize your two or three anchor papers, the published studies closest to yours, in one or two sentences each. Say what your project adds to them.
Cite each study by author and year, and list it in the references.

3. Approach

About half a page, plus your causal graph if you include one. Start from your meeting 2 graph and from the lines you wrote at meeting 8 on whether an experiment, an instrument or a discontinuity is open to you. A photograph of your meeting 2 graph is fine if its labels are legible.

Every project states these three things.

Name the unit of observation, such as a store-week or a household.
Name the outcome, the variable that measures it, and its source.
Describe the sample: which units, where, which years, and roughly how many. Say whether it covers the people or places that the decision in section 2 concerns.

If your question asks what a change caused, state these four as well. A preference and pricing question belongs here, because the treatment is the price or attribute you vary in your task.

Name the treatment, the change whose effect you want to estimate.
Describe the comparison group or period. Its outcomes stand in for the counterfactual, what the treated units would have done without the treatment.
Name the research design, such as an experiment, instrumental variables, regression discontinuity, difference-in-differences, or panel data with fixed effects.
State the identifying assumption in one sentence. It is the assumption that makes your comparison a valid stand-in for the counterfactual.

If your question asks anything else, state this instead. That covers a prediction, a measurement, a diagnosis, a feasibility study and a valuation.

Say what the answer has to get right to be useful, and how you will check it against something outside your own data. A prediction is checked by its error on data you held out. A measurement is checked against a count from another source. A valuation names its comparables and what makes them comparable. A diagnosis names the benchmark it measures the gap against. A feasibility study names the original data your team will analyze itself.

4. Data and access

About three quarters of a page. Start from your meeting 3 resource map. I read this section first, because most capstones that fail had a good question whose data never arrived.

Give one row per dataset, in a table with these six columns.

Dataset	Holder	Unit and period	Variables you need	Access time	Status
Example: Weekly Patterns Plus	Dewey, through Cornell’s subscription	Store-week, 2019–2024	Weekly visits, store category, location	Days	File opened Oct 12

Access time is your dataset’s row on the access ladder: now, days, weeks, months, or longer.
Status is one of four entries: file opened, access requested, or holder contacted, each with its date, or not started.
If you collect the data yourself, give the survey or experiment its own row. The holder is your team. The unit is the participant, and the period is when you will field it. The access time is weeks. The status is your stage, with its date: protocol submitted, instrument drafted, or not started.

Below the table, add three things.

Give the date by which you expect to hold every dataset.
Give your plan B, the question your project becomes if the data has not arrived by that date. If you already hold every dataset, give the question your project becomes if the data turns out not to contain what you need.
Say whether the project needs an IRB application or a data-use agreement, or that you have asked the IRB office to decide, with the date you asked. Use your meeting 4 IRB and DUA checklist. If the project needs an approval, give the date the slowest one has to start.

5. Anticipated challenges

About a third of a page. Start from the one-page threat your team wrote at meeting 7.

Name two or three challenges. For each one, say what you will do if it is not resolved, and by what date. One of them must be the main threat to your estimate.

For a causal question, name the source of bias in meeting 7’s terms, such as selection into treatment or omitted variables. Say the direction in which it pushes the estimate, and whether you can measure the variable that would close it.
For any other question, name the main way the answer could be wrong, and how you would find out.

The other challenges can be the row of your resource map most likely to stop the project, or an approval you are still waiting for.

6. Team plan

About half a page, in two parts. Start from your meeting 2 ground rules.

Roles. Give one row per member, in a table with these five columns. Each Delivers cell names a specific output. Each By cell gives a date before December 11, or a weekday for a task that repeats every week. Name a coordinator, who runs the weekly check-in and uploads the team’s work. Name a backup for every main task. For example:

Member	Role	Delivers	By	Backup
Farah	Coordinator	Runs the check-in. Uploads each deliverable	Every Wednesday	Gabe
Gabe	Data	Dewey account approved. First extract opened and described	Oct 30	Hana
Hana	Design	Causal graph and identifying assumption, written for the outline	Nov 4	Idris
Idris	Literature	Two more studies opened. What is already known, written for the outline	Nov 4	Kofi
Kofi	Writing	Outline merged and edited into one voice	Nov 10	Farah

Missed work. Write the steps your team follows when a member does not deliver. A miss is a task not done by its date without warning. A member who warns the team before the date has not missed it, and the team sets a new date at the next check-in. Give at least three numbered steps, ending with the team bringing the problem to me. You may use these three as written:

First miss. It is raised at the next check-in and written down, with a new date.
Second miss. The coordinator talks to the member that week. The task goes to the backup or is split, and the check-in records it.
No change, or a week with no reply. The team brings it to me, with the check-ins as the record.

References and AI disclosure

These two items go on a separate page after the six sections. They do not count toward the three pages, and nothing else goes on that page.

References. List every work you cite, in any consistent citation style, each with a link or DOI. Your datasets belong in the section 4 table, so list one here only if you cite it in the text. Cite only works you have opened. A citation to a work that does not exist is treated as fabrication under the AI policy.

AI disclosure. Write the three lines from meeting 5: the tools you used, what you used them for, and what you checked. If you used none, write “Tools. None.” For example:

Tools. Claude (claude.ai).
Used for. Finding candidate anchor papers. Tightening the wording of section 2.
Checked. We opened every reference and wrote each summary ourselves.

Format and submission

Use letter paper, one-inch margins, and 11-point type or larger. Single spacing is fine. Tables and figure labels may use 9-point type.
Fit sections 1 to 6, including tables and figures, within three pages. I do not read anything past page 3 except the references and the AI disclosure. Do not add appendices.
Name the file Team-A_project-description.pdf, with your team’s letter.
One member uploads it to the Canvas group assignment for the team. I do not accept submissions by email.
You may resubmit before the deadline. I grade the last submission.
The syllabus late policy applies. A late submission loses 10 % up to 8 hours late and 50 % from 8 to 72 hours. After 72 hours it is not accepted.

The talk at meeting 10

Every team presents in session B of meeting 10, on Wednesday, October
1. The order is drawn at random, so be ready from the start of session B.
You have three minutes, timed, and I stop you at three. One question from the room follows.
Cover the question, who acts on the answer, your approach, and your data with its status. For a causal question, the approach includes the design and its identifying assumption.
Your team decides who speaks.
Slides are optional, two at most. Attach them to the same Canvas submission as the description, as a second file named Team-A_slides.pdf. If you resubmit, attach both files again. I show the slides from the room computer.

How it is graded

The document and the talk are scored together out of 100. That score is worth 5 % of the course grade, and every member of the team receives it.

Part	Points
1. Research question	15
2. Why it matters	10
3. Approach	15
4. Data and access	20
5. Anticipated challenges	10
6. Team plan	10
The talk	20
Total	100

Each part is scored on the syllabus scale. Excellent earns full points. Adequate earns three quarters and weak earns half. A part that is missing, or so thin it shows no thought, earns zero. The talk is scored on whether it covers the four items within three minutes, and on how the team answers the question.

Questions students ask

Can we change our question after this? Yes. Record the change and the reason in your team’s decision log.

We do not have the data yet. Is that a problem? No. Show what you have checked and requested, with dates, and give your plan B. A dataset you have named but not checked scores weak.

Our question does not ask what a change caused. Do we need a research design? No. Complete the last part of section 3 instead.

We plan to use difference-in-differences, which meeting 9 covers two days before the deadline. How much do we write? Name the design, describe the comparison in plain words, and state the identifying assumption in one sentence, as the solar example does: without the treatment, the two groups would have followed the same trend. Remix chapters 8 and 9 cover it, and you refine it in the outline.

Can the team lower a member’s grade for not contributing? No. Every member receives the same grade on team work. Follow your missed-work steps, which end with the team bringing the problem to me.

Why do we have to write roles and a missed-work procedure? Writing them down early is standard practice, and there is evidence for it. Mathieu and Rapp followed 32 teams of MBA students through a business strategy simulation. The teams that sustained the best performance had written both a good team charter and a good plan for the work. Oakley and colleagues give a template for both.