# Project 2. Unclaimed Property at Silicone Motors

**Part 3 project. 100 points. Due Sunday, November 1, 11:59 PM MT.**

Submit through the Project 2 quiz on Canvas. It has three upload questions, one each for your slide deck (PDF), your executive one-pager (PDF), and your data support file, plus a few closing questions. Late work follows the syllabus: 10 percent of the points per day, capped at 50 percent.

**This scenario is fictional.** Silicone Motors, its records and its law firm are made up, and the legal rules in this project are invented for it. Escheatment is a real process that every company holding customer balances has to follow, but please do not cite this project or its made-up legal guidance at work.

## TLDR

**The two questions.**

1. What do we owe, to whom, and by when?
2. What should we do about it before the clock runs out?

**What you turn in.** A slide deck (PDF), an executive one-pager (PDF) and a data support file, uploaded through the Project 2 quiz on Canvas. Due Sunday, November 1, 11:59 PM MT.

**What you need.** Each resource is labeled, here and in 3.3.

- **Required:** this guide, the [counsel memo](https://seanmccaman.com/acctg6155/2026-fall/project-2/memo.html) (Schedule A is an image on pages 3 and 4), the data (3.2), and the [data dictionary](https://seanmccaman.com/acctg6155/2026-fall/project-2/data_dictionary.html).
- **Optional help:** the starter notebook, the [data explorer](https://seanmccaman.com/acctg6155/2026-fall/project-2/interactive/data_explorer.html), and the [example one-pager](https://seanmccaman.com/acctg6155/2026-fall/project-2/onepager_example.html).
- **Background:** the [escheatment model](https://seanmccaman.com/acctg6155/2026-fall/project-2/interactive/escheatment_model.html). Use it only if escheatment is new to you.

The documents and the data also come as one zip from the Project 2 page.

**The steps, in order.** Read Schedule A off the scan and build the rule table. Rebuild the balances and tie them to the GL. Build the owner table. Apply the rules and build the letter list. Write the recommendation. Build the deck and the one-pager. Section 4.1 has the detail.

**Using AI.** I encourage you to use it, especially to read Schedule A off the scan and to write the SQL and the pandas. Check its work against the memo and the image, not against the AI. Whenever you use it, make it explain what it did. If I ask you about your work and you cannot explain it to me, you don't own it. Section 4.5 has a prompt for this.

**Grading.** Section 5 has the rubric. Before you submit, you can have an AI review your draft against the rubric, as the end of section 5 explains.

## 1. Scenario: you are the Controller's analyst

Silicone Motors builds electric vehicles and runs a charging network. Drivers load money into a wallet and spend it at the company's chargers. The wallet can't be cashed out, only spent. People who want a car put down a refundable deposit to hold their place in line. Both are customers' money sitting on the company's balance sheet. When an owner goes quiet long enough, state unclaimed property law says the company has to report the balance and turn it over to the state of the owner's last known address. That process is called escheatment. Two of the states where the company has drivers exempt the wallet, so there the company keeps what nobody spends.

It is July 10, 2026. The company's outside counsel has sent the Controller a memo summarizing the rules for the nine states where the company has drivers. Counsel needs this year's due diligence letter list by August 15, so the letters can go out before the mailing window closes on September 2. The full schedule of what the company owes is due to counsel by October 25, and the report goes to the states on November 1, which, not by coincidence, is your project due date. August 15, September 2 and October 25 are dates in the story, and you turn in the letter list with everything else on November 1.

The Controller has two questions, and they are the whole project:

1. **What do we owe, to whom, and by when.** Which balances are dormant as of June 30, which state each one reports to, how much goes out the door on November 1, and which owners get a letter before the money goes.
2. **What we should do about it before the clock runs out.** An owner who acts before the dormancy anniversary keeps their money, because their own activity resets the clock. When does outreach still work, and what is at stake in dollars?

We will use Silicone Motors' fake little database: the wallet ledger, charging sessions, logins, reservations, the notices the company sent, the letters it mailed in prior cycles, and the general ledger control balances. The rules are in the counsel memo. Part of the memo is an image instead of text, which makes getting the rules out a bit of work. Schedule A, the dormancy periods by state, is a scanned image on pages 3 and 4 of the memo, and a clean copy of the same two pages is linked under it. It lists all 50 states by full name, so extracting it and mapping each name to the state codes in the data is part of the work.

### 1.1 Learning objectives

By the end of this project you will have:

- Rebuilt a subledger from transactions, tied it to the general ledger, and explained the difference instead of forcing it.
- Turned the counsel memo into a rule table you can compute with, and checked it against the memo rather than against the AI that helped you.
- Read Schedule A off a scanned image into a table, and checked it against the image row by row.
- Applied the rules to every owner, where the hard part is usually deciding what counts rather than the arithmetic.
- Written a recommendation to a specific decision maker, with a deadline and a dollar figure.

Where the skills come from: the subledger rebuild and the joins are Week 3, and the pandas is Week 7. Reading tables out of images is the Week 8 topic, and Schedule A is where this project uses it; you can start on it before then with any AI that reads images. Turning the memo into a rule table takes careful reading more than a new technique.

### 1.2 Deliverables

| Deliverable | What it is | Weight |
|---|---|--:|
| **Slide deck** (PDF) | The analysis, for the Controller. About 8 to 10 slides, plus an appendix. See 4.2 | 70 |
| **Executive one-pager** (PDF) | The same analysis compressed to what a busy reader needs. See 4.3 | 20 |
| **Data support file** (ipynb, xlsx, or another format that shows your work) | This can be any format you worked the data in, as long as it shows how you did it. Every number in 3.4, with the work that produced it. See 3.5 | 10 |

The data support file is graded on completeness: every required number present, in the requested units, with the method visible. Whether the numbers are right is graded in the deck, under Findings & Evidence, and I check them against the unrounded values in your data support file. The deck and one-pager are graded on judgment and communication, and that is where the points are.

You submit through a Canvas quiz with three upload questions: the deck, named `lastname_p2_deck.pdf`, the one-pager, named `lastname_p2_onepager.pdf`, and the data support file. The data support file can be xlsx, ipynb, sql, csv, zip, or pdf. If your work is in more than one file, such as a notebook and a .sql file, zip them into one. The quiz ends with a few closing questions, which ask which AI tools you used, how you used them, how long the project took and where the time went, plus a box for feedback. They are not graded. If you submit more than once, upload all three files every time, because each new attempt starts empty. I grade your latest attempt, so a resubmission after the deadline replaces your on-time attempt and is marked late.

## 2. Grading

A strong deck answers both of the Controller's questions with correct numbers for items 1 to 7, a clear narrative, and a recommendation. Excellent ratings come from the judgment calls: getting Schedule A out of the image and checking it, handling the traps in the data, including some this guide does not point out, explaining the reconciling differences, and addressing the outreach process. A deck that reports numbers without saying what they mean, or that gets the population wrong because it trusted a column name instead of the memo, may not get full credit.

Every deduction traces to something in this guide or the memo. If a number is wrong, the feedback names which one. If something on the Professionalism & Formatting line is missing, it names the item. Section 5 has the rubric.

## 3. Data

### 3.1 The database

One SQLite file, `p2_wallet.sqlite`, about 10 MB, plus the same tables as CSVs for anyone working in pandas or Excel. Open it in the course SQL explorer (linked from the Project 2 page), in DB Browser for SQLite, in Python with `sqlite3` or `pandas.read_sql`, or load the CSVs. The AI tutor on the explorer page can see the schema, your query, and the column names and row count of your result, but never the values in it.

| Table | Grain | What it holds |
|---|---|---|
| `drivers` | one row per wallet holder | opened date, the address as captured at signup, the most recent date postal mail came back |
| `wallet_ledger` | one row per wallet movement | the date it happened, the date it posted to the GL, the type, the signed amount in cents, the channel it came through, and a link to the row it reverses (on a reversal) or the charge it corrects (on a refund) |
| `charging_sessions` | one row per charging session | station, energy, and the ledger row that billed it |
| `app_login_days` | one row per driver per day with a login | day grain, no time of day |
| `reservations` and `reservation_events` | one row per reservation, one per event on it | deposits placed, applied to a vehicle order, and refunded, each with the channel it came through |
| `company_notices` | one row per message the company sent | balance statements, promotional email, and service alerts, with a bounced flag |
| `due_diligence_letters` | one row per letter mailed in a prior cycle | 2023 to 2025, with return and response dates |
| `gl_wallet_control` | one row per month-end per GL account | the general ledger balances you tie to: 2110 for cash and 2115 for promotional credit (2120 holds the deposits, and the project does not ask you to tie it) |
| `run_params` | one row | the cutoff and reporting dates for this cycle |

The full column list with descriptions is in the data dictionary, on the Project 2 page and in the CSV download, and the same text is in the `_data_dictionary` table so the tutor can read it. The dictionary describes fields. It does not tell you which fields count for what. The memo does that.

Two things to know about the data before you start. The `state_raw` column is what was typed at signup, and it was typed by people, so the same state appears several ways, some of them differing only by case or a stray space. You can list every spelling, and cleaning them is your job, because a join that silently drops the rows it cannot match is one of the most common ways this analysis goes wrong in practice. And every amount is in integer cents, which is deliberate: Excel, pandas, and SQLite round differently, and cents keep your numbers exact until you choose to divide.

### 3.2 Getting and loading the data

Everything is on the Project 2 page and in the `2026-fall/project-2/` folder of the course GitHub repository. Load the CSVs from their raw URLs, or download the SQLite file. The course web host blocks programmatic fetches, so a notebook should always pull from `raw.githubusercontent.com`, never from the course site. The repository is named `acctg5150-090` from an earlier term, and it is the right one for this course.

To download every table at once as CSVs, use [p2_csv.zip](https://seanmccaman.com/acctg6155/2026-fall/project-2/data/p2_csv.zip). Download everything on the Project 2 page has them too. The same CSVs, one file per table, are in the [GitHub data folder](https://github.com/sean-mccaman/acctg5150-090/tree/main/2026-fall/project-2/data/csv), and a notebook loads them from the raw URLs like this:

```python
import pandas as pd
base = "https://raw.githubusercontent.com/sean-mccaman/acctg5150-090/main/2026-fall/project-2/data/csv/"
drivers = pd.read_csv(base + "drivers.csv", dtype={"postal_raw": str})
ledger = pd.read_csv(base + "wallet_ledger.csv")
```

The starter notebook, `project2_starter.ipynb`, has a load cell for every table, a stub for each step in 4.1, and a numbers section for every 3.4 item. Open it in Colab from the Project 2 page. It is optional.

### 3.3 The documents you were given

Each resource has a label. **Required** means the graded work depends on it. **Optional help** makes the work easier, and you can skip it. **Background** explains the setting for anyone new to escheatment.

| Resource | Label | Use it for |
|---|---|---|
| This guide | Required | The assignment, the definition of every number, and the rubric |
| [The counsel memo](https://seanmccaman.com/acctg6155/2026-fall/project-2/memo.html) | Required | The rules. Schedule A is an image on pages 3 and 4 |
| The data, as SQLite or CSV (3.1 and 3.2) | Required | The records you analyze |
| [The data dictionary](https://seanmccaman.com/acctg6155/2026-fall/project-2/data_dictionary.html) | Required | What every table and column holds |
| The starter notebook (3.2) | Optional help | A load cell for every table and a stub for each step |
| [The data explorer](https://seanmccaman.com/acctg6155/2026-fall/project-2/interactive/data_explorer.html) | Optional help | The database in a SQL editor that runs in your browser |
| [The example one-pager](https://seanmccaman.com/acctg6155/2026-fall/project-2/onepager_example.html) | Optional help | The format of a one-pager, on a different topic |
| [How an Unclaimed Property Balance Is Built](https://seanmccaman.com/acctg6155/2026-fall/project-2/interactive/escheatment_model.html) | Background | The escheatment model on practice data, if escheatment is new to you |

**The counsel memo**, on the Project 2 page, is the authority for this project. It gives the dormancy periods for each state, which activities count as owner contact, which state a balance reports to, and the due diligence rule. Read all of it, including the footnotes. Then turn it into a rule table, one row per state, with columns for each rule the memo sets, and check every cell against the memo. If you are using the web SQL editor, a CTE works fine for the rule table (or even CASE statements).

Schedule A is an image on pages 3 and 4 of the memo, and getting it into a table is part of the work. Any AI that reads images can do the first pass, and so can an OCR tool. A few things make that pass better: give it one page at a time, ask for CSV with the column headers exactly as printed, and ask it to list every cell it was unsure of. Run it on the scan and on the clean copy linked under it, and compare the two tables; where they disagree, the image decides. Then check every row against the image yourself, all 50 states, including the footnote marks, because a footnote can change a number. Last, map each state's full name to the code the data uses, so your table can join to the owners. If you need an accessible copy of the schedule, ask me by Canvas message.

The memo says this plainly and so will I: its rules are a summary of the law as it applies to this company after elections its lawyers made, and they will not match what a general AI or a web search says about any real state. **The memo is the authority for this project** (again, please do not apply this logic to a real business). A student who asks an AI what California's dormancy period is, and uses that, will get the wrong answer, and the feedback will say so. This is not a trick. Real unclaimed property law often has no clear answer for a case like this, so a Google search can send you on a wild goose chase.

### 3.4 The numbers you report

These are the numbers the Controller needs. Put every one of them in your data support file, dollars to the cent unless told otherwise. Keep full precision while you work, including after a reportable percentage and in the $50 test, and round only the totals you report. The deck and the one-pager can round for the reader.

Definitions, in the words the grading uses:

- **Owner contact, wallet.** A qualifying wallet transaction as the memo defines it, an authenticated login in a state whose schedule says logins count, or a documented response to a prior due diligence letter. An owner has one wallet, and it runs its own clock, separate from any deposit the same owner holds.
- **Owner contact, deposit.** An owner-initiated event on that reservation. Each deposit runs its own clock, and activity on the wallet is not contact for a deposit or the reverse. A login is contact for the wallet only, never for a deposit. In the states that exempt stored value, a wallet is never reportable and never enters the dormant count; its deposits still follow the deposit period.
- **Dormant as of June 30, 2026.** The last owner contact date plus the dormancy period falls on or before June 30, 2026. An anniversary on June 30 itself is dormant. Schedule A gives periods in years, so add calendar years rather than a count of days: in Excel `EDATE(date, 12 * years)`, which counts in months, and in pandas `+ pd.DateOffset(years=n)`. Counting 365 days a year drifts around leap years, and one day can move an owner across the cutoff.

The numbers:

1. **Wallet cash liability** at June 30, 2026, on the date each transaction happened. Promotional credit is not part of it; it is item 3.
2. **The GL difference.** Account 2110's balance at June 30 minus your rebuilt cash liability (keep the sign; it can be negative), the number of transactions that explain it, and the cause in one sentence.
3. **Promotional credit** on the books at June 30, and confirmation that it ties to account 2115.
4. **Dormant wallets**: the count and the reportable amount, each by state and in total. Count only wallets that still hold cash at June 30. Wallets in the states that exempt stored value are not in this count.
5. **Dormant deposits**: the count and the reportable amount, each by state and in total. Count only deposits with a balance still held at June 30.
6. **Reportable to Delaware**, both property types together, and why each group of those owners reports to Delaware.
7. **Due diligence letters:** the list of owners (driver_id) who get a letter this cycle, and the count. The $50 test uses the reportable amount, after any reportable percentage, with the owner's wallet and deposits added together. The letter is the company's last required attempt to reach an owner before the money goes to the state. It sets the mailing cost. A reply to a letter counts as owner contact for the wallet only; a deposit needs an owner-initiated event on that reservation. The list has no bearing on your other numbers: compute everything as of June 30, before any of this year's letters go out. Letters from earlier cycles are different. They are already in `due_diligence_letters`, and an owner's reply to one counts as owner contact for the wallet.
### 3.5 The data support file

One file, or one zip, with every number in 3.4, labeled with its 3.4 item number, and the work that produced it visible. A number whose method you cannot show here loses credit. However you worked:

- **Colab or Jupyter.** Run every cell before you download the notebook, so the outputs show. Keep long outputs short, `.head()` rather than a whole table. The two exceptions are the item 7 letter list and your Schedule A table, which have to be there in full: print them, or save them as CSVs and zip them with the notebook. The starter notebook ends with a numbers section that lays out every 3.4 item for you to fill in.
- **Excel.** Organize your workbook so I can easily find your answers and the supporting figures, and can see the formulas you used to connect your data. Put the item 7 letter list and your Schedule A table on sheets of their own.
- **SQL.** If you used SQL alongside Excel or a notebook, put the queries in that file, or in one .sql file zipped with it. If you worked purely in Excel, this is not needed. If you worked purely in SQL, put every query in a single .sql file, each labeled with its 3.4 item number and with its result written as a comment under it.

Whatever format you use, include the table you extracted from Schedule A, all 50 rows, with the state code you mapped each one to. A sheet, a CSV in your zip, or a table in your notebook is fine.

## 4. Suggested workflow

### 4.1 Step by step

1. **Extract the memo inputs.** Build your rule table first, before any code, and check it cell by cell against the memo. Schedule A is an image, so extract it (3.3 has a few tricks for that), map each state's full name to how that state appears in the data, and decide which of the 50 rows you need. The memo has a footnote that changes a number. Find it.
2. **Rebuild the balances.** Sum the ledger per driver as of June 30 on the date the transaction happened, cash types and promotional types separately. Tie each to its GL account. One of them will not tie. Find the transactions that explain the difference, and you will find the GL posts on a different date than the one you summed on. The difference is a reconciling item with a cause, not an error to plug.
3. **Build the owner summary table.** Start from `drivers`, because it holds the whole population, and bring everything else in with a left join, so no owner drops out. Summarize each table to one row per owner before you join it, so the join cannot repeat an owner. Aim for one row per owner with these columns: `driver_id`, `clean_state`, jurisdiction, cash balance, promo balance, last owner wallet transaction, last login, last letter response, last owner contact, and mailable. Some columns fill in now and the rest in step 4. Deposits run their own clocks, so build a second table the same way, one row per reservation.

   Make sure you don't read a company (internal) event as a customer event. Company-generated entries look like activity and are not, and the channel column tells you which is which if you read the memo rather than guessing from the names. And a transaction that was reversed never happened, so it is not contact either. Here is why. A card load is credited to the wallet when the card is authorized. If the payment then fails to settle, the system reverses the load, usually within a few days, and the reversal row points back at the load in `reverses_txn_id`. A load that never settled is not contact. A charge that was later refunded in part did happen; the refund is a correction, not a reversal, and both the charge and the refund are the owner's doing. A refund also fills in `reverses_txn_id`, pointing at the charge it corrects, so read the type: only a reversal cancels the row it points to. Judge contact as of June 30 using the records you have. The table below shows the difference.

A reversal next to a partial refund, in four made-up rows (not from the data):

| txn_id | txn_type | amount_cents | reverses_txn_id | Counts as owner contact? |
|---|---|--:|---|---|
| 90001 | load | 4000 | | No. It was reversed, so it never settled |
| 90002 | reversal | -4000 | 90001 | No. A reversal is a company entry |
| 90003 | charge | -1800 | | Yes. The charge happened |
| 90004 | refund | 900 | 90003 | Yes. The owner disputed part of the charge |

4. **Apply the rules.** Normalize the state, resolve the jurisdiction using the memo's priority rule, apply the period, apply any reportable percentage, and build the letter list. A reportable percentage works as a multiplier: the reported amount is the dormant balance times that percentage. Keep the fraction of a cent; round only the totals. Owners with no usable address are not dropped; the memo says where they go.
   - **Hint.** The web SQL editor does not keep your changes. A table you create or update there is gone when you reload the page. To clean the states, use a CTE or CASE in your query, or work in your own copy of the .sqlite file.
   - **Tip.** Add a `clean_state` column next to `state_raw`. Never overwrite the raw field, so you can sample rows and check your mapping.
5. **Recommend.** Use the anniversaries to show how much money reaches dormancy in a window you choose, and say why you chose it. If no window stands out, use the twelve months after the cutoff, through June 30, 2027, the next cycle's cutoff. Then say what the Controller should do about that money before it goes, who does it, by when, and why. Show the window calculation in your data support file.
6. **Build the deck and the one-pager.**

### 4.2 Building the deck

About 8 to 10 slides for the Controller, who has ten minutes and will forward it to counsel. It is turned in, not presented, so it has to stand on its own. Explain terms such as dormancy and owner contact in plain words. Lead with what you found, then show how you got there. This is a checklist of what the deck covers, in roughly this order, not a slide template. Combine or split slides as your story needs.

1. **Title.**
2. **The answers.** The Controller's two questions, each answered in a line with its dollar figure.
3. **How you connected the data.** The rule table from the memo, with the footnote you found. The balances rebuilt and tied to the GL, with the reconciling item explained. How you decided last owner contact, and how you cleaned the states and resolved each owner's jurisdiction, with every owner accounted for. How you read Schedule A off the image and checked it.
4. **The data problems.** The address problems, the reversals, the standing top-ups, and what each did to the number.
5. **What you found.** The dormant population: counts and dollars by state and property type, with a chart of the dormant dollars by state, the Delaware amount and why, and the letter count.
6. **The recommendation.** Outreach timing, dollars at stake, and your call: what the Controller should do, who does it, and by when. Name any process change your findings call for, or say why none is needed.
7. **Limitations.** What your numbers assume, and what would change them.
8. **Thank you, then the appendix.** Extended tables and anything else a reviewer might want to check.

Polish counts, under professionalism and formatting in the rubric: slide numbers, labeled axes, a title on every body slide and every chart that states the finding rather than the topic (the title slide, the thank-you slide and the appendix are exempt), consistent number formats, no code on slides, and clean writing free of spelling and grammar errors. Extended tables go in an appendix.

### 4.3 The executive one-pager

One page, for a reader who will not open the deck. It answers both questions, with one chart, the dollar figures that matter, and the recommendation. It is a summary, not a second deck, so it should be scannable in a minute. A stakeholder should be able to act on it without asking you anything. I strongly suggest you open with the action you want the reader to take and let one chart make the main point.

A strong one-pager has:

- A title that states the point.
- A clear signal of what it is: information, or a decision you are asking for.
- An answer to every key question.
- One chart.
- The dollar figures that matter.
- The main caveat on your numbers.
- Bullets rather than paragraphs.
- Action language: who does what, by when.
- No wasted words.

[An example one-pager, on a different topic](https://seanmccaman.com/acctg6155/2026-fall/project-2/onepager_example.html), shows the format.

### 4.4 Using AI, and checking it

I encourage you to use AI throughout this project. It is good at reading a table out of an image and at writing the SQL and the pandas, and it is a reasonable partner for turning the memo into a table, as long as you check the table against the memo and not against the AI. When it reads Schedule A for you, check its table against the image row by row. Where it will hurt you is the rules: it knows a version of unclaimed property law, and that version is not the one in the memo. It will also happily count every ledger row as activity if you let it, because it does not know which channels are owner-initiated unless you tell it what the memo says.

Write your prompts the way the Example Managing-Up Document from Week 1 describes: the context, the goal, and the shape of the answer you want.

You do not need an AI slide in the deck. The quiz's closing questions ask which AI tools you used and how.

The syllabus rule on collaboration applies: talk through approaches with classmates, but the analysis and the write-up are your own.

### 4.5 Make the AI explain what it did

This applies to everything in the project. If I ask you about your work and you cannot explain it to me, you don't own it. Owning it means you can say what each step does, why it is there, and how you checked one number.

Whenever an AI builds something for you, such as a query, a formula, a chart or a method, make it explain what it did before you use it. Ask it to walk through the work one step at a time, in plain words, and to say why each step is there. Keep asking until you can explain it to the Controller with the AI closed. To test yourself, ask the AI for two questions about it, answer them without scrolling back, and ask whether you got them right. Then check one number by hand: pick one owner or one row and follow it through.

A prompt that works:

> Here is a [query, formula or method]: [paste it]. It should [say in one sentence what it produces, and which rule from the memo it applies]. Explain what it does, one step at a time, in plain words, as if I have to defend it to an auditor. For each step, tell me why it is there and what would change if I left it out. List every assumption it makes about the data, so I can check each one against the memo. Then tell me how to check one row of my output by hand: which rows of the raw data to pull, and what to add up or compare. Do not make up the numbers. I will run the check myself.

The second sentence matters most, because the AI can only check the work against what you say it should do.

You don't turn these explanations in. Keep your own notes, so you can still answer if I ask after you submit.

## 5. Rubric

The rubric has the same shape as Project 1's. Each criterion has a few levels, and I pick the one that matches what I see: Excellent, Good, Needs Work, or Missing. Excellent is full points, Good is three quarters, Needs Work is half, and Missing is zero. The Excellent line works as a checklist, so to earn Excellent your work has to show everything it names. The last criterion, the Schedule A table, is all or nothing.

| Part | Criterion | Points |
|---|---|--:|
| Slide deck | Storytelling & Slide Flow | 16 |
| Slide deck | Data Preparation & Documentation | 12 |
| Slide deck | Findings & Evidence | 8 |
| Slide deck | Conclusions & Recommendations | 8 |
| Slide deck | Visuals | 16 |
| Slide deck | Professionalism & Formatting | 10 |
| Executive one-pager | One-Pager: Leads with the Point | 4 |
| Executive one-pager | One-Pager: Format | 4 |
| Executive one-pager | One-Pager: Answers & Numbers | 4 |
| Executive one-pager | One-Pager: Chart | 4 |
| Executive one-pager | One-Pager: Recommendation & Caveat | 4 |
| Data support file | Data Support: Every Number, Labeled | 4 |
| Data support file | Data Support: Method & Letter List | 4 |
| Data support file | Data Support: Schedule A Table | 2 |
| | **Total** | **100** |

The slide deck is 70 points, the executive one-pager 20, and the data support file 10, the same as the three upload questions in the quiz.

### Slide deck (70)

**1. Storytelling & Slide Flow (16).** How quickly the Controller gets the answers, and how well the slides build to them.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 16 | The answers to both questions, each with its dollar figure, come first. Every body slide's title states a finding (the title slide, the thank-you slide and the appendix are exempt). Every slide supports a question or the recommendation. Dormancy and owner contact are explained in plain words. |
| Good | 12 | The answers come first, but some titles are topics, one slide wanders, or one key term goes unexplained. |
| Needs Work | 8 | The answers are buried or missing a dollar figure, titles are topics throughout, or the slides are hard to follow. |

**2. Data Preparation & Documentation (12).** How well the deck shows the data was extracted, connected and cleaned.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 12 | Schedule A read off the image into a table, and how it was checked against the image. The rule table from the memo, including what the footnote changes. Cash and promotional balances rebuilt and tied to GL 2110 and 2115, with the cash difference explained by cause and amount, not plugged. How last owner contact was decided. How states were cleaned, with every owner accounted for. The data problems found, and what each did to a number. |
| Good | 9 | Most of this is shown, with one piece missing or thin, such as the footnote missed, no check of the Schedule A table, or the GL difference plugged. |
| Needs Work | 6 | Several pieces are missing, so the reader cannot tell how the numbers were built. |
| Missing | 0 | Not turned in, or nothing to grade. |

**3. Findings & Evidence (8).** Whether the deck reports what is owed, to whom, and the letters, and gets items 1 to 7 right.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 8 | The dormant schedule by state and property type, with counts and reportable dollars. The item 6 amount, and why each group of owners reports there. The letter count. Every number in items 1 to 7 is right at the published tolerances: counts exact, dollars within $1. A mistake that flows into later numbers counts once, where it starts. |
| Good | 6 | All of it is shown, with one number off or one piece missing. |
| Needs Work | 4 | Two or more numbers are off, or the schedule is incomplete. |
| Missing | 0 | Not turned in, or nothing to grade. |

**4. Conclusions & Recommendations (8).** Whether the recommendation is specific and uses the numbers.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 8 | A specific call: what the Controller should do, who does it, and by when. How much money reaches its anniversary in a window you state, and what is at stake if owners act before then. The recommendation uses your numbers. Any process change the findings call for, or why none is needed. |
| Good | 6 | A clear call with one piece missing, such as no owner or date, or no window. |
| Needs Work | 4 | A vague recommendation that is not tied to the numbers. |
| Missing | 0 | Not turned in, or nothing to grade. |

**5. Visuals (16).** How clear and informative the charts are.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 16 | The dormant dollars by state shown visually. Labeled axes, with units, on every chart, readable at normal zoom. Every chart's title states the finding, not the topic. Every chart supports a finding the deck states. |
| Good | 12 | Clear charts with one gap, such as a topic title or a missing chart. |
| Needs Work | 8 | Charts that are hard to read or not tied to the findings. |
| Missing | 0 | Not turned in, or nothing to grade. |

**6. Professionalism & Formatting (10).** How professional the slides look and read.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 10 | Slide numbers. Consistent number formats for dollars and counts. No code on slides. Extended tables in an appendix. Clean writing, free of spelling and grammar errors. |
| Good | 7.5 | A reasonable appearance with minor formatting inconsistencies. |
| Needs Work | 5 | Formatting or grammar issues, or inconsistent design. |

### Executive one-pager (20)

**7. One-Pager: Leads with the Point (4).** Whether the page says what it is for.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | The title states the point, and the page says whether it informs or asks for a decision. |
| Good | 3 | One of the two: a topic title, or no inform-or-decide. |
| Needs Work | 2 | Neither. |
| Missing | 0 | No one-pager. |

**8. One-Pager: Format (4).** Whether it reads like a one-pager.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | One page, mostly bullets. A short bottom line is fine. |
| Good | 3 | One page, but dense. |
| Needs Work | 2 | More than a page, or paragraphs instead of bullets. |
| Missing | 0 | No one-pager. |

**9. One-Pager: Answers & Numbers (4).** Whether a busy reader gets both answers.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | Both of the Controller's questions are answered, with the dollar figures that matter. |
| Good | 3 | One question or one key figure is missing. |
| Needs Work | 2 | Two or more are missing. |
| Missing | 0 | No one-pager. |

**10. One-Pager: Chart (4).** Whether the chart earns its place.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | One chart that makes the page's main point. |
| Good | 3 | A chart on a secondary point. |
| Needs Work | 2 | A chart that is unclear. |
| Missing | 0 | No chart. |

**11. One-Pager: Recommendation & Caveat (4).** Whether the page ends in an action.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | The recommendation as an action, with who owns it, the date and the dollars, and the main caveat on your numbers. |
| Good | 3 | One of the two is missing or thin. |
| Needs Work | 2 | Both are thin. |
| Missing | 0 | No one-pager. |

### Data support file (10)

**12. Data Support: Every Number, Labeled (4).** Whether every number can be found.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | Every 3.4 item, 1 to 7, is present, labeled with its item number, and in the requested units. |
| Good | 3 | One or two items are missing or unlabeled. |
| Needs Work | 2 | Several items are missing. |
| Missing | 0 | No file, or it is empty. |

**13. Data Support: Method & Letter List (4).** Whether the work behind the numbers shows.

| Level | Points | What it looks like |
|---|--:|---|
| Excellent | 4 | The code, formulas or queries that produced each number, with their results showing. The item 7 letter list in full. If you used SQL, all of it is in one file or inside your workbook or notebook. |
| Good | 3 | The method shows for most numbers, or the letter list is cut short. |
| Needs Work | 2 | Numbers with no visible method. |
| Missing | 0 | No file, or it is empty. |

**14. Data Support: Schedule A Table (2).** Whether the table you extracted from Schedule A is there.

| Level | Points | What it looks like |
|---|--:|---|
| Exists | 2 | The table you extracted from Schedule A, all 50 rows, with the state code you mapped each one to. |
| Does not exist | 0 | No Schedule A table. |

**An optional AI review before you submit.** The [Example Managing-Up Document](https://seanmccaman.com/acctg6155/2026-fall/week-1/example_managing_up_doc.html) from Week 1 explains how I read and review your work. Give an AI this guide, the rubric, and that document, and ask it to review your draft against the Excellent lines. It is the same first pass I do.