Midterm Progress Review

From the organization chart to a midterm audit

Take out your Week 8 organization chart. It shows Judy as director, with three direct reports: Mina the coordinator, a part-time performer liaison who speaks Korean, and the bookkeeper. The translator is a freelancer, and the performer is not an employee. Put your Week 3 decision memo next to it. The memo says which option the company is testing, why, and who decides.

Since Week 3, each artifact has added claims about the outside world, the company's goals and its people. Nobody has yet checked whether these claims still fit together.

That is the job of a midterm. A midterm evidence audit is a structured check of the claims a plan depends on. For each claim, it asks what the source actually shows, what Judy's staff added by reasoning, and what would change their minds. It ends with an updated decision memo. The audit looks for the places where the argument has become stronger than its evidence.

To make this concrete, the chapter continues the practice scenario, which is invented for learning. Judy's company is testing Option A · Expansion (expand cheerleading and sports-entertainment talent management) for 12 weeks. The Week 5 goal tree set a review in test week 6, and this chapter treats the midterm audit as that review. Game 1 was played on the Saturday of test week 4. Because the home team pays each fee 30 days after the game, the game 1 fee is not yet due.

This week's question is: which parts of our argument survive a careful check, and what must change before the company commits more people, time or money?

Where this week fits

Week 9 sits at the middle of the course. It takes the artifacts from the first eight weeks and checks the claims that connect them. The result is an evidence audit and an updated decision memo. Week 10 turns the audit's findings into communication rules, starting with one question: who actually tells the performer about a change?

Four kinds of statement

In Week 1 you learned to sort statements into evidence, inference and unknown. A midterm adds a fourth kind: design, which records something the company decided or invented in order to test an idea. Each kind needs a different kind of support.

• Evidence is an observation you can point to in a source. Example: "A public source describes a collaboration." Its support is the source, with the date you checked it.
• Inference is a reasoned interpretation of the evidence. Example: "The collaboration may show access to a network that could support a cross-border service." It needs a believable reason and at least one other possible meaning.
• Unknown is a question that matters but that the sources cannot yet answer. Example: "We do not know whether the relationship is repeatable, contractual or profitable." It needs a next step and an owner.
• Design is a choice the company makes in order to test something. Example: "For a 12-week test, we assign a coordinator and a budget limit." It should be labeled as a choice.

The most common error is compression: one kind of statement quietly turns into a stronger kind. A public event becomes proof of a capability, or an invented budget becomes proof of affordability. Each step sounds more confident, and the uncertainty a manager most needs to see disappears.

The claim chain: one frame for the whole audit

This chapter uses one frame for the audit: the claim chain. A claim chain is the path from something observed in a source to a decision the company makes. It has five links.

1. Observation. What does the source directly show?
2. Inference. What might the observation mean, and what other meanings are possible?
3. Hypothesis. What business idea are we proposing to test?
4. Design. Who will test it, with what resources and safeguards?
5. Review. Which result would support the hypothesis, narrow it or reject it?

Each link carries a different burden. An observation needs an accurate source, and an inference needs a believable reason. A hypothesis needs a clear Target, Value, Capability and Revenue Model. A design needs an owner, a safeguard and a signal. A review needs a trigger written before the result is known, so that nobody can change the standard after seeing the result.

The other tools fit inside this chain. Evidence sits at the observation link, and inference at the inference link. The hypothesis and the design are the company's own choices, so they count as design. Unknowns can appear at any link. Confidence is a label on each link that says how well that link is supported.

Give each link its own level, and do not average them. A claim can be high-confidence as an observation and low-confidence as an interpretation. A low-confidence link about consent or safety cannot be rescued by a high-confidence link about publicity.

Worked example: tracing the Taipei East page

The case gives two public observations to work with. Judy's public YouTube channel, @judy777, publishes Korean-language videos. And Taipei East, a professional volleyball team in Taipei, has published material about Tokki Cutie. The team describes Tokki Cutie as a cheerleading collaboration involving OM Entertainment, a company it presents as led by Judy.

The Option A test grew from the second observation. The table follows it through the chain. The first link comes from the case; the later links belong to the practice scenario.

Notice where the confidence drops. The jump from "one team worked with the company" to "another team will sign and pay" is the weakest link. That is why the company chose a small test, aimed at that link, instead of a large commitment.

One over-claim and its correction

Now look at a sentence that one of Judy's staff wrote into the Week 7 capability inventory.

Over-claim: "Taipei East's page proves that Judy's company can already run cheerleading appearances for professional teams, so it has the capability to serve more teams."

The sentence skips two links. It treats an observation as proof of a capability, and one collaboration as a routine the company can repeat. The page is a team's public description of a collaboration. Who did which work, and how problems were handled, are not known from public sources. And as Week 7 showed, a capability is something the company can do again and again without depending on one person's effort.

Corrected sentence: "Taipei East's public page describes a cheerleading collaboration involving a company it presents as led by Judy. This suggests a relationship with at least one professional team. Whether the company can deliver game appearances as a repeatable routine is what the Option A test is designed to find out."

The correction keeps what is useful and removes only what the evidence cannot carry.

The evidence matrix: two filled rows

An evidence matrix is a table with one row for each important claim. It helps the company decide which claims need attention now and which can wait. The two rows below come from the worked example: a narrow claim the source can carry, and the over-claim with its correction.

Two habits make the matrix useful. Give each claim its own row, so that one source is not asked to support claims it never made. And use the "if false" column to decide what to research. A detail that could not change the decision can stay as background. A detail that could reverse the decision needs a specific request, an owner and a date.

Match the strength of the sentence to the source

A sentence that uses a research source should be no stronger than the source itself. Here is one example from Week 6. Palich, Cardinal and Miller (2000) combined the results of many earlier studies and found that firms with moderate diversification tended to perform better than those with very little or a great deal. Compare three sentences:

• Too strong: "Research proves that Judy's company should run exactly two business lines." The studies were about large firms, and they give no number of lines for a small company.
• Too weak: "Research has nothing to say about this." The finding still warns against "more is always better".
• Matched: "Research on large firms suggests that both too little and too much diversification can hurt performance. So our portfolio table should show what the content work and the Option A test share, and what running both costs."

The matched sentence keeps the finding and its setting, and turns it into a question the company can check.

Audit 1: outside conditions and goals

The first audit checks the Week 4 environment map and the Week 5 goal tree. Does each market claim name a mechanism, such as a rule or a payment practice? And does each goal still fit the outside condition it depends on?

In the practice scenario, the key outside condition was whether the performer may be paid for game appearances in Taiwan. The goal tree said the test would be redesigned if this was not confirmed by test week 3. The audit finds that a qualified adviser answered in writing in test week 3, and Mina recorded the date. The claim moves from unknown to evidence.

The goal audit also asks what behavior each goal rewards. Locke and Latham (2002) showed that specific goals help when people are committed, get feedback and understand the task. Ordóñez and colleagues (2009) warned that narrow targets can push people to hide problems or take excessive risks. Operating goal 2 asks Judy to hold payment talks with two more teams by test week 8. Read as "sign two more teams", it could push Judy to promise dates before anyone checks Mina's hours.

Audit 2: scope and capabilities

The second audit checks the Week 6 portfolio table and the Week 7 capability inventory. During the test, the company keeps making videos, including sponsored ones, and runs a small food-service line. The goal tree said that new sponsored-video deals would wait. The audit finds that one new sponsored request was accepted in test week 4 because it seemed small, and it used several of Judy's hours in the week of game 1. This is scope creep: extra work that nobody formally decided to take on.

The audit also tests claims of synergy. Kumar (2013) found that in diversified firms, related business units could become less productive when shared work mainly served the core business. In the scenario, Mina's bilingual coordination serves both the content work and the Option A test. Both draw on Mina's same 40-hour week, so the sharing may simply move work onto her.

Week 7 asked whether "cross-border coordination" is a capability of the company or a skill of Mina, and what would happen if Mina were sick on game day. The audit checks whether the answer is written down. It is only partly written. The Week 8 chart names the translator as Mina's backup, but that exists only on paper: her backup training has not happened, and Mina's pre-game checks exist only in Mina's notes. Until that changes, the capability claim stays at low confidence.

Audit 3: organization and people

The third audit compares the Week 8 organization chart with what happened around game 1. The chart gives messages to the performer to the performer liaison. In practice, most of those messages went through Mina, and twice Judy messaged the performer directly. Nobody broke a rule on purpose; the chart simply did not match the way information moved.

The plan assumed that Mina's test work would take about 3 test hours in an ordinary week (plus about 1 hour briefing the partner), and about 7–8 in the week before a game. In the week before game 1, test week 3, she used about 10 hours: all of her reserved hours. Part of that time went to work the chart had given to someone else. This staffing assumption is not supported by the evidence, and it is the main finding Week 10 will work on.

Next, ask who carries the burden. Week 8 gave the performer the right to refuse a schedule that breaks her 12-hour rest rule; the audit checks whether she knows how to use it. Week 8 also taught psychological safety, and Edmondson (1999) found that teams with more of it showed more learning behavior. The practical check: can the translator or the performer raise a concern, get an answer, and keep their future work?

Audit findings at the test week 6 review

The table collects the findings from the three audits.

Contradiction tests

A claim can look fine on its own and still conflict with a claim from another week. A contradiction test puts two claims side by side and asks whether both can be true at once.

When a conflict is real, do not simply keep the more optimistic statement. Decide which assumption needs evidence and which safeguard cannot be traded. In the scenario, the first row is live: the goal tree requires focus, but the portfolio shows a new sponsored deal. A written rule for exceptions is the correction.

What would weaken a claim, and which lines hold

An audit becomes useful when it says what could weaken the current interpretation. For each major claim, write an observation that supports it, a competing explanation, and an observation that would count against it. The claim "our coordination is repeatable" would be rejected if it works only through Mina's unrecorded overtime.

Next, separate ordinary criteria from red lines. A criterion is something the company can trade off within a range, such as cost or speed of learning. A red line, also called a stop condition, means the company pauses, redesigns or refuses the commitment, whatever else looks good. Week 5 introduced red lines. At midterm, check four groups:

• Consent: the performer has agreed in writing to the terms and the use of her image.
• Lawful, clear payment: the performer may be paid, and everyone knows who pays what, and when.
• Safety and rest: someone with authority can act at once on a safety or rest problem.
• Correction: a harmful decision can be corrected, and the person who raised it is not punished.

A red line stops the company from continuing after a known boundary has been crossed. The team may revise one when new evidence arrives, but it should record the change and the reason. Otherwise the standard changes after the result is known, and every decision looks successful.

The midterm audit in seven steps

The audit should leave a trace that another manager can follow, from the recommendation back to the evidence and forward to the next action. It corrects without blaming: it keeps what earlier weeks got right and marks what still needs testing.

1. List the recommendation, not the facts. For example: "Continue the Option A test to test week 12."
2. List what the recommendation needs. Name the customer, the capability and the safeguard it depends on.
3. Connect each need to a claim and its source. Use the claim chain to see which link each claim sits on.
4. Label each claim. Is it evidence, inference, unknown or design, and how confident are you?
5. State the consequence if the claim is false. If it would change only a word, it is minor. If it would change whether people are safe, whether the cash plan works or whether the promise can be kept, it is major.
6. Write the next information request. "Research more" is not an action. "Judy asks the next team whether it can pay within 30 days of the game, and records the answer by test week 8" is an action, with an owner, a method and a decision it could change.
7. Decide the level of commitment. A weakly supported claim may justify a conversation, a paper exercise or a small test. It does not justify a full launch.

The three options at midterm

Each option depends on different claims and has its own weakest link.

You will compare the three options as whole systems, and in Week 17 you choose one and defend it.

When sources or teammates disagree

Disagreement is useful information. First, check whether the two sources are about the same question. Then name the type of disagreement. A factual one needs a check of the original record. One about interpretation becomes a hypothesis to test, and one about values needs a decision by the person responsible.

Record the disagreement in the decision memo, and do not let the most confident writer win. Treat a classmate's experience the same way: as in Week 4, ask which rule or practice produced it. The answer may become a new row in the evidence matrix.

Checking AI help during the audit

An AI summary is not a source. In a midterm audit, the specific risk is that a tool makes a claim chain sound stronger than it is, for example by turning "linked with" into "caused" or dropping an open question. Either sends the claim back to the original source.

Updating the decision memo

The audit ends by updating the Week 3 decision memo with conditional language, which states the conditions under which the recommendation holds. Conditional language shows that the writer understands the evidence. Compare two sentences:

• "The boutique service is the best strategy."
• "If the customer problem is verified, if the team can staff the required routine, and if the safeguard can be funded, the boutique service is the next option to test."

The second sentence tells a manager what must be checked and what would make the recommendation wrong. A good update also says what it sets aside for now, and why.

Worked example: the updated decision memo

Here is how Judy's staff update the memo, using the audit findings.

The update declares neither success nor failure. It records what must happen next.

Class practice: a midterm evidence audit

This practice is done without AI tools. Your teacher will tell you which worksheet to use and how to submit it. Bring your decision memo and your artifacts from Weeks 4 to 8.

1. Trace one claim. Choose the claim your option depends on most. Write its observation, inference and hypothesis, and give each link a confidence level.
2. Check the design and the review. Write the test that would check the hypothesis, and one result that would make you narrow or stop it.
3. Run one contradiction test. Put your claim next to a claim from another week's artifact. If they conflict, write the correction.
4. Update your decision memo. Write one conditional recommendation, one next evidence request with an owner and a date, and one review trigger.

A classmate should be able to challenge your reasoning using only what you wrote. Label any detail you invent, and keep the updated memo for Week 10.

Key points and what you bring to Week 10

• A midterm evidence audit checks the claims a plan depends on and ends with an updated decision memo.
• Use one frame, the claim chain: observation, inference, hypothesis, design and review. Label each link and give it its own confidence level.
• Watch for compression, where an observation silently becomes a capability. Correct the sentence, and keep the useful idea.
• Test claims from different weeks against each other, and keep red lines out of any average.
• A strong recommendation is conditional. It says what must be true, what would change it and who checks.

For Week 10. Bring your evidence audit and your updated decision memo. Week 10 turns audit findings into a communication protocol. In the practice scenario, it starts from the gap between the organization chart and the way messages actually reached the performer.