Algorithms, Data and Digital Business

Lab 1 · Data · 30 minutes · Practice context: ibds-26-s03-d04

What does a viewing record actually tell us?

Your brief: A fictional community-event platform wants to estimate which topics interest visitors. Decide whether its viewing log is suitable evidence. Reading: pp. 72–74. Preparation: Data and analytics teaching deck.

Different-case example: A 12-second museum clip played for 6 seconds has a recorded completion ratio of 0.5. That is a duration ratio, not proof of enjoyment. A replay can make total viewing exceed one clip length; investigate before calling it impossible.

1. Predict individually — 3 minutes: Which topic seems most popular? State your measurement and one assumption.
2. Investigate in pairs — 10 minutes: Annotate every questionable record. E1/E2 might be a duplicate or a genuine repeat; E5 might record replay or an error. State what extra information would distinguish these possibilities. Keep the original log unchanged.
3. Compare — 7 minutes: Calculate Music’s mean watched seconds with both rows, then with one provisionally excluded. Why can an unchanged mean still conceal a distorted count? Calculate Art’s mean for recorded values only, then with the blank assumed zero. Label the denominator and assumption in each result.
4. Explain individually — 5 minutes: Choose a defensible cleaning policy and give an alternative interpretation of viewing. Identify one veracity issue and one limitation of using this small dataset.
5. Correct — 5 minutes: Swap explanations. Challenge one assumption. Retain your first response and add a dated correction. Do not silently replace it.

Check your work: Your record identifies observed versus inferred information, shows both denominators, explains a cleaning choice, and states a limit. A blank is not automatically zero. More rows cannot by themselves repair a biased collection method.

Fresh individual check: A second platform records only signed-in viewers. Explain one claim it can support and one population-level claim it cannot yet support.

Real-world connection: Read the interaction-signals section of TikTok’s 2020 provider explanation. Locate one recorded signal and explain a possible inference. It is a historical provider account, not disclosure of current scoring weights.

Lab 2 · Algorithms · 30 minutes · Practice before ibds-26-s03-d08

One dataset, two objectives

Your brief: A fictional event platform ranks notices. Its editor wants newcomers to encounter accessible activities as well as familiar interests. Explain the rule before proposing a change. Reading: core 3.2A–E, pp. 81–89. Preparation: Algorithms and algorithmic dilemmas.

The practice rule is score = 3 × interest_match + completion_ratio. Match is 0 or 1. Rank larger scores first; break equal scores by notice ID alphabetically. All notices are eligible initially.

Different-case example: A gardening notice with match 0 and completion 0.4 scores 0.4. A condition such as “include only notices with complete measurements” is a filtering rule; the score orders those that remain. Keep these operations distinct.

1. Predict individually — 3 minutes: Sketch the likely order without calculating every score. Identify the tie-break rule.
2. Trace in pairs — 8 minutes: Calculate the four scores and sort them. Write a finite loop that visits each notice once. Name the input, changing variable, condition if any, and stopping point.
3. Redesign — 8 minutes: Choose a newcomer weight from 0 to 4 and add weight × newcomer_suitable to the practice rule. Predict which positions may change, recalculate, and explain one changed or unchanged position. You choose the weight; there is no universally correct value.
4. Test — 6 minutes: Write a new boundary case with a missing ratio or a tie. State the handling policy before testing it. Then use Tune the Feed to examine a similar relationship on its separate dataset. It cannot check your notice-table arithmetic.
5. Defend individually — 5 minutes: Explain the intermediate calculation, one stakeholder trade-off, and what evidence would be needed before claiming the rule improved participation.

Check your work: Retain prediction, original scores/order, chosen rule, new scores/order, boundary test and defence. Correct arithmetic does not settle whether an objective is fair. This synthetic rule cannot predict a real platform’s audience behaviour.

P1-style formative application: Explain how the changed objective could alter visibility for a notice publisher. Use an intermediate operation and qualify the consequence. No new scored task or maximum is assigned.

Independent boundary: Put this rehearsal away before the registered ibds-26-s03-d08 /8 task. Use its original dataset, instructions and permitted conditions.

Lab 3 · Business and labour · 30 minutes · Practice context: ibds-26-s03-d06; revisit d07/d09

Percentage of what—and evidence for whom?

Your brief: A fictional creator is choosing between long-form and short-form production. Evaluate what a revenue-share statement establishes. Reading: pp. 213–219. Preparation: Business and labour teaching deck.

Advertiser pays → platform applies its revenue rules → eligible creator receives a share

The arrows describe an exchange. The amount available to share and the applicable conditions still need evidence.

Real source: YouTube’s partner earnings overview, checked 11 October 2026, describes a 55% share of net Watch Page advertising revenue and a 45% share of allocated Shorts Creator Pool revenue under the relevant modules. Those are different bases. The rates alone do not establish an individual creator’s income.

Different-case example: In a fictional licensing agreement, 20% of a $50 eligible base is $10. Twenty per cent of total sales would be a different calculation if only part of those sales was eligible.

1. Predict individually — 3 minutes: Is the larger percentage enough to choose a format? Give a reason before calculating.
2. Calculate in pairs — 7 minutes: Find each hypothetical payment and payment per stated creation hour. Label both bases. These ratios omit expenses, unpaid work and tax; do not call them wages or profit.
3. Read and locate — 7 minutes: Find the provider’s revenue-share terms. Record the heading, provider purpose, one supported claim and one unanswered question. Identify a source you would seek for workers’ experiences; do not invent its findings.
4. Decide individually — 8 minutes: Explain a conditional recommendation. Connect a platform rule to a creator opportunity or risk. State a cost or uncertainty and the evidence that could change your recommendation.
5. Converse and correct — 5 minutes: A partner challenges your denominator or inference. Record the challenge, retained original claim and your correction. Your teacher may ask you to explain it aloud.

Check your work: Accurate bases and calculations; a claim tied to an exact source locator; a clear distinction between provider terms and evidence of lived outcomes; a conditional judgement.

P2-style formative preparation: Explain how an independent worker study could complement this provider source and where their purposes might differ. This exercise has one supplied source, so it is preparation for synthesis—not a full two-source Paper 2 response. Use both assigned provider/regulator sources in the original ibds-26-s03-d07 synthesis.

Later retrieval: Without notes, explain a new percentage statement’s denominator and one reason a technical efficiency gain might not improve every worker’s experience.