How to create a case-study assessment

What you'll need

• A SkillCort workspace (or a demo)
• A real situation the role has faced — a deal, a client problem, a strategic call
• The decision or recommendation you want the case to surface
• Someone senior in the role to pressure-test the brief

Step 1: Decide what the case should reveal

A case study is only as good as the question behind it. Before writing anything, name the two or three capabilities you want to observe: does the candidate separate signal from noise in messy information, structure a recommendation, quantify a trade-off, communicate a decision persuasively? Different roles weight these differently — a sales case leans on discovery and prioritization, a consulting case on structure, a management case on people trade-offs.

Then choose the decision the case forces. Good cases end in a recommendation with consequences: which account to save, whether to enter the market, how to handle the underperformer. Cases that only ask for description produce essays; cases that force a decision produce evidence.

Step 2: Write the brief from real material

Build the brief from an anonymized real situation, keeping the messy parts: numbers that do not quite reconcile, a stakeholder quote that cuts against the data, one genuinely irrelevant detail. Strong candidates show themselves by what they choose to ignore.

Keep the brief to one or two screens — enough context to reason with, not a data dump that turns the exercise into a reading test. Include two or three exhibits at most: a short table of figures, an excerpt from an email, a summary of the account history. If you draft with AI generation, upload the real source document as grounding and edit heavily by hand.

• One central decision, stated plainly at the end of the brief
• Real (anonymized) numbers, including at least one imperfection
• A constraint that forces a trade-off — budget, time, or politics
• No insider knowledge a new hire could not have

Step 3: Build the item with the side-by-side stimulus layout

Create a folder for the role in the items library, then create the item using the side-by-side stimulus layout: the brief and exhibits stay pinned on one side while the answer area sits beside them. Candidates reference the case while writing instead of scrolling back and forth — the difference between testing analysis and testing short-term memory.

Tag the item by capability and difficulty in the folder's tag set. Case studies are expensive to write well, so treat each one as a reusable asset: when you assemble future assessments, the picker's folder, tag, and type filters bring it back in seconds.

Step 4: Choose the response format: rich text or file upload

For most cases, the rich-text response item is the right default: candidates structure a recommendation with headings, lists, and emphasis, and evaluators read every submission in one consistent place with no downloads. It suits consulting-style memos, sales account plans, and management write-ups — anywhere the reasoning itself is the deliverable.

Use the file-upload item when the artifact itself is the skill — a slide deck for a consulting role, a spreadsheet model for a commercial one. You can also combine them on one page: a rich-text executive summary plus an uploaded model. Ask for the same artifact the job would demand, because that is the evidence the Decision Board will hold later.

Step 5: Build a matrix rubric for analysis quality

Open-ended analysis needs a level-based matrix rubric, written before any candidate responds. Typical criteria: problem structuring, use of evidence from the brief, quality of the recommendation, and communication. Anchor every cell with observable behavior — "recommendation follows from the analysis and names its main risk" is scoreable; "good strategic thinking" is not.

Weight the criteria for the role: recommendation quality heaviest for a sales case, structuring for consulting, stakeholder handling for management. Consider making the rubric candidate-visible so people optimize for substance over length. Attaching the rubric freezes it into a snapshot, so mid-run edits to the bank can never shift the standard under your candidates.

Step 6: Assemble and timebox in the builder

Create the assessment with the case in its own section and import the item through the picker. Give the case section the dominant scoring weight — ×3 or more against any warm-up section — so weighted role-fit on the Decision Board reflects the analysis, not the preliminaries.

Timeboxing is where case studies most often go wrong. Set a total limit of 60–90 minutes and a per-page limit on the case page: generous enough for real thinking, firm enough that the test does not reward whoever had a free weekend. State the timebox in the candidate instructions along with the expected format, and say explicitly that a focused, imperfect answer beats a polished, incomplete one.

Step 7: Preview, publish, and evaluate on evidence

Run Preview as candidate from the builder header and attempt the case in the timebox yourself — or give it to a strong current employee. If they cannot produce a solid answer in the limit, cut exhibits or narrow the question. Then clear the publish checklist, which blocks Publish until scoring, timing, and instructions are in place.

Publish and create the delivery with a name, time window, and light-touch security — for a take-home-style case, copy/paste blocking is often all that is proportional. Invitations go by email and candidates pass the consent and preflight gate before starting.

In evaluation, multiple evaluators score the same frozen matrix by clicking level cards, and AI can draft a summary of each response with suggested criterion scores — model and version logged, with a named person confirming or overriding every one. The Decision Board then compares candidates on weighted role-fit, with each memo or deck one drill-down away.