Fictional practice, not an employer interview question.
Invent one million equal-size event records. Region A contributes 950,000 events, including 95 failures. Region B contributes 50,000 events, including 250 failures. Every failure is included in its region’s event total. These are chosen counts, not actual customer data or a vendor cost estimate.
| Recorded population | Events | Failures | Failure rate |
|---|---|---|---|
| Region A | 950,000 | 95 | 0.01% |
| Region B | 50,000 | 250 | 0.50% |
| Both regions | 1,000,000 | 345 | 0.0345% |
Work through the case
B has 5% of the events but about 72.5% of the failures: 250 ÷ 345. Its failure rate is 50 times A’s. A combined failure rate hides where the problem is concentrated.
Proposal one retains all 345 failed events and a sample of 99,655 successful events, keeping 100,000 records. It also retains aggregate event totals by region. Proposal two keeps the same record count but removes the region field and keeps only the combined denominator. Both reduce retained record count by 90%; only the first preserves the data needed to calculate the regional rates in this example.
Prepare your application
Do not make “less telemetry” the whole result of an observability project. Show whether a person can still answer the question the data was collected for. A small synthetic event file is enough to demonstrate what a cost-cutting proposal preserves and what it destroys.
Match the example to the internship
Datadog’s early-career page separates technical and business internships. Read the named department rather than assuming every student role requires the same coding portfolio. An engineering candidate can implement the exercise below; a product or customer candidate can write the investigation questions, compare options and explain the operational consequence.
Use the exact posting for timing and location. Joining a talent community is different from submitting that application. A project can help you prepare, but do not delay a suitable open application to finish an unnecessarily large monitoring stack.
Do not mistake the retained sample for the original population
Under the first proposal, 345 of the 100,000 retained records are failures: 0.345%. That is ten times the original overall rate, because failures and successes were sampled differently. Use the retained population counts or a justified weighting method when estimating rates; do not treat the retained sample as a simple random sample.
Turn this into a local exercise with a script that creates the fictional records, applies each proposal and prints the questions still answerable. Ask: Which region accounts for the failures? What was its rate? Which individual event would you inspect next? Document any question you can no longer answer. A field removed permanently cannot be recovered by changing the dashboard.
Describe the work without inventing savings
A truthful project entry, after doing it, could be: “Compared two retention rules on synthetic event data; preserved regional denominators and showed how selective sampling changes the apparent failure rate.” The 90% record-count reduction is not automatically a 90% bill reduction, because a real bill may use other measures.
In an interview, explain which fields you retained, why the regional denominator matters and what your toy example leaves out. For the team, ask: “Which investigation should still be possible after this data is aggregated or discarded?” That question ties the proposed change to the person relying on it.
Sources and related guides
Official sources
Official sources and current instructions
The source collection was revised September 5, 2026. These links help you find the employer’s current guidance. This guide does not establish a current opening, eligibility, deadline, or interview process.
About this guide
By InternshipDeadlines · AI-assisted editorial adaptation · Updated September 17, 2026
Original practice material, not a reported employer interview question. Scenarios are hypothetical. Original fictional telemetry exercise; not Datadog pricing or a reported interview question.
Published by InternshipDeadlines. Guide updated . How we use sources.
Cite this guide
InternshipDeadlines. Datadog internship application: make telemetry useful for a decision. Updated . https://internshipdeadlines.com/career-guides/datadog/
Use the dates attached to the sources behind your claim. An editorial update does not reverify every source or confirm that an internship is open.