Document Review & E-Discovery Math: Data Volumes, Culling, Review Rates, Sampling & Recall

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Convert the gigabytes collected from each custodian into estimated documents with a docs-per-GB lookup, total the data per source with SUMIFS and estimate the whole document population.
• Run the collected population through a culling funnel step by step – system files, duplicates, date range, search terms and family expansion – and measure how much of the collection reaches review.
• Turn document counts and review speeds into hours, count the working days to the deadline with NETWORKDAYS, size the first-level team with ROUNDUP and cost all three review tiers with SUMPRODUCT.
• Follow a review that works through batches in order of a relevance score: batch richness, cumulative relevant documents, the share of the estimated total found so far and the first batch that falls below a stop signal.
• Estimate how many relevant documents were left behind from a random sample of the discard set, turn that into recall and check how the margin of error moves the result.
• Measure how often second-level QC overturns each reviewer's calls, flag reviewers above a threshold, count the overturned "Not relevant" calls and colour every overturn in the log.
• Count the withheld documents, add up the produced pages and give every produced document its begin and end page number with a prefix and leading zeros.
• Forecast each phase's final cost from actual spend and progress, total the forecast variance with SUMPRODUCT, compute the cost per reviewed document and chart budget against actual spend.