In-house governance and a primary platform/provider shrink variance, tighten forecasts, and cut invoice surprises

eDiscovery is expensive—and unpredictable. And that variance wrecks budgets. The fix isn’t cheaper eDiscovery; it’s controlled eDiscovery cost forecasting through an operating model that forces cost-driving decisions upfront. The operating model works if you (1) know your typical eDiscovery spend for that matter archetype and apply that baseline to new matters; and (2) manage overruns using two signals—scope delta (approved changes to inputs like custodians, date ranges, or sources) and execution drift (unapproved drift like rework, throughput failures, or QC churn).

Practically, it works like this: at each matter’s outset, establish your baseline scope and phase budget forecast tied to that scope. I’ve dubbed this Gate 1. Then, watch for the two signals: scope delta and execution drift. Either signal triggers an enforcement point—Gate 2. Gate 2 is the change-control checkpoint: you classify the variance as scope delta or execution drift, then either re-price the scope or intervene operationally. This operating system—baselining, Gate 1, the two signals, Gate 2—works most elegantly within a single system of record, often anchored in a primary platform/provider. You can build this in-house, but most teams anchor it in a primary platform/provider because otherwise normalization becomes a permanent job.

• Gate 1: Baseline scope + phase forecast
• Track two variance metrics: (1) scope delta and (2) execution drift
• Gate 2: Change control—re-price scope delta or intervene for execution drift
• System of record: capture baseline, deltas, phase forecasts/actuals, approvals

True, some variance will remain no matter what any in-house team does. But many overruns come from preventable drift (i.e., execution drift): inconsistent scoping, late protocol changes, and rework that only surfaces on an invoice—after an invisible string of decisions. The goal isn’t perfection. It’s reduced variance, higher predictability, and clear explanations when spend moves. These outcomes follow if you, the in-house team, own the cost-driving decisions and enforce guardrails. For if you outsource the decisions, you outsource control. But put the right decisions behind gates, and discovery becomes explainable and forecastable even when matters are genuinely different.

That’s the frame. Now the mechanics.

Benchmark eDiscovery Spend by Matters First—or Your Baseline Is Built on Vibes

Mechanics start with standardization. Let me explain: eDiscovery spend variance stems from teams running each matter like bespoke project work—new tools, new lawyers, new workflows. And true, each matter is unique, but that doesn’t mean every matter warrants a custom-built workflow, for litigation has a repeatable operational skeleton, even when the facts change. The phases recur: early case assessment, collection, processing, review, production, and case support. Lather, rinse, repeat. Scale and emphasis differ, but the workflow is consistent enough to standardize.

Here’s my argument: if an in-house team benchmarks eDiscovery spend by phase, the data will reveal a typical range per phase. If the range is tight, congratulations—you can forecast. But if it isn’t, that’s telling. It tells you what levers to pull to tighten things up and make spend forecastable. The biggest lever is governance: in-house ownership of scope decisions, change control, and the operating record. Benchmarking is the prerequisite.  

Yet benchmarking only works when cohorts of similar matters match the cot drivers. Without comparable cohorts, the numbers lie. The difference in review spend in a slip-and-fall versus a class action is astronomical—and, as a metric for this exercise, entirely useless. A truer assessment requires two filters:

• Matter archetype. Group matters by posture that indicates eDiscovery investment (for example: single-plaintiff employment, commercial contract disputes, regulatory inquiries, internal investigations, wage-and-hour/class actions).
• Discovery complexity bands. Within each archetype, sort matters into a few operational bands that actually drive cost: custodian count, data volume, data types (email/files only vs. chat/mobile vs. cloud/structured), and review model (linear vs. TAR/CAL, onshore/offshore mix).

If two matters diverge in more than one band, don’t benchmark them together. Either separate the cohort or normalize the phase using unit costs like $/GB processed, $/GB hosted/month, and $/doc reviewed. Otherwise, you’re averaging noise.

The end game is understanding typical spend per phase by archetype and complexity so budgets start with data, not vibes—lest you find yourself worshipping the false god of a vibes-based budget. Because a data-backed one is how you establish your baseline scope.

Gate 1: Lock In Baseline Scope Before Work Begins

Gate 1 requires a baseline scope and a phase forecast priced to that scope. Benchmarking gives you a baseline scope for cohorts of matters. And now, when your outside counsel gives you a budget, you have actual-spend data to negotiate with. Baseline scope is the input; the budget is the priced output. In other words, your eDiscovery budget should reflect the scoped inputs—not assumptions or optimism.

For example, say you benchmark by comparing all doc review costs for single-plaintiff slip-and-falls over the last four years and realize you’ve spent $10,000–$17,000 on 98% of your matters (excluding true outliers). That’s your baseline scope band for that cohort—and it should anchor your phase budget. If your outside counsel budgets $25,000 for doc review on that type of matter, you can address it upfront—ideally before the work begins—and retool the budget to fit historical data.

Now, in an active matter in that same cohort, you’re only halfway through review but you’ve already spent $12,000. If nothing changes, you’re on track to blow past the band.

And that’s the point. Armed with historical and current spend data, you can change things mid-workflow. Invoice variance—or better yet, active monitoring—flags an issue. Next: attribution. Did scope change, or did execution drift?

Track Two Cost Drivers: Scope Delta vs. Execution Drift

Say you ask your outside counsel what spiked the eDiscovery costs. They may tell you that everything has changed—the plaintiff added eight other defendants and seventeen claims; discovery has ballooned beyond the normal range. That’s scope delta. Or they may say they had to re-review half the dataset because they changed the review protocol midstream and invalidated 25% of prior issue tagging. That’s execution drift—unapproved cost driven by rework, workflow drift, or preventable inefficiency. Probably the most maddening thing to see on an invoice. And if you’ve managed litigation budgets, I bet you know the feeling all too well.  

Either way, the overrun should map to one or both metrics: scope delta or execution drift.

• Did you authorize a change in inputs (custodians, sources, dates, protocol tier, deadlines)? If yes, it’s scope delta.
• If you didn't authorize a change to inputs, it’s execution drift (rework, protocol changes, throughput issues, or QC churn).

Once you categorize the root cause, you’re at Gate 2.

Gate 2: Price Scope Delta or Intervene on Execution Drift

At Gate 2, the in-house owner classifies the variance and approves the response; outside counsel and vendors supply the pricing and operational plan. Gate 2 is your enforcement mechanism. And the management logic is direct: price scope at decision time; arrest execution drift before it snowballs. Here’s what I mean. If scope creeps—and it will—price it. Adding 10 GB of custodian data? That’s a function of dollars per gigabyte. Adding 13 custodians’ phones? That’s a function of dollars per phone. These things aren’t nebulous; they’re concrete and priceable. And as in-house counsel, you can (and should) ask your vendors—law firms and eDiscovery providers alike—to price them so you face fewer surprises.

Now, if execution drifts, that’s a different ailment with a different fix. Execution drift often stems from the operating model—the thing you, as in-house counsel, can control—and springs from inconsistent scoping, workflow drift, late protocol changes, and change control (or lack thereof) that manifests as invoice surprise. Left unmanaged, it hides until the invoice like a bedeviling Jack-in-the-box.

That’s the framework, but what of enforcement? To give this gate real teeth, in-house teams need to own the eDiscovery operating system: specifically, the decision rights when scope changes or execution drifts. This doesn’t require a new department or a complicated process. It just requires one owner (person or team) with authority to (1) set a baseline at Gate 1 or (2) pause work at Gate 2. If that owner can’t pause the vendor queue or freeze the review workflow, the gates are all bark, no bite.

Of course, total discovery spend still moves with data volume, deadlines, and posture. The point isn’t to pretend every matter costs the same. The point is that owning the decision rights shrinks the unexplained part of cost swings and turns eDiscovery from invoice roulette into something your team can plan around. That’s governance. And that’s the end game of mature eDiscovery governance: decision rights, documentation, and enforcement tied directly to spend.

But governance fails without documentation. If your baseline scope lives in email, your re-baselines live in Slack, and your invoices live in PDF purgatory, you don’t have governance; you have folklore. So the next step isn’t another meeting. It’s a single system of record that pins each dollar to an approved decision.

One System of Record: A Force-Multiplier for Spend Data

A "system of record" is where you store the operating truth of discovery: baseline scope, re-baselines, phase forecasts, phase actuals, and the approvals that connect spend to decisions. That linkage turns variance from unexplained to governable—every dollar ties back to an approved choice.

But the system only works if the data is consistent. You must compare apples to apples: consistent fields, consistent taxonomy, consistent phase definitions across vendors and matters. You could cobble together vendor and firm data, normalize it manually, and maintain it for every matter. But that's a permanent data-cleaning job your lean in-house team probably can't absorb. The scalable option is to anchor your program to a primary platform/provider that normalizes inputs, enforces phase taxonomy, and ties invoices to scope decisions automatically. This approach aligns with broader corporate eDiscovery strategy that emphasizes eDiscovery vendor management as a strategic competency rather than an administrative afterthought.

So let’s say you opt for a primary platform/provider model. Your system of record should capture—and require—these fields:

• Matter segmentation: archetype (e.g., employment single-plaintiff, commercial contract, regulatory inquiry) and operational bands (custodians, volume, data types, languages/geo, timeline, and review model)
• Baseline scope: custodians, sources, date range, protocol tier, and review model—plus explicit assumptions
• Phase forecast: budget by phase using a consistent phase taxonomy
• Phase actuals: actual spend by phase, tracked in the same taxonomy
• Variance attribution: scope delta vs execution drift, with a short note
• Re-baseline approvals: timestamp, approver, and what changed

This isn’t a wish list. It’s the minimum data you need to forecast. For when those fields are mandatory, three things happen. First, matter segmentation transforms your “apples-to-apples cohorts” from theoretical to actual. Second, baseline scope and phase forecast make enforcing Gate 2 eminently doable. They take your kickoff budget from vibes-based to data-backed. And third, phase actuals, variance attribution, and re-baseline approvals mean invoice friction is less of an argument about fairness; it’s more of a check against the operating record. In other words: fewer fights, earlier decisions.

The payoff isn’t just cleaner reporting. It’s leverage. Because once you separate scope delta from execution drift, AFAs stop collapsing at the first change order.

AFAs: The Compounding Payoff  

Standard inputs change the dynamic. If you can forecast by phase and separate scope deltas from execution drift, you can structure alternative fee arrangements that survive reality: a set baseline, clear assumptions (volume, protocol tier, languages, deadlines), and an agreed method for pricing deltas when scope changes. That’s what aligns incentives. Counsel can run the case without managing client expectations through invoice surprises, and the department can re-price change without renegotiating the matter from scratch.

The structure may look like this: fixed fees by phase for baseline scope, unit pricing for scope deltas ($/GB prodcessed, $/GB hosted/month, $/doc reviewed), and operational remedies for execution drift.

Now, not every matter belongs in a fixed fee. Some matters are inherently unstable—expedited timelines, cross-border constraints, volatile privilege density, unusual data types, bet-the-company litigation. The point isn’t to force AFAs everywhere. It’s to make the same kind of change trigger the same kind of pricing conversation early, with documentation that holds up later.

Three Steps to Start Forecasting eDiscovery Costs

You don’t need a multi-year transformation. You need a minimum viable operating model that makes the two metrics move. You’re building a repeatable decision loop. Start with three things:

• Build the minimum data schema: baseline scope, phase forecast, phase actuals, and a scope-vs.-execution attribution when variances appear.
• Enforce two gates immediately: baseline scope (Gate 1) and re-baselines (Gate 2).
• Run a monthly variance review that asks only two questions: How far off were we by phase? And was it scope delta or execution drift?

Then judge success with three outcomes that finance will recognize:

• Invoice variance tightens by phase (forecasts become credible, not aspirational).
• Scope changes get priced earlier (re-baselines happen when decisions are made, not when invoices arrive).
• Execution drift falls because preventable rework gets caught early instead of cascading.

These outcomes can’t be faked. Either forecasts tighten and variances become explainable in real time, or they don’t. When they do, you’ve turned discovery from bespoke scramble into a managed operating function—without pretending complex matters stop being complex.

Discovery won’t stop being complex, but it can stop being unpredictable. Most in-house teams know what needs to happen. The challenge is building it into the operating rhythm without adding permanent overhead.

If you're ready to make eDiscovery costs forecastable, TransPerfect Legal can help. Our discovery consultants work with in-house teams to implement the baseline scoping, two-gate enforcement, and system-of-record infrastructure that make eDiscovery budgets and variance explainable in real time. Whether you need phase tracking or strategic guidance on vendor management and change control, we've built these models for legal departments managing complex, multi-matter portfolios. Connect with us to discuss how baseline-driven budgeting can work in your program.