The Training Budget Gate

The Training Budget Gate Scenario

Your training budget gets spent every year. Most of it buys activity, not performance. The difference is decided at the moment of approval, and most approvals are a formality. Someone senior asks, someone in L&D nods, the money moves.

The Training Budget Gate changes that moment. Seven criteria, each a yes or no. Every yes must be evidenced, not asserted. Pass all seven and the spend is approved. Fail any one and the decision is not “approved with concerns.” It is one of three: Fix the request, Defer it, or Stop it. Those are the only valid answers at a gate. This is the same discipline behind the Domino Map’s decision gates: leaders earn the right to proceed by making the conditions for success explicit, owned, and verifiable.

1. A named business outcome

The request must name the business outcome it serves, in plain language, with a person accountable for it. Not a learning objective. Not “improve leadership capability.” Something like: “Reduce regretted attrition among first-year engineers from 18 percent to 12 percent by Q3.”

The owner matters more than the wording. If no one is accountable for the outcome, no one will notice if the training fails to produce it. That is exactly when training becomes theater: everyone did their part, nothing changed, nobody’s fault.

Fail mode: the outcome is “develop our people.” Fix: send it back until a business owner names the outcome they will answer for.

2. A measurable gap with a baseline

You need two numbers: where performance sits today and where it must be. Without a baseline, no amount of spending can ever be proven worthwhile, because there is nothing to compare the result against.

This does not require a research department. It requires someone to write down the current state before the money is spent. Attrition rate. Error rate. Cycle time. Customer effort score. Pick the metric the business already trusts.

Fail mode: “We will measure after the program.” That is a promise to argue about the numbers later. Fix: establish the baseline first. It takes a week, not a quarter.

3. A diagnosed cause, with training as the matched intervention

Someone ran the diagnosis and the evidence says the gap is a skills gap. Not an incentive problem, not a broken tool, not a process nobody follows. Training is the matched intervention for the diagnosed cause.

This is where most requests die, and where most budgets are saved. Run the cause domains: clarity, tools, incentives, skills, manager reinforcement, context. If the cause is not capability, approving training spend is funding the wrong solution with confidence.

Fail mode: the diagnosis was skipped because the requester was senior. Seniority is not evidence. Fix: diagnose, then re-present.

4. Economics that clear the bar

The expected value must exceed the total cost, with both stated honestly. Total cost means everything: design and build, delivery, technology, plus the hours performers spend away from real work, plus maintenance. Most business cases count the vendor invoice and ignore the rest. The hidden costs are usually larger than the visible ones.

On the value side, use conservative math: the operational change you expect, times its unit economic value, times a realistic attribution factor. If the honest math does not clear the bar, the request is a donation, not an investment.

Fail mode: “The benefits are intangible.” Some are. Most are not; they are just unmeasured. Fix: quantify or defer until you can.

5. Transfer conditions in place

Even the right training fails without support, opportunity, and reinforcement. Will performers get help when they try the new behavior. Will they get real chances to practice it in actual work. Also, will managers notice and reinforce it.

Approving training spend without transfer conditions is buying a course and hoping the workplace cooperates. Hope is not a condition. If the conditions do not exist, the approval should fund fixing them first, not the course.

Fail mode: “Managers are on board.” Verbal support evaporates. Fix: name the specific manager actions, with dates, before approval.

6. A measurement plan with a decision attached

The request states what will be measured, when, and what decision the evidence will trigger. Not “we will evaluate.” Something like: “At 90 days, if the error rate has not moved from 6 percent toward 3 percent, we stop funding the rollout and revisit the diagnosis.”

Measurement without a pre-committed decision is decoration. The decision is what makes the measurement honest, because everyone knows in advance what the numbers will be used for.

Fail mode: evaluation planned for “after implementation,” owned by no one. Fix: write the decision rule into the approval.

7. A stop condition

Every approval names the condition that kills the spending. A date, a metric threshold, a trigger event. Open-ended funding is how zombie programs survive for years: nobody approved their continuation, nobody stopped them either.

This criterion protects the approver as much as the budget. When the stop condition is agreed up front, ending a failing program is executing the plan, not admitting a mistake.

Fail mode: “We will review annually.” Annual reviews review; they rarely stop. Fix: define the tripwire now.

A worked pass

A VP requests a $180,000 leadership program for 120 managers. Twelve workshops, external facilitator, offsite venue.

Criterion 1: the outcome is “stronger leadership bench.” No named business outcome, no accountable owner beyond the VP who requested it. Fail.

Criterion 2: no baseline. Engagement scores exist but nobody ties them to this program’s intent. Fail.

Criterion 3: no diagnosis. Exit interviews mention manager quality, but no one checked whether the cause is skills, incentives, or workload. Fail.

Criterion 4: $180,000 plus roughly 2,880 manager-hours off the floor, against unquantified benefit. Fail.

The verdict is Stop, not “approved with concerns.” The fixed version comes back in a month: the business owner is the COO, the outcome is regretted attrition among critical-role holders, the baseline is 14 percent, the diagnosis found a genuine feedback-skills gap in 40 percent of managers, the economics clear the bar at half the original cost, and the stop condition is set at 90 days. That version passes all seven. Same company, same budget pool, completely different spend.

Run it in writing

The gate takes one page. Each criterion gets a line: the evidence, the verdict, the owner. The approver signs the page. When a request fails, the page says exactly what would make it pass, so “Fix” is an instruction, not a rejection.

Most organizations do not have a spending problem. They have an approval problem. Seven criteria, honestly applied, fix the approval. Everything downstream gets easier: vendors sell against real outcomes, L&D builds against diagnosed causes, and finance sees a budget tied to performance instead of activity.

Put the gate in front of the next request that lands on your desk. If it passes all seven, approve it with confidence. If it does not, you just saved the money before it was spent, which is the only time saving it is easy.