Budgeting, Forecasting and Cost Management Case Studies
Finance5 Case StudiesUpdated Sep 2026
Budget work rarely fails on arithmetic. It fails when the people who own a number disagree about what it is for, or when nobody compares the forecast with what actually happened. The five cases below come from finance teams of different sizes, from a regional parts distributor to a clinic network. Each begins with a visible symptom: a missed target, an angry department head, or a process that keeps costing more than planned. The aim is to show how roles, service levels, workflow data, stakeholder expectations, and delivery targets interact when money is tight and information is partial.
Case Study 1
Missed invoice targets in a shared services team
Quillon Industrial Supply, a parts distributor with six regional depots and about 1,350 employees
The situation
Quillon runs a shared accounts payable team of eleven people that processes invoices for three divisions. Its targets are simple: approve 95 percent of invoices within five business days and finish the month-end payables cut by the third business day. Over two quarters the approval rate sat near 71 percent, and the cut slipped into the sixth business day most months. Ownership is blurred. Intake clerks log invoices, divisional managers approve them, procurement issues the purchase orders, and the controller signs off on accruals. Each group ranks its own priority first.
The challenge
Leadership has started to treat the team as the problem, and two divisional managers have asked to pull their invoices in-house. The team lead, Priya, believes the team is doing what it was asked to do, and she has the logs to show intake is current. The real issue is where the time goes and who is supposed to act at each stage. Any fix must work within current headcount, because the budget has no room for additional hires.
Analysis
Start with the workflow data, because the team's own timestamps tell a different story from the complaints. Pulling the approval queue showed that invoices sat with divisional approvers for an average of six business days, while intake took under one. The team was measured on approval within five days, so the clock started when an invoice landed in a manager's queue and kept running while the manager was away. Teams grounded in Budgeting, Forecasting and Cost Management usually separate the stages of a process before they assign blame, and that separation here showed three distinct failure points: approver absence, purchase orders that did not match invoices, and accrual entries that were reopened after close.
The second layer is expectations. Divisional managers believed the five day clock was a promise to them, but the policy was written as a promise to suppliers, which is a different standard. Nobody had agreed whether the target measured the team or the approvers. Procurement, meanwhile, had never been asked how often purchase orders were amended after the invoice arrived, and that amendment rate turned out to be the largest single source of rework. A target that does not name an owner for each stage invites every group to point at another.
Actions taken
- Split the five day target into two service levels: intake within one business day, approval within three days from the divisional queue, with the clock paused only for documented vendor disputes.
- Name one accountable approver and one backup for each divisional queue, and publish their names on the shared dashboard so overdue items route to a person rather than a general inbox.
- Require procurement to confirm purchase order amendments before invoices enter the queue, and track the amendment rate weekly as a separate measure so rework has a visible owner.
- Hold a fifteen minute Monday review with the controller, procurement lead and team lead, using one page of queue age and amendments, and escalate any stage that misses its service level two weeks running.
Results
Within the first full quarter, the approval rate rose from 71 percent to 88 percent, and the month-end cut moved to the fourth business day. Divisional managers stopped asking to bring invoices in-house once they could see their own queue ages, and the amendment rate fell by roughly a third, which removed most of the rework. The team still does not hit the 95 percent target, and the controller has said the accrual error rate needs its own plan next. The most useful change was smaller than expected: a named backup for each queue cleared the overdue items that had sat for weeks.
Questions for discussion
- Which of the three stage level targets would you tie to individual performance reviews, and which should stay with the team as a shared measure?
- How would you handle a divisional manager who refuses to name a backup approver because the workload is already unmanageable?
- If the amendment rate is driven by procurement behavior, what evidence would you need before asking procurement to change its own process?
Case Study 2
Escalation over late variance reporting at a clinic network
Brenmoor Clinical Group, an outpatient network with nine sites and about 2,100 staff
The situation
Brenmoor Clinical Group runs nine outpatient sites under one finance office of six people. The regional operations director, Marcus, relies on the monthly variance report to decide where to hold staffing and where to release overtime budget. Over the last quarter, three reports arrived after the ninth business day, the cost center labels changed without notice, and two sites were shown with the wrong headcount. Marcus sent a written complaint to the chief executive, copied to the board's audit member, and asked for a meeting within the week. The finance manager, Dana, has been covering the reporting alone since a colleague moved to payroll.
The challenge
The complaint is partly fair, and the response has to separate what was late and wrong from what was merely unclear to Marcus. He expects a report that is accurate, on time, and explained in terms of his sites. Dana can produce an accurate report by the sixth business day if she stops reconciling cost centers by hand, but that requires a mapping decision the controller has not made. Marcus also wants a commitment in writing before the meeting, which means the team must promise dates it has not yet tested.
Analysis
The first step was to pull the reporting workflow data rather than rely on memory of the complaint. The log showed that data pulls were completed on time in most months, but cost center mapping took four to six days and was done by one person. The service level in the finance charter said reports would be issued by the sixth business day, yet no one had been assigned the mapping step, so it had no owner and no deadline. The late reports were a symptom of an unowned step, not of a slow analyst. Practitioners in this area tend to ask whether a delay sits in the work itself or in the handoff before they judge anyone's speed.
Marcus's expectations were never written down. In conversation he wanted variance by site, explained against approved headcount, with forecast impact for the next two months. The team had been delivering a total variance by cost category, which satisfied the charter but not the way he managed. Rebuilding trust meant agreeing a short specification with him, then committing to measurable targets: issue by the sixth business day, reconcile every site before issue, and flag any unexplained variance above a threshold to him directly. A commitment is only credible when the stakeholder can check it against a date.
- Agree the spec Confirm with Marcus which sites, categories and forecast horizon the report must cover.
- Stabilize the calendar Give the mapping step a named analyst and a fixed completion day.
- Report and explain Issue the variance by site with a short explanation for every line above threshold.
- Review and recommit Meet monthly to check the last three dates, changing targets only by agreement.
Actions taken
- Call Marcus before the meeting to acknowledge the three late reports and the wrong headcount, and ask him to name the two decisions the report must support this quarter.
- Assign the cost center mapping to a named analyst, move it ahead of the data pull, and set a fixed completion day so the sixth business day target is realistic.
- Agree a one page specification with Marcus covering site level variance, approved headcount, and a two month forecast, then confirm it in writing before the next report is produced.
- Send a short written commitment with three measures: issue by the sixth business day, every site reconciled before release, and any unexplained variance above the agreed threshold raised within one day.
Results
The next two reports reached Marcus on the fifth and sixth business days, both reconciled at site level with a short explanation for each line above the threshold. He used the first one to move overtime budget from a site that had recovered into one under strain, which he had not been able to do before. Marcus still asked for a third measure, forecast accuracy over the following two months, and the team agreed to report it from the next cycle. The escalation closed by the end of the quarter, but the mapping step still rests on one analyst.
Questions for discussion
- What would you have said to Marcus in the first five minutes if the late reports had been caused by a system outage rather than an unowned step?
- How should a finance team decide which stakeholder requests become written commitments and which stay as informal preferences?
- When a commitment depends on a single person, what backup arrangement would you accept, and how would you check that it works?
Case Study 3
Recurring rework in change order approvals
Sorrell and Vance Engineering, a civil contractor with four project offices and about 1,100 employees
The situation
Sorrell and Vance manages construction projects across four offices. Every change to a signed contract, called a change order, must be priced, approved, and loaded into the cost ledger before the project forecast can be updated. The cycle runs monthly. Site engineers submit change requests, the estimator prices them, the project manager approves, and the cost accountant posts them. Over the last two quarters, the average change order took twenty three business days to post, and about one in five came back from the accountant with a missing quantity or a wrong cost code.
The challenge
Project managers say forecasts are stale for most of each month, which makes margin conversations with clients uncomfortable. Estimators say they cannot price requests that arrive without quantities. Accountants say the rework comes from the estimators. Nobody has measured where the time actually goes. The company wants a fix that does not lengthen the approval chain, because the approval step exists in response to an audit finding and cannot simply be removed.
Analysis
Mapping the handoffs produced a clearer picture than the argument did. Across the sampled change orders, most of the elapsed time was spent in queues rather than in work. A request waited about six business days before the estimator opened it, and another four days sat with the project manager. Active pricing took under a day in most cases. The data also showed the rework loop. Roughly a fifth of requests returned to the site engineer at least once, most often because the quantity field was empty. Practitioners who work in cost control usually count returns and waiting time before they propose any change to the steps themselves.
The root causes sat at the handoffs. The request form did not require quantities, so estimators guessed or sent it back. Project managers approved in batches at the end of the week, because the approval service level was written as five business days and nobody had defined when a request counted as received. Accountants posted only after approval was signed, which meant an error found late restarted the loop. Three options were on the table: tighten the form with mandatory fields, add a fixed weekly approval slot, or move pricing closer to the site engineer with a template. Each was compared on cycle time, error rate, and effort on people already overloaded.
Actions taken
- Add mandatory quantity and cost code fields to the change request, and reject incomplete submissions at intake rather than sending them back after pricing has started.
- Use a fixed Wednesday approval slot, and define receipt as the moment a request lands in the project manager's queue so the five day service level has a start point.
- Pilot site pricing templates on one project office for one quarter, measuring pricing time and return rate against the current process before any system change is funded.
- Cut the accountant's post-approval review to a sample of changes above a set value, since the loop is driven by upstream data quality and checks at the end catch errors too late.
Results
Within two quarters, median posting time fell from twenty three business days to under twelve, and blank-field returns dropped sharply. Project managers see updated forecasts within the week, and the client conversations that had been postponed are running on current numbers. The site pricing pilot showed a real gain in one office but needed a second quarter before the company was willing to fund system work. The Wednesday slot frustrated project managers who preferred daily approvals, and exceptions are still handled case by case.
Questions for discussion
- Which of the three options would you fund first if the only resource available was the time of the estimator who currently prices every request?
- How would you measure whether a control removed at the end of the workflow was actually replaced by upstream checks?
- What evidence would convince a skeptical project manager that a fixed approval slot is faster than approving items as they arrive?
Case Study 4
Resistance to a new spend approval policy
Hollis Ridge Credit Union, a member-owned lender with twelve branches and about 640 staff
The situation
Hollis Ridge Credit Union operates twelve branches and a central operations office. The finance committee approved a new spend policy: discretionary purchases above a set value need pre-approval through a purchasing tool that replaces email requests. The policy takes effect at the start of the next quarter. Branch managers had been approving their own small purchases for years, and the tool adds a step to purchases that used to take a day. Two weeks before launch, a survey showed that most branch managers expected delays, and several had started routing purchases through personal cards to avoid the queue.
The challenge
The rollout lead, Tomas from the finance office, has three weeks until go-live and a single trainer. The policy is not optional, and the committee wants uptake above ninety percent in the first month. The concern is not that people reject the purpose of the policy. Branch managers say they were not consulted on the threshold, that the tool does not handle emergency purchases, and that they are still measured on the same service levels as before.
Analysis
Before sequencing anything, the team sorted the complaints by who raised them and what each group actually lost. Branch managers worried about speed and autonomy. Central operations worried about approval volume and the audit trail. Frontline staff worried about being blamed for purchases they could not pre-approve. The tool's own logs, from a pilot in two branches, showed that approval took under a day once a request was complete, but nearly a third of requests were incomplete when first submitted. Teams that run policy rollouts usually separate a stated objection from the underlying loss before they design a response, because the same complaint can hide very different problems.
The adoption sequence followed the losses rather than the calendar. First, the threshold was revised for branch emergency purchases after a short consultation, which answered the autonomy concern at low cost. Second, the team added an emergency path with after-the-fact review, so managers were not forced to choose between the policy and a broken refrigerator. Third, training went to branch managers before frontline staff, since they would be asked to answer questions. Uptake was tracked weekly as the share of discretionary purchases routed through the tool, against a target that rose by a fixed step each week rather than jumping to ninety percent on day one.
Actions taken
- Revise the threshold for branch purchases after a two week consultation with branch managers, publishing the reasons for each figure so the change reads as a decision made with them.
- Add an emergency purchase route with after the fact review within two business days, so that no branch is forced to choose between the policy and a critical repair.
- Train branch managers in the first week, ahead of frontline staff, and give them a short script for common questions so they can answer without routing every query to the finance office.
- Report uptake weekly as the share of discretionary purchases routed through the tool, with a target that rises by a fixed step each week, and review any branch that falls below its line.
Results
Uptake reached 62 percent in the first week and 84 percent by the end of the month, short of the ninety percent target but well ahead of the 40 percent the pilot branches had shown at a comparable point. Two branches still route most purchases through personal cards, and the team has asked their managers to explain the pattern rather than assuming defiance. Branch managers now describe the emergency route as the reason they accepted the policy. The finance office has asked for a second trainer before the next policy change, since the first month showed how quickly the single inbox fills up.
Questions for discussion
- Which stakeholder concern in the readiness chart would you address first, and what would you give up to do it?
- If uptake plateaus at the same level for three weeks, what would you examine before deciding whether to tighten the policy or loosen the target?
- How would you decide whether a training script is helping branch managers or quietly shifting questions to a different team?
Case Study 5
Choosing among three capital initiatives with partial data
Ferris Bay Utilities, a regional water utility serving about 38,000 customer accounts
The situation
Ferris Bay Utilities has a capital budget that covers one major project in the coming cycle. Three proposals are on the table: pipe replacement in older districts, an upgrade to a pumping station, and a replacement billing system. Each has a cost estimate, but the estimates differ in reliability. The pipe program depends on a condition survey that covers only a third of the network. The pumping figures come from a vendor quote with a short validity period. The billing benefits rest on a vendor demonstration. The finance director, Ines, is preparing the board paper.
The challenge
The board wants one recommendation, and staff are divided along departmental lines. Operations argues for the pumps, because a failure there would stop supply to a district. Customer service argues for billing, because complaints about estimated bills drive most of the call volume. Asset management prefers pipes, because the condition data points to the highest failure rate. The paper must justify a choice, state what would change it, and avoid committing the full budget before the data is better.
Analysis
The team started by agreeing criteria before scoring any proposal, because scoring first tends to pick a winner and then build the criteria around it. Four criteria were set: service risk avoided, cost certainty, reversibility, and effect on customer accounts. Each initiative was then assessed on how wrong its inputs could be. The pipe program had the weakest data but the most reversible structure, since districts could be sequenced and paused. The pumping upgrade had a credible failure consequence but a quote that would expire before the board's decision. Experienced analysts usually weight risk by how quickly a wrong estimate can be corrected, not only by how large it might be.
The recommendation was staged rather than binary. The board would commit now to design of the pumping station upgrade, because the failure consequence was severe, the quote was time limited, and design could proceed inside the same budget envelope. Pipe replacement would receive a funded condition survey in the first stage, with a go or no go decision once the survey covers the full network. Billing replacement would be deferred with a written trigger: a vendor's confirmed cost and a measured benefit from a reference site. Service levels for each stage were stated, including the response time for the survey and a monthly variance report to the board.
- 1Set criteria firstAgree four criteria with the board before any initiative is scored or ranked.
- 2Test input reliabilityCheck how wrong each estimate could be and how quickly it could be corrected.
- 3Commit the pumping upgradeFund design now, since failure consequences are severe and the quote expires.
- 4Fund the pipe surveyPay for a full network condition survey, then decide on replacement districts.
- 5Defer billing with triggersSet a written trigger: confirmed vendor cost and measured benefit from a reference site.
Actions taken
- Write the four criteria and their weights into the board paper before any scoring, and invite the department heads to challenge the weights at a short workshop.
- Ask the asset team to state the confidence range for each cost estimate and the point by which the range would narrow, so the board can see which numbers are provisional.
- Commit only the design budget for the pumping station, and set a service level for the design stage with a fixed review point at the next board cycle.
- Attach a written trigger to each deferred item, naming the evidence required, the owner who will bring it back, and the variance threshold that would reopen the decision early.
Results
The board approved design for the pumping station and funded the pipe condition survey, and it deferred billing with a trigger tied to the reference site review. The full survey later identified two districts needing replacement sooner than planned, a gap a single large pipe commitment would have hidden. The vendor quote expired before design was complete, so the team re-priced the work, and the variance report showed the change openly. Estimated bill complaints continued, and customer service has asked for interim meter reads as the next item to test.
Questions for discussion
- If the vendor had offered a longer quote validity period, how would the staged recommendation have changed, and what would you have needed to see?
- Which criterion in the paper carried the most weight in practice, and how would you test whether the result depended on that weighting?
- When a department head argues that a deferred item is more urgent than a funded one, what evidence would make you reopen the decision?
Learning Resources
Finance CASE STUDIES
China, India, United States, Indonesia, Pakistan, Nigeria, Brazil, Bangladesh, Russia, Japan, Mexico, Philippines, Vietnam, Egypt, Turkey, Germany, United Kingdom, France, Italy, Canada, South Korea, Australia, Spain.