# Every team's numbers are right. The money is lost between them.


Each system records its own side of a decision. The loss appears only when records from two teams sit on one row and someone owns the answer.

> By Saad Bin Shafiq, Founder of Nodes · Nov 9, 2026
> Canonical: https://www.nodes.inc/blog/money-lost-between-teams

**Answer:** Hidden costs between departments stay hidden because each system records only its own side of a decision. Purchasing books the lower price, support books the extra work, and finance sees two unrelated lines. The loss appears when records from both teams are joined to the same decision and one person owns the answer.

---
Your purchasing system is right. Your support system is right. Your general ledger is right. All three can be right while the company loses money on one decision.

A team moves a part to a cheaper supplier. Purchasing records the saving, and the saving is real: the unit price fell and the invoices prove it. Four months later, warranty claims climb and support adds a shift. Support records that too, in a different system, under a different cost center, with no field that names the supplier change.

Finance closes the quarter with a favorable price variance on one line and an overrun on another. Nobody lied. Nobody made a mistake inside their own job. No report in the company has both facts on one row.

The example is made up. The pattern behind it has been measured for decades.

## Why the loss never reaches a report

Reports follow the org chart because budgets do. A cost center has one owner and one number to defend. A decision that helps one owner and hurts another has no natural home, so it gets recorded twice, as two facts with no relationship between them.

Integration projects do not fix this on their own. When MuleSoft surveyed 1,050 IT leaders in late 2022 for its [Connectivity Benchmark](https://www.salesforce.com/news/stories/connectivity-report-2023/), the average organization ran 1,061 applications and had integrated 29% of them. Those figures are a few years old and will have moved. The harder point has held up: most of the integration that exists was built to complete a process. An order reaches the warehouse. An invoice reaches the ledger. A pipe built to finish a process has no reason to ask whether the process was worth running.

## Accountants found this decades ago

Robert Kaplan spent a long part of his career on one version of the problem. In a [2005 article for Harvard Business School](https://www.library.hbs.edu/working-knowledge/a-balanced-scorecard-approach-to-measure-customer-profitability), he describes an insurance company that ranked its customers by profit after assigning its indirect costs to each one. The top 40% produced 130% of annual profits. The middle 55% broke even. The bottom 5% lost an amount equal to 30% of annual profits.

Kaplan says that shape appears in "virtually every customer profitability study ever done," with 15% to 20% of customers typically producing all of the profit or more.

Look at what made the losses visible. Those customers looked fine in the sales numbers, because their revenue was real. The cost of serving them sat in marketing, distribution, technical, service and administrative lines, which many companies treat as fixed or spread as a flat percentage of sales. Put the two together per customer, and part of the book is being paid for by the rest of it.

Then look at why so few companies kept doing it. Kaplan told [Working Knowledge in 2007](https://www.library.hbs.edu/working-knowledge/adding-time-to-activity-based-costing) that the employee time surveys behind activity-based costing were expensive and slow, so models were refreshed rarely, and many companies abandoned them or let them go stale. The join was worth doing. Keeping it current cost more than most finance teams could spend.

That cost is what has changed. Reading records across a dozen systems and proposing which ones belong to the same decision used to be a consulting project. Software can now do the first pass, and people can spend their time on the part that needs them.

## Where to look first

Four places are worth checking before anything exotic.

### Bills paid twice

Accounts payable checks each invoice against its own purchase order. A second vendor record for the same supplier passes that check. [APQC's benchmark](https://www.cfo.com/news/metric-of-the-month-detect-and-prevent-duplicate-or-erroneous-payments/), published in 2020, put duplicate or erroneous payments at about 0.8% of annual disbursements for top performers and about 2% for the bottom tier. Not all of that stays lost, since some gets caught and recovered. Each percentage point is still a large number at enterprise scale, and the vendor master, the invoice queue and the bank file each look clean alone.

### Software nobody uses

Finance holds the contract and the renewal date. The identity provider holds the log-ins. Zylo's [2026 SaaS Management Index](https://zylo.com/blog/how-much-wasted-on-saas-spend) says the average organization uses 54% of its licenses, and puts the waste at $19.8M a year on average and $80.6M at organizations above 10,000 employees. Zylo sells software for this problem and its summary does not describe the sample, so read those as a vendor's count. The renewal still gets approved by someone who cannot see the log-ins.

### Work never billed

The project system knows what was delivered. The billing system knows what was invoiced. The contract, often a PDF in a shared folder, knows what was allowed. A change order that was approved in an email can fall through all three, and each system will report that nothing is wrong.

### A customer about to leave

The account owner sees a renewal date a few months out and a good relationship. The product team sees log-ins falling every week. Support sees the same complaint for the third time. Any one of those is noise. Together they are a warning you can act on while there is still time.

## Why a dashboard does not close the gap

The usual answer is a warehouse and a cross-functional dashboard. Keep them. They answer questions someone already thought to ask. The supplier row at the top of this piece exists only if somebody suspected the supplier change, knew which two systems to join, and had the access to do it.

Finding the gap is also half the work. A finding with no owner stays a slide. The buyer who booked the saving has no reason to reopen it, and the head of support has no authority over suppliers. Somebody has to see both sides, put an estimate on each and take it to a person who can decide. We wrote separately about [the cost of acting versus doing nothing](/blog/cost-of-inaction-is-the-hard-half), which is the comparison that person needs.

This is the design we chose for Nodes. It connects the systems a company chooses and looks for places where one team's choice costs another team money. It suggests a fix and shows the evidence from both sides. A person says yes or no and gives a reason, and [an approval gate](/blog/what-is-an-approval-gate) means Nodes changes only what was approved. Those answers stay with the company in its [context graph](/blog/context-graph-vs-retrieval-pipeline), so the next finding starts from what the last decision taught.

## What a join cannot tell you

Two facts on one row are a coincidence until someone tests them. Warranty claims can climb for reasons that have nothing to do with the supplier: a design change, a bad batch from a different plant, a new support policy that made claims easier to file.

A finding worth taking seriously shows its estimate, its assumptions and what would prove it wrong. Potential value is a number to test. It becomes money after the change is made and the result is measured over a period both teams agreed to in advance. If the result comes back flat, that belongs in the history too.

## One test for this quarter

Pick the largest saving your company booked last year. Name the two teams most likely to have absorbed its cost. Ask for their records from the two quarters after it took effect, matched to the decision.

If that takes an afternoon, you are in good shape. If it takes a project, you have found your first gap. [The CFO guide](/buyers/cfo) covers what to ask next.

Every team in your company can close the quarter on plan while the company misses its number. Nobody at that table is wrong. They are each holding half.

## Sources

- [Robert S. Kaplan, A Balanced Scorecard Approach to Measure Customer Profitability (HBS Working Knowledge, 2005)](https://www.library.hbs.edu/working-knowledge/a-balanced-scorecard-approach-to-measure-customer-profitability)
- [Susan Gilbert, Adding Time to Activity-Based Costing (HBS Working Knowledge, 2007)](https://www.library.hbs.edu/working-knowledge/adding-time-to-activity-based-costing)
- [MuleSoft, Connectivity Benchmark Report coverage (Salesforce News, 2023)](https://www.salesforce.com/news/stories/connectivity-report-2023/)
- [Perry D. Wiggins (APQC), Detect and Prevent Duplicate or Erroneous Payments (CFO, 2020)](https://www.cfo.com/news/metric-of-the-month-detect-and-prevent-duplicate-or-erroneous-payments/)
- [Zylo, 2026 SaaS Management Index summary](https://zylo.com/blog/how-much-wasted-on-saas-spend)

*Saad Bin Shafiq is the founder of Nodes, serving data-sensitive enterprises.*