--- title: "Every handoff leaks: why things fall through the cracks between your tools and teams" description: "Things fall through the cracks where work changes hands. What refineries, airlines and hospitals learned: name who owns the work next, and check that it reached them." publishedAt: 2026-09-28 updatedAt: 2026-09-28 author: Ena Pragma url: https://enapragma.co/field-notes/every-handoff-leaks-why-things-fall-through-the-cracks-between-your-tools-and-teams tags: ["erp-crm-handoffs", "how-to-think-in-systems"] --- *How to think in systems, episode 2.* ## Why do things fall through the cracks between my tools and teams? Because a handoff is where what one person knows has to reach the person who will do the work next, and it often reaches someone else, or arrives incomplete, without anyone noticing. A form or a rule does not fix that by itself: in the two investigations below, a logbook, a turnover form and, in one case, a written procedure all existed. What the investigations and the guidance point to is simpler and harder: name the person who owns the work after the handoff, and check that it reached them. The UK safety regulator's way of checking is a short two-way handover in which the person taking over cross-checks what they now own, with the few facts that matter written down as well. The rest of this note is the evidence, including where it is weak, and a version your team can try this week. ## A Friday order, pictured *This is an illustration, not a client story.* Picture a sales rep closing a custom order late on a Friday. The customer asked for delivery before 8 a.m. because the site gate is locked after that. The rep puts the order in the CRM and mentions the delivery window in an email to the customer. On Monday, someone in operations re-types the order into the ordering system. The delivery window was never in the order, only in the email. The truck arrives at ten. Nobody did anything careless. The rep knew, the order was entered, and the system shows no error. The information simply did not travel with the work. ## What the evidence says Almost all of the measured evidence here comes from hospitals, heavy industry and aviation, not from businesses like yours. It shows how handoffs fail and what helped in those places. It does not tell you how many of your problems are handoffs. We found no measurement of that for small or mid-sized businesses. **Communication breakdowns are common, and handoffs are part of them.** When a malpractice insurer's benchmarking arm [reviewed 23,658 US malpractice cases in its database](https://cdn2.hubspot.net/hubfs/217557/Documents%20-%20CBS%20Report%20PDFs/Malpractice%20Risks%20in%20Communication%20Failures%202015.pdf), "Communication was a factor in 30% of 23,658 cases," and more than half of those involved miscommunication between two or more providers. That counts every kind of communication failure, not handoffs alone. You will see much bigger figures quoted, such as the claim that 80% of serious medical errors involve handoffs. The earliest source we found for that one was a 2010 press release that cites no study, so we do not use it. **At Texas City, the talk that happened was with the wrong person.** In 2005 an explosion at the Texas City refinery killed 15 people. The [Chemical Safety Board](https://www.csb.gov/assets/1/20/csbfinalreportbp.pdf) found that "A crucial face-to-face discussion between the operator who began the process of starting up the raffinate unit, and the operator who finished the job did not occur at shift change." Two board operators did speak, but "because the Night Board Operator was not the one who filled the tower, he provided few details." The logbook entry was "minimal and unclear," and the day operator "read the logbook and interpreted the entry to mean that liquid was added only to the tower." The board found several causes, including operator fatigue and a high-level alarm that did not activate; the handoff was one of them. **On Continental 2574, the rules were sound and the handoff still missed.** In 1991 a Continental Express plane broke up in flight over Texas, and all 14 people aboard died. The airline's procedure already required conversations and paperwork: "These procedures included briefings by mechanics to supervisors, briefings by outgoing supervisors to incoming supervisors, completion of maintenance and inspection shift turnover forms." That night, the [National Transportation Safety Board](https://www.ntsb.gov/investigations/AccidentReports/Reports/AAR9204.pdf) found, a mechanic "gave a verbal turnover to the second shift supervisor (who was responsible for the C check on another airplane)." The turnover was passed along twice, ending with a mechanic who later "did not recall receiving a turnover." The incoming inspector read the turnover form and found nothing, because the outgoing inspector "had not yet made his log entries." The next shift's supervisor and mechanics "were not verbally informed." Upper screws had been removed from the leading edge of the tail, and it separated in flight. **Copying by reading is where data goes wrong.** A [review of 93 studies](https://pubmed.ncbi.nlm.nih.gov/38196643/) of clinical research data found that reading one record and typing what it says into another had a "pooled error rate of 6.57%," against 0.29% for typing straight from a form. These are pooled figures from different studies, from 1978 to 2008, not a head-to-head test, and the review is a preprint. But the pattern is familiar: plain typing was fairly accurate, and the step that involves reading and interpreting was not. **What helped, as far as anyone has measured, is mixed.** - In nine hospitals, a program for handoffs between resident doctors called I-PASS combined a standard format for spoken and written handoffs with training, observation and a campaign to keep it going. Afterward, [the medical-error rate fell](https://pubmed.ncbi.nlm.nih.gov/25372088/) "by 23% from the preintervention period to the postintervention period (24.5 vs. 18.8 per 100 admissions, P<0.001)," and the spoken handoff took no longer: "2.4 and 2.5 minutes per patient." It was a before-and-after study, not a randomized trial. "Site-level analyses showed significant error reductions at six of nine sites," so three did not show a significant drop. And because it was a bundle, the study cannot say which part did the work. - When one hospital [randomized teams](https://pubmed.ncbi.nlm.nih.gov/29672526/) to a bundle with training and a fixed time and place for face-to-face handoffs, more patients were handed over, but it "did not result in a significant improvement in patient safety." It was a small, short trial. - A review of SBAR, a widely used handoff format, found only "moderate evidence for improved patient safety," and said "there is a lack of high-quality research on this widely used communication tool." So the honest summary: structured handoffs changed behavior in these studies, the large before-and-after study saw fewer errors, the small randomized trial did not, and the evidence is thinner than the confident claims you will read elsewhere. **Saying it back helps, if someone listens.** In US air traffic control, pilots read instructions back to the controller. In a 1996 analysis of controller and pilot recordings, "Approximately 1% of the readbacks contained an error," and "Forty percent of these errors were not noticed by the controller or not corrected through further communication." A read-back is a check, not a guarantee. It is only as good as the listening on the other end. The UK's safety regulator describes the handover it wants in the same terms. A good [shift handover](https://www.hse.gov.uk/humanfactors/topics/shift-handover.htm) is "two-way, with both participants taking joint responsibility," "done using both verbal and written communication," and includes "cross-checking of information by in-coming personnel as they assume responsibility for the task." ## Try this for one week This is our advice, built from the evidence above, not a tested method. 1. **Pick one handoff that has hurt.** Sales to operations, operations to billing, one shift to the next. One, not all of them. 2. **Name two people.** The one who knows, and the one who will own the work next. Not "ops," and not whoever happens to be nearby. At Texas City and on Continental 2574, the conversations that did happen were with people who had not done the work or did not own it. 3. **The sender prepares three things.** What is done, what is not done, and anything unusual about this one, like the 8 a.m. gate. 4. **Talk, and have the receiver say it back.** Two minutes, both ways. The receiver repeats what they now own, and the sender listens for anything wrong or missing. 5. **Write down the few facts that must survive, in the system where the work happens.** Not in an email thread, and typed once. If someone is reading one system and re-typing into another, that is a good place to look first. 6. **Check that it happens.** Continental's rules were sound, and they were not followed. For the first few weeks, look at a handful of handoffs yourself. I-PASS paired its format with training and observation, too. ## Where AI fits AI can take over some of the copying in a handoff: moving an order's details from one system to another, or flagging an order that is missing a field it usually has. Summarizing an email thread is different. That is reading and interpreting, the step that went wrong most in the data-entry review, so an AI summary needs the same say-back check as a person's. We wrote about a similar problem between AI agents in [When AI agents talk to each other, structure gets lost. New research points to where it starts.](https://enapragma.co/field-notes/when-ai-agents-talk-to-each-other-the-loss-starts-at-the-writer) There too, most of the loss started with whoever wrote the handoff, and the cheapest check was a round trip. Episode 1, [You are the bottleneck](https://enapragma.co/field-notes/you-are-the-bottleneck-how-to-make-your-business-less-dependent-on-you), was about decisions that wait on one person. A handoff is what those decisions travel through once you let them go. *How this was researched:* our AI research agent searched for measured evidence on handoffs and wrote a graded evidence sheet. We then re-read the sources quoted here directly on September 25, 2026: the PubMed abstracts of the I-PASS, handoff-trial, SBAR and data-entry studies, the Chemical Safety Board's Texas City report, the NTSB's Continental Express report (from a library copy with searchable text, because the NTSB file is a scan), the insurer's malpractice report and the UK regulator's guidance. The air traffic control quotes come from our research agent's full-text copy of the FAA-sponsored report, because the government site blocked our own download. A separate critic agent checked this note's quotes and claims against those sources before it was published. The Friday order is an illustration, not a client. ## Sources - [Starmer et al., "Changes in medical errors after implementation of a handoff program," New England Journal of Medicine, 2014 (PubMed abstract)](https://pubmed.ncbi.nlm.nih.gov/25372088/) - [CRICO Strategies, "Malpractice Risks in Communication Failures," 2015 Annual Benchmarking Report](https://cdn2.hubspot.net/hubfs/217557/Documents%20-%20CBS%20Report%20PDFs/Malpractice%20Risks%20in%20Communication%20Failures%202015.pdf) - [US Chemical Safety Board, "Refinery Explosion and Fire," BP Texas City, final report, 2007](https://www.csb.gov/assets/1/20/csbfinalreportbp.pdf) - [National Transportation Safety Board, Aircraft Accident Report NTSB/AAR-92/04, Continental Express Flight 2574, 1992](https://www.ntsb.gov/investigations/AccidentReports/Reports/AAR9204.pdf) - [Garza et al., "Error Rates of Data Processing Methods in Clinical Research: A Systematic Review and Meta-Analysis of Manuscripts Identified Through PubMed," preprint (PubMed abstract)](https://pubmed.ncbi.nlm.nih.gov/38196643/) - [Tam et al., "Structured patient handoff on an internal medicine ward: A cluster randomized control trial," PLoS One, 2018 (PubMed abstract)](https://pubmed.ncbi.nlm.nih.gov/29672526/) - [Müller et al., "Impact of the communication and patient hand-off tool SBAR on patient safety: a systematic review," BMJ Open, 2018 (PubMed abstract)](https://pubmed.ncbi.nlm.nih.gov/30139905/) - [Cardosi et al., "An Analysis of TRACON (Terminal Radar Approach Control) Controller-Pilot Voice Communications," DOT/FAA/AR-96/66, 1996](https://rosap.ntl.bts.gov/view/dot/8711) - [UK Health and Safety Executive, "Shift handover"](https://www.hse.gov.uk/humanfactors/topics/shift-handover.htm)