ApiaryActiveLive
Try: pause · settings · learn · wipe
← Community / Reading Room
WA
craft · 24 min read

Write a Service Handoff That Survives a Different Technician

The technician returning tomorrow should not need to call the technician who left today just to find out which room is affected. Yet that is what a work order…

The technician returning tomorrow should not need to call the technician who left today just to find out which room is affected. Yet that is what a work order labeled “handled, waiting on part” often requires. Was something fixed, or did someone just talk to the customer? Is one piece of equipment back in use while another remains untouched? Does the office know who will buy the part? Half of the work can be done correctly and still leave tomorrow's crew arguing about the other half.

A useful handoff starts with a simple obligation: another person has to be able to look at it cold and understand which actions are supported by evidence. That doesn't take a large software rollout. It takes a record that separates evidence from interpretation and gives every open item a next move.

The examples below are fictional, limited to ordinary noncritical fixtures, and are about documentation rather than a repair method. Safety-critical equipment follows its own procedures, documentation, qualifications, formal notices and access controls. A work-order note is not a substitute for any of them.

A service visit, a work order and an issue can all end at different times

A visit ends when the person leaves or no more work is scheduled for that visit. An order can remain active after the technician leaves. An issue can remain open on an active order even while another issue has been verified resolved. Combining all three into a single “complete” pill gives you an attractive board and a poor answer.

Suppose a venue asks a shop to correct two problems with a countertop literature stand assembly. One paper tray does not sit square against its guides. A second has a cracked cosmetic trim kit and needs replacement. The first can be corrected and checked against the venue's requested outcome today. The second remains untouched while the correct kit is identified. The office cannot honestly tell the venue both were completed; it also shouldn't schedule another half-hour to repeat the work that passed today.

Give the visit its own end time and the issues their own statuses. That makes it possible to distinguish "visit ended, issue open" from "visit ended, both repairs accepted." It also makes repeat calls easier to interpret. If a customer calls again about the first tray, review the evidence from its prior test. If they call about the second, that's a request for an update, not proof that a previous repair failed.

Large facilities already separate these concerns. The Whole Building Design Guide's commissioning documentation overview describes how issue logs preserve an issue's status and responsible party alongside other verification and operation records. The VA/USACE commissioning specification, Section 01 91 00.01, also distinguishes corrective action, readiness for retest, and who verifies resolution. The U.S. Department of Energy’s federal commissioning process likewise identifies retesting after implementation and hands off a record plus a plan for future work. A small service shop doesn't need to adopt those specifications. It can borrow the useful idea: make correction and verification separate facts.

Build around three clocks without asking the technician to track everything

Capture the first clock as the visit window: who was present and when that visit began and ended. Capture the second as an issue's evidence sequence: report, observation, corrective work, verification. The third is the customer or office commitment, such as “tell the venue what the parts desk found by Wednesday.” It may start after the visit ends and fire before the issue has changed. One timestamp is rarely enough to distinguish all three. An app can expose separate fields, but a numbered paper record can do the job if the habits are clear.

Keep each time near the event it qualifies. “14:20 recording entered, observation made around 13:30 local” tells a dispatcher why a fresh update refers to an older state. Timestamp granularity should match the job. A description like “before the lobby opened” is sometimes critical to a reported condition, while the last seconds of a closeout message are usually incidental. If someone's device clocks differ or the site uses a different time zone from dispatch, write the intended clock context in the appointment. Never change an actual observation time simply to make a daily report sort more neatly.

Treat reminders and visit notices as a second view onto commitments, not their only existence. A supervisor might send an email promising Friday and then discover that a colleague captured Thursday in the route queue. Record who said what to whom, resolve the mismatch, and correct the scheduling record and affected notice deliberately. Do not choose whichever date is written in the newest interface without checking what was actually agreed. If the mistake was never sent to a customer, avoid sending an unnecessary apology that introduces a new confusion. Mark the corrected internal state and keep a trace of the change if the old value was already acted on.

One uncomplicated status summary can say: “Visit closed 11:40. Issue remains open on model suffix confirmation. Coordinator update due Wednesday by 15:00; no second visit scheduled.” Each sentence is about a different clock. The summary still works when a job spans shifts or when a service desk stays open longer than a retail venue. It lets the next person understand a late report without assuming somebody had stayed onsite at the time the entry was created.

Use observations that a second person can reproduce

Compare these two notes:

“Fixture acting weird. Fixed.”
“At the reception counter, left freestanding brochure stand, asset S-07. Empty upper tray sat visibly out of square. Visitor reported that paper caught at the left front corner. Technician observed the catching with one stack of 20 standard brochures supplied by reception. Lower tray not checked.”

The second note is not a more technical diagnosis. It is a better record of what someone saw. It includes an asset, a location, the claim's origin, a reproducible input, and a boundary. It is usable even if a future investigation discovers a different explanation. The first note may come from the same technician doing the very same work, but anyone who reads it must invent most of those facts.

The best way to get better observations is to stop asking for “more detail” in the abstract. Ask “Which thing?”, “What precisely happened?”, “Under what conditions?”, and “Who witnessed it?” A receptionist may describe a sound that happens only at opening. A tech may see the symptom only after pulling out a full tray. A manager may be looking at the wrong fixture. Store all three as claims with their sources and times. There is no requirement that one of them become a root cause before the visit ends.

A clear photo helps when it answers a particular question. A wide photo can establish the bay and fixture. A detail photo can show the tray guide and asset label. A photograph of a nameplate can establish which model number was visible, provided it is readable and belongs to the item pictured. A photo by itself cannot usually establish what happened while operating the fixture or what was agreed on a call. Name the piece of evidence by its job: “S07 location,” “S07 label,” “S07 upper tray before.” Leave irrelevant people, computer screens, paperwork and keys out of the frame.

Include the mismatch without assigning a cause

People standing at different angles can see different things in good faith. A desk employee facing patrons might call a drawer the one on the left, while the person at reception sees it on the right. Two displays in the same room might have asset tags that differ by one digit. A phone recording of “left stand” should not force the next person to guess an orientation. Describe the shared landmark and point of view. If the report names a location you haven't verified, record it as the location reported, not the location found. Put both in the same issue record so a future technician can see why an initial route might be wrong.

The same distinction matters when the event repeats in a new form. “Pamphlet slipped below the bottom lip when inserted diagonally” is not identical to “paper stopped moving at the front left.” The next person might relate the observations through actual checks, but the record should not make that inference automatically. A useful contradiction entry has both observations, their conditions, and a specific next comparison. If the responsible lead decides they belong to separate issues later, keep each observation connected to its original source and leave a pointer from the first combined record.

A shop's reporting tools may prompt for one category before saving. Use the narrowest supported category, then keep uncertain symptoms as descriptive text rather than choosing a cause just to complete the form. “Operation symptom, cause undetermined” may be enough. If the system offers only three wrong choices, update the form rather than teaching people to conceal the gap. These constraints matter because office reports often summarize only category fields. A claim made to discover which parts to bring can become a false trend by the time somebody counts ten instances as manufacturing defects.

Small examples should also respect permissions. If the employee tells you the issue looks worst in a staff-only area, ask the responsible contact to coordinate access or bring out a representative sample within policy. Don't treat a line in the job record as permission to enter a new area, and don't include passcodes, access links, or private diagrams in broadly distributed updates. The handoff can say “requires receiving-area appointment; site coordinator will arrange,” which gives the next person enough information without carrying credentials along with the job.

Set a pass condition before writing that it passed

A check should answer the customer’s agreed problem under described conditions. For a noncritical display using the example above, you might agree to check that a stack of the venue's standard brochures can be inserted and removed without catching in the upper tray. The test is modest because the promise is modest. It does not prove the tray can carry every kind of booklet or survive years of use.

Putting the test after the action matters. “Aligned existing parts” is an action. “Removed and replaced the 20-brochure stack three times; no catches observed” is an observation under recovery conditions. “Verified for 20 standard brochures used today” is a bounded status. Preserve all three rather than writing “working like new.” If someone later uses thick stitched catalogs instead of the original thin brochures, that change can be compared to the earlier baseline without misrepresenting either result.

Record where and when the check took place and which person did it. Independent verification may be required for some jobs; if it is not, say who checked rather than implying an independent observer. A customer watching the same test can be noted accurately. A customer who was merely told it passed is not a witness to the test. A signature approving receipt of a work order isn't necessarily acceptance of every issue on it. Keep customer acknowledgments in the field that matches what they actually acknowledged.

If the relevant condition cannot be checked, leave the result open. An operating problem may depend on something the crew cannot reproduce in the shop. An open item like “not reproduced with a 20-piece stack; customer reports only oversized programs catch; waiting for a sample” gives a very different next step than “resolved, no issue.” The point of a closeout is not to make every box green before the end of the shift.

State what would turn this test into a new test

Every comparison has a few variables worth preserving. The product may be different, the fixture may sit in a new spot, the usual insert may have become stiffer, or normal operation may be delayed when a customer flow starts. A later technician should see which of those changes would require a new test, without being asked to preserve a distracting inventory of everything in the lobby. Capture what could realistically affect the outcome and leave administrative details elsewhere.

For the standard-brochure example, “three cycles using stock supplied today” is a fair record. Add the stock name or a simple measurement if it matters to the next decision and is reasonable to capture. “Stock L, 20 thin folded brochures, ordinary straight-in motion from customer side” gives later work a useful reference. It says why a thick catalog or diagonal motion may not be covered. There is no need to generate confidence scores or claim a prediction about every paper variation. A short condition statement beats a precise but imaginary percentage.

If the proposed remedy can be performed but cannot be checked under the triggering conditions, document the split in the actual status, not just in a footnote. “Corrective action completed, awaiting test with usual material” is available for the next person to read during scheduling. In a low-risk shop there may be a sensible alternate check, such as confirming that the unchanged adjacent tray still accepts that exact stack. That could isolate a mistaken report, but it would not prove the affected tray is repaired. Give surrogate evidence its own scope. If the job has an applicable final testing requirement, a surrogate does not replace it unless that arrangement is actually approved.

Preserve honest negative evidence. A repeated test with unchanged conditions may find no observable recurrence. That is not equivalent to evidence that the original report was false or that a correction caused the absence. Set a practical rule for when to reopen, who observes, and how they can record an example if it happens again. Avoid asking casual users to follow invasive monitoring or keep customer data when their ordinary description or a benign sample would do. A clear next signal is more useful than a generic promise to “keep an eye on it.”

Make an open item an actual assignment

Every unresolved issue should name a next actor, next action, and timing or trigger. “Waiting on parts” is a state, not an assignment. A bounded note might instead say: “Parts desk to confirm kit against manufacturer’s parts sheet by Wednesday noon. Do not order until manufacturer confirms the model suffix. Scheduler updates venue by Wednesday 3 p.m. even if vendor hasn't replied.” Neither instruction makes up an arrival date.

Split the dependencies that have different owners. The parts desk and scheduler can each see one action. When no one responds from the vendor, the next move is an update, not disappearance. If the quote is required before work can continue, write “quote required, not approved” and assign a person to prepare it. An exploratory phone call is not approval to order parts. The record should be as clear about what has not been authorized as it is about the action that has.

Use a status vocabulary small enough that everyone can remember it:

  • Reported: received but not yet reproduced or independently confirmed.
  • Observed: a technician or clearly identified other source witnessed the symptom and recorded conditions.
  • In progress: authorized work is underway.
  • Blocked: a named decision, part, access window, or condition prevents progress.
  • Ready for retest: the corrective action is recorded; the relevant test isn't complete.
  • Verified resolved: the applicable acceptance check passed, with its results and scope recorded.
  • Closed without resolution: no further work is authorized for this order, with the reason and any remaining issue explicitly recorded.

A service desk may use different words. The discipline is not the exact labels; it's avoiding status jumps that hide facts. “Technician left” cannot transform blocked into verified resolved. “Parts promised” cannot transform quote required into authorized. An office should be able to see the difference without playing back a voice message.

Treat quiet progress and silence as different things

Once an issue has an owner, write an explicit trigger the owner can observe. “When received” can mean a tracking email, a parcel sitting in receiving, or a physically matched kit. If the trip depends on the correct item, “shipping notice received” is merely progress. “Parts desk confirms exact kit identifier on the physical packaging and records it on A-218-02” names the needed check. The next status can become “ready to offer a return window” only if the other dependencies, including approval and access, are met. This distinction makes reminders precise instead of forcing the coordinator to ask everyone whether anything changed.

A responsibility assignment must survive a calendar collision. A shop with one dedicated parts person may want a backup who can at least prepare an update when the assigned person is away. The backup doesn't inherit authority that was never granted. They can say “part identification is pending; next update Thursday” without ordering anything if that is within role. If there is no person responsible for a handoff period, record the uncovered gap. On small jobs this may lead to a different check time or a narrower promise rather than silent assumptions about availability.

Record the last known evidence as early as practical, not only while creating a progress reminder. If the vendor sent a general question and nothing else, write “vendor asked for model suffix; no match or delivery estimate received.” It tells the person responsible what an answer would look like. A bare “waiting” can conceal whether the shop is awaiting a supplier, a supplier is awaiting the shop, or both think the other side will act. On an ordinary Friday queue, the actor's next step is sometimes the most important thing to resolve. No amount of rewriting the status adjective can substitute for it.

Worked example: two findings, one return decision

Here is how the countertop stand job might appear in a readable record. The fictional model and dates are deliberately generic. No real job or venue is implied.

Order A-218 | Visit 1 ended Tuesday 11:40 | Two distinct items

A-218-01 – left stand upper tray catches stock Location: reception counter, left side facing the desk. Asset S-07. Reception reports catching at left front corner. Technician reproduced with 20 of the standard brochures supplied by reception. Lower tray, other paper formats, and long-duration use not assessed. Action recorded: corrected seating of existing tray and guide per applicable fixture instructions. Verification after action: stock inserted and removed three times without catching with the same 20 brochures; employee at desk watched final cycle. Status: verified resolved for the original reported condition with standard stock. Evidence: S07 location/photo 01; observed and verification note 02.

A-218-02 – right stand trim cracked Location: reception counter, right side facing the desk. Asset S-08. Hairline crack visible on decorative trim at corner, as shown in photo 03. Staff report cosmetic issue only; no additional mechanism or use change assessed as part of that observation. No work on S-08 performed. Site contact asked for a quote, no order approval recorded. Status: blocked on exact manufacturer kit / quote. Parts desk obtains model suffix and matches kit listing by Wednesday noon. Scheduler sends a status update by Wednesday 3 p.m., even if no vendor ETA. Return visit only after availability and scope approval are confirmed.

Customer-facing summary “The left stand's upper tray passed a test with the same brochures that were catching. The right stand's trim was not repaired today. We're identifying the matching trim kit and will update you by Wednesday afternoon; no part has been ordered.”

There is no mystery appointment in that note. The customer doesn't have to infer why only one stand changed. The next technician does not need to poke around a different asset looking for a cosmetic crack. The office can create one reminder and know which part of the note to share. Evidence stays attached to the specific issue instead of a generic folder called “Tuesday.”

A useful exercise is to have one person read the handoff without seeing the pictures and name the two affected assets, test conditions and next assignment. Then hand them the photos and ask which one proves the model suffix. If the answer to the second question is “none,” write “not visible” and schedule a better capture only if necessary. This catches an important class of errors: notes that are complete grammatically but still send the next person to the site without a decision.

Follow a materially different case through one additional visit

Imagine an independent print collective, not the venue in A-218. At the back counter, a shallow paper rack occasionally catches a wax-coated sample sheet during withdrawal. When an employee tries to show the problem, only standard uncoated stock is available. The technician observes no catching with that stock and records the affected rack's identity from its shop tag. The lead tells the employee, “The rack passed with ordinary paper. We couldn't reproduce your reported condition without the coated stock. Please set aside one reference piece before the next review.” No corrective work was attempted. This is a conditional finding, not a failed fix and not a resolved issue.

Two days later the office receives a low-quality video with a visible scrape but an unreadable asset tag. The note says “same issue.” This raises priority to gather evidence without changing status to confirmed. A short call through the normal site contact establishes that the video is from a different rack in the same room. Create a separate report ID for that recording; retain the original issue and explicitly mark the clip nonmatching. A shortcut that simply appends it to the older case might get a prompt second visit but send the worker to the wrong place. A short correction now preserves the capacity to group both records later if an actual shared condition is found.

At the second visit, the supplier's ordinary paper stock arrives along with one coated sample that the employee recognized. The technician can reproduce the original pattern with the coated stock by the agreed method and records that it does not occur on the supported ordinary paper. Suppose the operator's documentation specifies ordinary sheets; in this invented case the venue decides not to pursue an alteration to accommodate the coated sample. That is closure of this work order by a scoped decision. It is not a finding that the coated stock works or that all use is approved. Send the actual result plus the closure reason, not “nothing wrong.” If the program changes later and does need coated material, a new request can refer to this verified limitation without redoing the first two visits.

This case differs from a repair with a clear retest. The return proved a condition, not a successful fix. Compare its data shape to the earlier job: one reported behavior, a negative test under alternate conditions, rejected evidence from a different fixture, positive reproduction under the intended stock, then a customer decision not to alter. A template that has only “failed/open” and “fixed/closed” would misstate at least one step. The extra middle states earn their keep where a workshop has to explain after the fact why a visit was useful without a repair.

Keep the customer view separate from internal assignment

A site contact needs outcomes, limits, known next steps and any project-specific instructions that apply. The internal queue may also include clearance constraints, staffing notes, purchase proposals, supplier contacts and link permissions. A single shared document that mixes them will be either too sparse for the crew or too revealing for everyone else. Generate the customer-facing summary from the confirmed fields, not by forwarding the entire job card.

Record any restrictions as required by the relevant safety process and directly communicate them to the responsible people. A private note visible only to tomorrow's technician does not notify today's users. A customer-facing report alone does not execute any physical restriction. For serious issues neither a broken status pill nor a customer email substitutes for mandated actions. The example here deliberately avoids those hazards so the handoff design is the lesson.

Retain originals where appropriate and follow your organization's privacy and retention policies. A three-photo record that answers three questions is often better than an unreviewed gallery. Better yet, make sure the person checking the evidence can open the links with their actual access. A link that only works for its uploader is a poor handoff even if the picture itself is excellent. On equipment records kept over several visits, add a new verification entry rather than silently replacing the original. You then have a useful sequence rather than a polished statement detached from what really happened.

Correct records without turning them into a moving target

Even bona fide measurements are sometimes typed into the wrong issue row. If a record was already used for a routing, purchase or customer update, treat later correction as a new event. Preserve the old assertion in the appropriate record history and identify what replaces it. Then update the people who took action from the wrong claim, using the organization's rules. “The 20-sheet test was on S-07, not S-08; S-08 was not tested” is more useful than silently swapping a photo label. It lets the recipient undo any specific conclusion that depended on the incorrect asset number.

Minimize what can go wrong at rollout. The work cards may be generated through one system but displayed in another. Have an authorized person try viewing a sample entry, photo, and attachment from the actual role that needs them. Verify that the role can see an unchecked or blocked label, not just green success. A missing field in the external summary need not be a defect if it deliberately protects internal information. It becomes a defect when it conceals the only remaining help needed from the audience. Two views can share identifiers without disclosing the same full contents.

Suppose a team wants to export the record later for training. Use fictional or stripped-down examples rather than forwarding actual contact details, site photos or private communications. Keep relevant evidence in its approved place, and preserve originally reported language when needed for interpretation without treating it as a diagnosis. This protects the record from a common confusion: a classroom illustration can show how a good handoff works, but can't prove who said what in the shop. Training and recordkeeping are distinct tasks even when they borrow the same fields.

Start with one template and a return-visit review

For the next ten ordinary service visits, try this compact structure:

  1. Identify: order, visit, date, asset ID, location and issue ID.
  2. Observe: customer report and field observation labeled separately, plus test conditions and evidence links.
  3. Act: what was done and what was explicitly not done.
  4. Verify: check method, result, scope, date and person; mark unchecked items.
  5. Assign: status, owner, next step, due time or next trigger, and what the customer was actually told.

Change the template when a field repeatedly fails to answer a real question. If techs skip the location because the customer address is already in the system, show them that the missing location is “reception right,” not a second postal address. If the same issue picks up a fresh ticket number every visit, keep the work-order numbers but use a durable issue ID across them. Make it easier to fill the useful field than to write a vague paragraph somewhere else.

At the end of the pilot, review three return visits and one job that never needed a revisit. Ask whether the record correctly identified the affected asset, whether an unchecked condition got expanded into a broader promise, and whether an unresolved item had a named owner. You can also count the handoff clarification calls the team had to make, but don't assume fewer calls prove better repairs. Some calls are good, particularly where new facts matter. The target is fewer calls to recover facts already observed at the first visit.

If using AI to clean up dictation, give it the existing fields and require it to leave blanks visible. It can shorten the wording or pull a timestamp out of narration. It must not infer that a check happened because the tech said a part was installed, or convert “should get it tomorrow” into a promise. Review any generated customer wording against the original observations and approvals before sending it. Avoid uploading photos, customer details or confidential documentation to an unapproved service.

The best closeout isn't the longest note or the one with the fewest open tasks. It's the one that lets the next person find the right asset, repeat the right check and know exactly who acts next.

Hear the next shift read the customer update aloud

Put a customer update next to its source record and ask someone to explain every verb. “Ordered” names an actual purchase, not a stock inquiry. “Scheduled” explains that someone established a time, not that a technician felt available then. “Sent for review” tells the recipient only who is expected to look next. A defensible update can sound less busy than the internal work that produced it, and that is fine. The recipient doesn't need a performance of certainty; they need one clear, accurate next expectation.

There is useful information in the expected negative case. A message due at three can be sent even if a vendor failed to respond by two. In a simulated handoff, have the person write the update for that case using only the last confirmed facts and the existing authority. The output should neither imply an order was placed nor shift the burden onto the recipient unnecessarily. That small rehearsal tests both the visible record and whether the promised checkpoint is genuinely deliverable.

Try a handoff the receiver cannot rescue by calling the writer

For a short pilot, ask a relief supervisor to review a fictitious set of three entries when the writer is not present. One has a completed repair and a physical verification. One has a completed action but no valid conditions for a test. One was observed and needs a proposal, but no corrective work has approval. The supervisor should be able to assign each to a next owner and answer a customer status question without speaking to the writer. For the unverified one, ask what fact would make it ready rather than what reassuring words make it sound ready. Make the accepted answers and known limitations part of the exercise.

Then introduce trouble with the display layer. Pretend the dispatch dashboard has grouped any row written today under “closed today” because it uses the work order status, not the issue status. The pilot should reveal that misread. Correct the view or label its scope rather than giving every dispatcher a list of secret exceptions. If the tool can't distinguish the clocks introduced above, manually track open outcomes in an appropriate queue for now. A deadline or platform upgrade may eventually remove the duplication, but the number shown today should not overclaim what happened.

Measure interruption recovery as well as volume. One useful ratio is how many sampled open records have a next action that can be executed by the named person with no preventable clarification. Keep counts and examples separate; three missing model suffixes can often be fixed by one field prompt, while one missing set of retest conditions might call for a bigger design change. Avoid creating incentives for closing difficult items early. A high closure rate can be earned by dropping hard problems. A useful record keeps the trouble visible at the right level until an authorized decision really ends it.

Sources and scope

AI disclosure: Drafted with AI assistance for Apiary. This workflow is an editorial adaptation, not a claim of compliance with cited commissioning specifications. The fixture work, people, schedule, observations and test records are invented for illustration.

Frequently asked
What is Write a Service Handoff That Survives a Different Technician about?
The technician returning tomorrow should not need to call the technician who left today just to find out which room is affected. Yet that is what a work order…
What should you know about a service visit, a work order and an issue can all end at different times?
A visit ends when the person leaves or no more work is scheduled for that visit. An order can remain active after the technician leaves. An issue can remain open on an active order even while another issue has been verified resolved. Combining all three into a single “complete” pill gives you an attractive board and…
What should you know about build around three clocks without asking the technician to track everything?
Capture the first clock as the visit window: who was present and when that visit began and ended. Capture the second as an issue's evidence sequence: report, observation, corrective work, verification. The third is the customer or office commitment, such as “tell the venue what the parts desk found by Wednesday.” It…
What should you know about include the mismatch without assigning a cause?
People standing at different angles can see different things in good faith. A desk employee facing patrons might call a drawer the one on the left, while the person at reception sees it on the right. Two displays in the same room might have asset tags that differ by one digit. A phone recording of “left stand” should…
What should you know about set a pass condition before writing that it passed?
A check should answer the customer’s agreed problem under described conditions. For a noncritical display using the example above, you might agree to check that a stack of the venue's standard brochures can be inserted and removed without catching in the upper tray. The test is modest because the promise is modest.…
References & sources
  1. Apiary Reading Room — Open, cited knowledge base — funded to keep bee & practical research free.
From the Apiary Reading Room. Opinion & editorial — not financial advice. We don't overclaim.
More from the Reading Room