Telemetry Is Not Truth

Sync recommends. A named human decides. Execution stays with authorized systems. On the Decision Case and in Recovery, telemetry is a stream of observations. Truth for the decision is whether those readings, under their actual sensing, sampling, calibration, and context limits, justify the action or the clearance at hand. Those are different records. A dense, high-frequency, green-looking feed can still be wrong for the decision: wrong sensor, wrong unit, stale sample, aliased signal, or a proxy that does not track the failure mode. Telemetry is not truth. It is not a diagnosis. It is not root cause. It is not authorization for corrective work. It is not a verified operational outcome. It is not a Decision Case. Treating telemetry as truth is false clearance. Verification stays open until a verified operational outcome is recorded from the plant, not from a green trace. The honesty boundary and the verification boundary still apply to the feed.

In this essay, telemetry names those five records and no figure of speech. Tags are the named points a system stores: a pressure, a temperature, a speed, a state bit, a counter. Sensors are the devices that produce those points, each with a range, a location, and a way they can be wrong. Historians are the stores that keep the points over time, at the rate and the compression the historian was given. SCADA is the picture and the live list operators watch: screens, trends, and alarms built on those points. CMMS counters are the tallies a maintenance system increments: hours, starts, cycles, completed tasks. A stream of observations is that flow. It can be dense. It can be high-frequency. It can look green. Truth, in this essay, is not a mood and not a slogan. Truth for a plant decision is whether those readings, under the sensing, sampling, calibration, and context limits that actually apply, justify the action or the clearance the room is about to record. A feed that has not been held to those limits is not that justification.

The meeting puts the feed on the screen. The tags are updating. The sensors are green. The historian trace is dense. The SCADA overview has no red tile. The CMMS counters are moving. Someone says the asset is known, and the room treats the stream as truth for the decision: run on, defer, or close. Nobody named the sensing. Nobody named the sampling. Nobody named the calibration. Nobody named the context. Nobody asked whether the point was the wrong sensor, the wrong unit, a stale sample, an aliased signal, or a proxy that does not track the failure mode. Nobody wrote accept, reject, escalate, or return. The Decision Case still has to show the question, the evidence, the recommendation, the named human decision, and the check. A telemetry feed is not that record, and it is not truth.

This essay follows Accuracy Is Not Precision and stands beside Calibration Is Not Validation, Simulation Is Not Proof, Forecast Is Not Fact, Model Is Not Reality, Proxy Is Not Outcome, Verification Is Not Optional, and Learning Requires a Verified Outcome. Accuracy keeps repeatability — tight residuals, fine resolution, and stable decimals — from being stored as truth relative to the plant decision and observed outcomes. This essay keeps the stream itself — tags, sensors, historians, SCADA, CMMS counters — from being stored as truth for the action or the clearance. A precise feed can still be the wrong sensor, the wrong unit, a stale sample, an aliased signal, or a proxy that does not track the failure mode. Stable decimals on that feed are still not accuracy, and the feed is still not truth. Calibration keeps a model, a twin, a sensor, or a forecast tuned so historical residuals look small from being stored as validation that the representation is fit for the decision. A sensor that was calibrated is still a sensor. A small residual on that sensor is still not validation, and the feed is still not truth. Simulation keeps a controlled representation of possible futures from being stored as proof the plant will behave that way. Forecast keeps a forward-looking compression from being stored as a fact about the plant. Model keeps a compressed representation from being stored as the plant. Proxy keeps a stand-in, including a tag that does not track the failure mode, from being stored as the verified operational outcome. Verification keeps the case open until the observed outcome is on the case. Learning keeps a later question from inheriting telemetry clearance. A dense feed is still not truth.

Orville Davis states the Decision Case order in Field Manual v0: Question, Evidence, Recommendation, Human decision, Action, Verification, and Learning. The manuals index lives at /manuals. Evidence is step 02. The Evidence chapter keeps a visible split between observed fact, assumption, and what is still missing. A stream of tags, sensors, historians, SCADA, and CMMS counters is not a fourth column that records the decision, names the failure mode, authorizes the work, or closes the case. A proposal whose only basis is the feed has not shown that the readings, under their sensing, sampling, calibration, and context limits, justify the action. A proposal with no named human decision is not a Decision Case. Honesty boundaries sit beside the spine: recommend is not authorize, evidence is required, and Stage-1 does not treat a telemetry feed as live plant control. Sync refuses false precision. A number the historian printed is not truth the edition did not hold.

A stream of observations is not the decision

A stream keeps the points it was given and drops the points it was not. Tags keep the names that were configured and drop the names that were never mapped. Sensors keep the physics they can see and drop the physics they cannot. Historians keep the samples they stored and drop the intervals they compressed away. SCADA keeps the screen that was drawn and drops the equipment that was not on the screen. CMMS counters keep the increments someone recorded and drop the work that was never written. Density is not a fault in the tool. High frequency is not a fault in the tool. A green-looking overview is not a fault in the display. Each of those is a limit of the record. The limit does not become truth because the feed looks finished.

Sensing is what the device can actually observe: where it sits, what it touches, what it cannot see, and which failure mode it was never built to track. Sampling is how often the point is taken, how it is filtered, and what the historian throws away between stored values. An aliased signal is a sampling failure: the stored wave looks smooth, or looks like a slower wave, because the sample rate did not hold the motion that mattered. Calibration is whether the point still sits on a known reference, and when that reference was last shown. Context is the regime, the unit, the asset, the operating state, and the question the reading is being asked to answer. A reading without those four is an observation with the limit hidden. Hidden limits are how false clearance gets stored.

The feed can still be wrong for the decision in ordinary ways. The wrong sensor is the point from a different asset, a different leg, or a spare that was never in the flow. The wrong unit is a number that looks familiar in a scale the decision does not use. A stale sample is a value that stopped updating while the screen still paints it as live. An aliased signal is a trace that cannot show the cycle the failure mode lives in. A proxy that does not track the failure mode is a tag that moves, or stays green, for a reason that is not the mode the room is deciding. A dense, high-frequency, green-looking telemetry feed can be any of those and still fill the wall. Filling the wall is not justification.

A feed can inform a recommendation to investigate when sensing, sampling, calibration, and context limits are explicit, and when the case can still refuse if the readings do not justify the action. Without that refusal, the green trace is a claim with the limit hidden. Refusal is the recorded stop. The honesty boundary refuses to call a practice feed a live conclusion. The verification boundary refuses to call a feed a check. Unknown stays unknown. The operator must not treat a missing refusal as go.

Treating telemetry as truth is false clearance

False clearance is a record that says go, defer, or closed when the check is not on the case. The tags are in band, so the room stores in-band as ready. The sensors are green, so green is stored as healthy. The historian is dense, so density is stored as knowledge. The SCADA overview is quiet, so quiet is stored as controlled. The CMMS counters are incrementing, so the counters are stored as the work. Each of those moves is false clearance. Treating telemetry as truth is another path to false clearance under the honesty boundary and the verification boundary. The case looks released. The decision was not justified by the readings under their limits. The feed was treated as if it were the justification.

False precision is the same failure in a finer unit. A tag printed to a decimal the sensing does not support, a historian value printed as if compression had not dropped the peak, a SCADA number printed in the wrong unit as if the scale were a measurement of the mode, a CMMS counter printed as if the tally were the outcome — those are false precision. Sync refuses false precision. This essay states no OEM limit, no sample rate, no residual tolerance, no remaining-life figure, no probability, and no savings figure. A number the meeting wants so the feed will look decisive is not a diagnosis, and this page does not publish one. The case holds the approved evidence that was stored, the sensing that was named, the sampling that was named, the calibration that was named, the context that was named, and what is still unknown. It does not hold a precision this essay made up.

Evidence from the plant beats the feed. When the stream and the plant a later reader can inspect disagree, the case keeps the disagreement in view. It does not overwrite the inspection, the failed part, or the missing point with a green trace. It does not overwrite a stale sample with the last good value painted as live. It does not overwrite the wrong unit with a familiar number. A feed that cannot show its sensing, sampling, calibration, and context limits — and that cannot refuse when those readings do not justify the action — is not ready to sit beside the recommendation as if it were truth. A dense, high-frequency, green-looking telemetry feed that cannot show those limits is still wrong for the decision until the observed outcome of that decision is on the case.

The feed is not a diagnosis and not root cause

A diagnosis names a failure mode a later reader can inspect: what failed, on what evidence from the plant, and what is still missing. Root cause is a recorded causal check that names why, on that evidence, and leaves the remaining uncertainty in view. Tags name neither. Sensors name neither. Historians name neither. SCADA names neither. CMMS counters name neither. A green-looking feed names neither, unless the feed is a pointer to approved evidence that already names them — and then the evidence names them, not the stream. The feed is not a diagnosis. It is not root cause. It is not truth.

Symptom Is Not Cause keeps a vibration spike, a hot bearing, a pressure drop, an operator report, a CMMS complaint code, a red tile, or any other observed symptom from being stored as the cause. A sensor that shows the symptom has not recorded the cause. Threshold Is Not Diagnosis keeps a numeric threshold, a limit, an alarm band, or a set-point from being stored as a diagnosis. A tag that sits inside a limit is still a tag. Correlation Is Not Causation keeps co-moving tags from being stored as a recorded causal check. A dense cluster of co-moving points is still co-movement. Trend Is Not Truth keeps a rising or falling line from being stored as a recorded check. A slope drawn on the historian is a slope. The tags under it are not truth for the decision. Noise Is Not Signal keeps chatter from being stored as an actionable signal. A high-frequency chatter that looks busy does not promote it. An aliased signal can hide the motion and still look smooth.

Sample Is Not Population keeps a handful of observed cases from being stored as the fleet. Sampling on one asset, one shift, or one compressed historian window is still that window. The sample rate does not make the window the population the decision claims to cover. Assumption Is Not Evidence keeps an unverified input from being stored as a recorded fact. The belief that the tag is the right sensor, in the right unit, still live, and still tracking the mode is an assumption until those limits are named and the comparison is a measurement.

If the case cannot show the asset configuration, the failure and maintenance history, the sensing and the sampling, and the human judgment that connects the feed to a mode, the trace is a picture without a trail. Evidence Lineage Is Not Optional already requires that trail. Tags do not shorten it. Sensors do not shorten it. Historians do not shorten it. SCADA does not shorten it. CMMS counters do not shorten it. The feed can cite the trail. Citing is not the same as being the trail, and citing is not truth for the decision.

Sensing, sampling, calibration, and context stay on the case

Sensing that is not named is a refusal condition. The wrong sensor is the ordinary case: a tag name that points at a different transmitter, a redundant instrument that was never selected, a tap that is isolated, a probe that is not in the flow. A green-looking value from the wrong sensor does not justify the action. The room can note that a point updated. Noting it records that a stream was available. It does not record that the point was the one the decision needed.

Sampling that is not named is a refusal condition. A stale sample is a point whose time is older than the decision, while the screen still paints it as current. An aliased signal is a stored shape that cannot represent the cycle, the impact, or the mode the question is about, because the sample interval was too coarse or the historian compressed the peak away. High frequency on a different tag does not repair the alias on this one. Density in the historian does not repair a gap the compression removed. When sampling is insufficient for the decision, the feed refuses. It does not become truth for a motion it did not hold.

Calibration that is not named is a refusal condition. Calibration Is Not Validation already keeps a sensor tuned so historical residuals look small from being stored as fitness for the decision. This essay keeps the feed from being stored as truth even when someone says the instrument was calibrated. A calibrated wrong sensor is still the wrong sensor. A calibrated stale sample is still stale. A calibrated proxy that does not track the failure mode is still a proxy. The fit can be close to a reference and still not justify the action. Context that is not named is the same stop: the wrong unit, the wrong asset, the wrong operating regime, the wrong question. A number that is accurate in inches does not justify a decision that required millimetres, and this essay states no conversion and no OEM limit. The context has to be on the case or the reading stays short of truth for that decision.

Confidence Is Not Evidence keeps a strong gut feel, a high model score, or a sure voice from being stored as a recorded check. Being sure the feed is dense enough is still not the check. A green-looking SCADA page is still not a recorded check. Anecdote Is Not Evidence keeps a story from being stored as evidence. A hallway claim that the historian always shows the truth is still a story. The honesty boundary and the verification boundary are the same stop: if sensing, sampling, calibration, and context cannot be shown, or if the readings do not justify the action, the feed refuses, and the case stays unknown. A stream that cannot refuse is not truth.

Telemetry does not authorize corrective work

Authorization is the named act. A drafted next action is a proposal. Recommend is not authorize. Recommend Is Not Authorize keeps that proposal off the decision. A proposal that cites tags, sensors, historians, SCADA, or CMMS counters is still a proposal. A recommendation to investigate is still a proposal. The feed can be dense. The trace can be smooth. The overview can be green. Neither the density nor the color accepts the operational consequence. The feed is not authorization for corrective work.

Corrective work is the change someone intends after a named decision: the repair, the deferral, the inspection, the isolation. A green tag does not authorize running on. A quiet SCADA page does not authorize deferral. A CMMS counter does not authorize the work the counter was used to justify. An aliased signal does not authorize the mode the signal could not show. Sync must not auto-authorize because the feed showed a next step. Sync recommends. A named human decides. If the only support is the stream, the proposal is not ready to be stored as authorized.

Recording an ACTION disposition is work intent. It is not plant execution. Action Is Not Execution keeps that intent off the plant. Sync does not write the work order. The systems that already hold execution authority write the work order or the isolation. Execution stays with authorized systems. A feed, however high-frequency, does not move execution onto Sync, and it does not replace the named decision that would have to come first.

An alert the feed raised is still an alert. Alert Is Not Decision keeps seeing, acknowledging, or silencing an alert from being stored as a decision. A dashboard of tags is still a display. Dashboard Is Not Decision keeps that display off the decision. Color on SCADA does not put it back as authorization, verification of outcome, work-order execution, or plant control. A map of the same points is still a map. Map Is Not Terrain keeps a diagram, a CMMS schema, a dashboard, a digital twin, or a process map from being stored as the operating reality. A SCADA picture of tags is that kind of picture when the screen is treated as the plant.

A feed is not a verified operational outcome

Verification records whether the authorized action did what it was supposed to do on the plant. The case stays open until that check is on the case. Verification stays open until a verified operational outcome is recorded. A later trace that says the tag recovered is a later stream of observations. It is not, by itself, the check. The historian can fill in. The SCADA tile can turn green. The CMMS counter can increment. The feed does not record what the authorized change did. A stale sample after the work can look like recovery. The wrong sensor can look like recovery. A proxy that does not track the failure mode can look like recovery and still miss the mode.

Proxy Is Not Outcome keeps the stand-in off the outcome. A sensor health percent is a stand-in. A green tile is a stand-in. A CMMS counter of closed tasks is a stand-in. A proxy that does not track the failure mode is a stand-in that misses the question. Telemetry clearance is not the check. Clearing the meeting because the feed looked dense, the trace looked smooth, or the overview looked green must not be read as verification of outcome. Sync must not auto-close because the feed was current. Sync must not treat telemetry clearance as Learning credit. A closed trace is not a closed case.

Learning is what a later question is allowed to inherit. Learning Requires a Verified Outcome says the later question inherits that verified outcome, not a hope. Learning requires a verified outcome, not a hoped-for state from tags, sensors, historians, SCADA, or CMMS counters. What was believed may include a labeled note that a feed was consulted and that its sensing, sampling, calibration, and context limits were explicit. Learning does not inherit the feed as if the check had named the cause and confirmed the result on the plant. A later crew that defers the work because the historian looked fine, without the verification on the prior case, is inheriting a stream. That is the wrong lesson into Learning. The Learning chapter publishes the closed record. It does not publish a feed as a verified operational outcome.

A feed is not a Decision Case

A Decision Case opens on a question a named person can decide. The question is recorded before evidence or recommendation. Recording that tags, sensors, historians, SCADA, or CMMS counters were opened does not record that question, and it does not record the decision. The case is the order: Question, Evidence, Recommendation, Human decision, then action, verification, and learning. A stream of observations is not that order.

Treating telemetry as truth shortcuts the Question → Evidence → Recommendation → Human Decision path. The feed jumps to a conclusion. The evidence step is skipped, or it is filled with the trace and called justification. The recommendation is implied by the color. The human decision is treated as already made because the overview is green or the counters moved. That shortcut is not a case. A named human records the Decision Case only against approved evidence. Until the readings are on the case with their sensing, sampling, calibration, and context limits, and until those readings justify the action or the clearance, the feed remains a stream of observations.

A named person accepts, rejects, escalates, or returns the recommendation. That act is the decision. Human Decision Is Not Optional says until that act is recorded, the case is still a proposal. A feed does not replace a named human decision. The tags do not sign accept. The sensors do not sign reject. The historian does not escalate. SCADA does not return the proposal. The CMMS counters do not record who decided. The signed-in Decision Case keeps who decided visible beside the recommendation. The recorded act is mandatory.

Sync may surface telemetry when the limits are named

Sync may surface telemetry beside approved evidence, with the honesty boundary that the feed is a stream of observations and not truth for the decision. Surfacing means the tags, the sensors, the historians, the SCADA picture, or the CMMS counters can sit beside a question so a person can see the sensing, the sampling, the calibration, and the context. The label is part of the surface. Without that label, the screen hides the limit and offers false precision. The honesty boundary is not optional.

Honesty Boundary Is Not Optional says Sync states what is simulated, seeded, or out of scope. Naming that limit does not promote a practice feed into plant truth. The Honesty boundaries chapter keeps Stage-1 from claiming a live conclusion the edition does not hold. A simulated or seeded historian is a practice record. A SCADA picture fed by that practice record is a practice picture. A CMMS counter on a practice asset is a practice tally. None of them is a live operational outcome, and none of them is a substitute for the approved evidence the operator still has to attach. A green-looking practice feed is still not truth for a plant decision.

Sync must not auto-close, auto-authorize, or treat telemetry clearance as Learning credit. Auto-close would store the feed as a finished case. Auto-authorize would store the green trace as permission for corrective work. Learning credit would let a later case inherit the stream. None of those acts is available to the system. A named human records the Decision Case only against approved evidence. Verification stays open until a verified operational outcome is recorded. Sync refuses false precision, including a sample rate, a tolerance, or a health percent this essay does not hold.

Stage-1 evidence is the record held on the case. A live connector that pulls historian or control-system tags sits outside this edition. Simulated or seeded telemetry is a practice record. A practice feed is still a stream of observations. It is not live plant evidence treated as truth, and it is not a substitute for the check the operator still has to record. Sync refuses to treat telemetry as truth.

A model built on the feed is still a model. Model Is Not Reality keeps a digital twin, a KPI model, a risk matrix, an RUL curve, a Monte Carlo run, or an LLM summary from being stored as the plant. Feeding that model a dense trace does not make the model the plant, and it does not make the trace truth. A forecast built on the same points is still a forecast. Forecast Is Not Fact keeps a projection, an RUL estimate, a demand plan, a Monte Carlo percentile, or an LLM-generated outlook from being stored as a fact about the plant. A simulation that consumes the feed is still a simulation. Simulation Is Not Proof keeps a digital-twin run, a Monte Carlo fan, a what-if scenario, or a failure simulation from being stored as proof the plant will behave that way. The input stream does not promote the run.

Blank is not zero. Unknown is not clear. Silence is not clearance.

Blank is not zero. Blank Is Not Zero keeps an empty field from being stored as a measured zero. Filling the blank with the last historian value, a default from SCADA, or a zero from a CMMS counter that was never incremented collects a feed value. It does not collect the missing plant value. A stale sample written into a blank field is still stale. In Recovery, economic assumptions that were left blank stay blank. They are not inferred from a tag. A blank verification field is not a verified outcome of none, and a feed that shows no deviation does not close it. Storing blank as zero invents false certainty and corrupts Recovery, verification, and Decision Case math.

Unknown is not clear. Unknown Is Not Clear records missing evidence as unknown, not as clear or ready. In the Decision Case and in Recovery, missing evidence stays unknown. Filling the unknown line with a tag, a sensor, a historian trace, a SCADA tile, or a CMMS counter does not change the state to clear, and it does not name root cause. Bad or not-ready evidence stays blocked. A feed that looks related to the blocked item does not relabel blocked as ready. The operator must not treat unknown as go because the feed was dense. Go would mean the readings, under their limits, justify the next recorded step. A stream of observations is not that justification by itself. A wrong unit that was not named is still unknown.

Silence is not clearance. Silence Is Not Clearance keeps no reply, no alert, and an empty inbox from being read as approval. No reply is not approval. An empty inbox is not clearance. A quiet feed is not controlled. A SCADA page with no red tile is not a decision. A historian that does not show a spike does not write accept, reject, escalate, or return. It does not record a check that nobody performed. A green-looking overview does not turn the missing response into the named decision. A feed that stayed silent on its sensing, sampling, calibration, and context limits is still not clearance.

Sync recommends. Humans decide.

Sync may surface telemetry. Sync may hold the tags, the sensors, the historians, the SCADA picture, or the CMMS counters beside a recommendation to gather approved evidence, with the honesty boundary that the feed is a stream of observations and that sensing, sampling, calibration, and context limits have to be explicit. That is a proposal. The person who can accept the operational consequence records the decision. Until that act is on the case, the case is still a proposal. The system does not auto-accept because the feed looked finished. Sync recommends. A named human decides.

Execution stays with authorized systems. Coverage of a feed is not control of the plant. Coverage Is Not Control already keeps sensors, CMMS rows, dashboards, and patrol routes from being stored as control authority. A Decision Case that records a feed as a stream of observations does not put Sync in control of the work, the isolation, or the change. The Verification chapter is the check. The feed is not the check. The Human decision chapter is the named act. The trace is not the act.

Tags, sensors, historians, SCADA, and CMMS counters are a stream of observations, not truth for the plant decision. Truth is whether those readings, under their actual sensing, sampling, calibration, and context limits, justify the action or the clearance at hand. A dense, high-frequency, green-looking feed can still be the wrong sensor, the wrong unit, a stale sample, an aliased signal, or a proxy that does not track the failure mode. Treating telemetry as truth is false clearance. It is not a diagnosis, not root cause, not authorization for corrective work, not a verified operational outcome, and not a Decision Case. Evidence from the plant beats the feed. Sync may surface telemetry when the limits are named. Sync refuses false precision. Sync must not auto-close, auto-authorize, or treat telemetry clearance as Learning credit. A named human still decides. Verification stays open until a verified operational outcome is recorded. Execution stays with authorized systems.

The neighboring boundaries still hold

This boundary is not the neighboring ones. Each of those essays still holds. None of them promotes a stream of tags, sensors, historians, SCADA, or CMMS counters to truth for the decision or to a Decision Case.

Calibration Is Not Validation is the fit: a model, a twin, a sensor, or a forecast tuned so historical residuals look small, stored as if the representation were fit for the decision. This essay is the stream under that fit. A calibrated feed is still not truth, and the calibration is still not validation. Simulation Is Not Proof is the run of possible futures. A feed used as input does not make the run proof. Forecast Is Not Fact is the outlook. A historian of the past is not a fact about the decision still ahead, and a projection of that historian is not a fact. Model Is Not Reality is the compressed representation. A model fed by tags is still not the plant. Proxy Is Not Outcome is the stand-in, including a proxy that does not track the failure mode. Verification Is Not Optional is the check that still has to be written. Learning Requires a Verified Outcome is the lesson a later case may inherit. A green trace is still not evidence that the readings justify the action. A high-frequency window is still not a check. A quiet SCADA page is still not the outcome. A later sample does not close the case. The label stays. The limits stay unwritten until they are on the case. The decision stays unwritten until a named person records it.

Map Is Not Terrain is the picture. A SCADA overview stored as the operating reality is still a map. Trend Is Not Truth is the slope. The tags under the slope are not a second kind of truth. Anecdote Is Not Evidence is the story. A retelling of what the historian showed is still not evidence. Anecdote Is Not Evidence keeps the retelling off the case. Sample Is Not Population is the window stored as the fleet. Noise Is Not Signal is the chatter stored as an actionable signal. Symptom Is Not Cause is the observed symptom stored as the cause. Threshold Is Not Diagnosis is the line that was crossed. Assumption Is Not Evidence is the unnamed sensing limit stored in place of a fact. Confidence Is Not Evidence is the sure voice stored in place of a check. Correlation Is Not Causation is the chart of co-movement. Recommend Is Not Authorize is the drafted next action. A fluent trace is still not truth. A tight band on a tag is still not a check. A green tile is still not the outcome.

Unknown, blank, and silence are still the missing record. A feed does not interpret them into clearance. On this record, blank is not zero, unknown is not clear, and silence is not the named decision. A case that is unknown, blank, or quiet is not made ready by a dense historian, a green SCADA page, a moving CMMS counter, or a high-frequency tag from the wrong sensor.

The check that closes the case is still verification. The lesson a later case may inherit is still the verified operational outcome, not telemetry clearance. The trail behind a recommendation is still evidence lineage. An alert is still not a decision. A dashboard is still not a decision, and it is still not the plant. Recommend is not authorize. Those records are either on the case or they are not. If they are not, a feed does not supply them. Evidence from the plant beats the feed.

Where the public statement lives

Field Manual v0 is the public contents of the Decision Case. Start at the manuals index or open Sync Field Manual directly. Evidence is step 02. Observed fact, assumption, and what is still missing stay distinct. A feed does not collapse that split into truth for the decision. A proposal with no basis is not ready for the human decision that follows. Verification is step 06. A feed that says the asset already recovered is not the check. Learning is step 07. A later case inherits the verified outcome, not the tags, the sensors, the historians, the SCADA picture, or the CMMS counters the last shift treated as the lesson. Honesty boundaries keep the edition from treating a stream of observations as live plant control, and they keep the edition from printing false precision. Later editions can deepen a chapter. The spine stays in this order.

What this article is not claiming

This is an essay about the record, not a customer case study. It names no plant, states no savings figure, and claims no prevented failure. It states no OEM limit and supplies no sample rate, no residual tolerance, no remaining-life figure, no probability, and no customer result. It does not claim that Sync executes plant work, writes a work order, isolates equipment, or controls a plant. It does not claim that tags, sensors, historians, SCADA, or CMMS counters are truth for the decision, that a dense, high-frequency, green-looking feed justifies the action, that a stream of observations is a diagnosis, that a trace is root cause, that a green tile is authorization for corrective work, that a current feed is a verified operational outcome, or that telemetry is a Decision Case. It does not claim that self-guided onboarding is a live product path.

Stage-1 readiness means a signed-in user can complete the Decision Case — question, evidence, recommendation, human decision, action, verification, and learning — and Field Manual v0 describes that journey. This edition does not describe plant execute, a live connector tag pull, SMTP invite delivery, or automatic revocation of access on expiry as live. It does not describe Sync writing work orders, isolating equipment, or controlling the plant. Simulated or seeded telemetry and assets are practice records. They are not live plant results, and a practice feed is not truth for a plant decision. Self-guided onboarding is not claimed as a live product path.

The series continues with Dashboard Is Not Control, on why a dashboard reads stored or streamed values and renders them as tiles, trends, and colors, and why that read is not a write that changes the plant. Companion reading: Accuracy Is Not Precision on why repeatability — tight residuals, fine resolution, and stable decimals — is not truth relative to the plant decision and observed outcomes, Calibration Is Not Validation on why tuning a model, a twin, a sensor, or a forecast so historical residuals look small is not validation that the representation is fit for the decision, Simulation Is Not Proof on why a digital-twin run, a Monte Carlo fan, a what-if scenario, or a failure simulation is a controlled representation of possible futures and not proof the plant will behave that way, Forecast Is Not Fact on why a forecast, a projection, an RUL estimate, a demand plan, a Monte Carlo percentile, or an LLM-generated outlook is a forward-looking compression and not a fact about the plant, Model Is Not Reality on why a digital twin, a KPI model, a risk matrix, an RUL curve, a Monte Carlo run, or an LLM summary is a compressed representation and not the plant, Proxy Is Not Outcome on why a stand-in measure, including a proxy that does not track the failure mode, is not the verified operational outcome, Verification Is Not Optional on why the case stays open until the check is recorded, Learning Requires a Verified Outcome on why a later case inherits the verified outcome, not telemetry clearance, Map Is Not Terrain on why a diagram, a CMMS schema, a dashboard, a digital twin, or a process map is a compressed representation and not the operating reality, Assumption Is Not Evidence on why an unverified sensing limit is not a recorded fact, Confidence Is Not Evidence on why a dense feed or a high score is not a recorded check, Trend Is Not Truth on why a slope is not a recorded check and the tags under that slope are not truth, Recommend Is Not Authorize on why a drafted next action, including a recommendation to investigate, is still a proposal, Dashboard Is Not Decision on why a display is not authorization, Sample Is Not Population on why a handful of observed cases or plants is not the fleet, Noise Is Not Signal on why sensor chatter is not an actionable signal, Symptom Is Not Cause on why an observed symptom is not the cause, Threshold Is Not Diagnosis on why crossing a numeric threshold is not a diagnosis, Correlation Is Not Causation on why co-moving tags are not a recorded causal check, Alert Is Not Decision on why seeing, acknowledging, or silencing an alert is not a decision, Silence Is Not Clearance on why silence, no reply, no alert, and an empty inbox are not authorization, Evidence Lineage Is Not Optional on why a recommendation you cannot reconstruct is not an industrial decision, Action Is Not Execution on why work intent is not plant execution, Blank Is Not Zero on why an empty field is not a measured zero, Unknown Is Not Clear on why missing evidence is recorded as unknown, not as clear or ready, and Human Decision Is Not Optional on why the named act is mandatory. A Reliability Assessment asks whether the records can support a conclusion. A Strategic Pilot is a governed proof around one operating decision. The Verification chapter is the public statement of the check this essay refuses to replace with a feed.