The fleet operations briefing - on the telematics you already run. This guide walks every role through the platform, defines the color language, explains how every number is made and which data feeds it, and answers the questions that come up in the first month. It is written to be quoted: every claim in here matches what the app actually computes.
Envision is read-only, always - it can look at your fleet's data but can never change, delete, or write anything on your telematics platform. Sign-in differs by vendor:
For Motive and Samsara, tick "Remember this token on this device" to skip re-pasting - it's stored only in your browser, removable anytime. The first load pulls your fleet's recent history; large fleets take a minute or two, and the progress lines narrate what's happening.
Every color in Envision is a claim, not a decoration. The same six colors mean the same things on every page, so once you've read this section you can read any screen.
| Color | What it claims | Where you'll see it |
|---|---|---|
| Red | Critical, declining, or needs a decision now. | CRITICAL briefing lane · Open queue items · Overdue coaching · offline vehicles · At-Risk score band · a delta moving against you (cost up, score down). |
| Amber | Caution, rising, or in flight awaiting a result. | SAFETY lane only when a behavior is genuinely rising · Actioned queue items · Awaiting-result coaching · quiet (24-72h silent) devices · Fair score band · idling on the map. |
| Green | Good, improving, or resolved. | WIN and SAVINGS lanes · Fleet Wins pill · Closed queue items · Completed coaching · Excellent score band · online vehicles · driving on the map · a delta moving your way. |
| Blue | Information without alarm - a fact that makes no trend claim. | The SAFETY lane when your largest behavior is not rising · Scheduled coaching · stopped on the map · evidence-source text on queue cards. |
| Gold | Attention and memory - the brand's own color, never an alarm. | Selected chips, tiles, and lenses · the Priorities pill · Recurred badges ("you closed this and it came back") · Honor Roll. |
| Grey / muted | Dark by name: no data, no prior, or not qualified - never a fake zero. | "prior - not held" ribbon stats · unranked below-floor drivers · tiles that say what's missing · Up-next coaching (no history yet). |
Score coloring everywhere follows two thresholds you can tune in Settings (defaults shown): Excellent above 85 · Fair 71-85 · At Risk 70 and below. The same bands color driver scores, the fleet score, and asset health.
Your three stops: the delta ribbon at the top of the briefing (four numbers, each vs the prior window - better or worse, before you've read a sentence), the green Fleet Wins pill (what's improving, with the numbers that prove it), and the idle-cost math - real dollars, formula attached. Five minutes, once a day, and you know whether the operation is getting better.
Start at the Fleet Operations Briefing - the ribbon, then one card per domain (section 4 explains its logic). Then the gold Priorities pill: the day's five, at most two per lane, each item clickable to where the action happens. Use the group filter to review one division at a time; use ↓ CSV on any tile for the meeting handout.
The Safety page is your stack rank - one score, computed the same way regardless of vendor, with a search box that reaches even unranked drivers. The Honor Roll podium shows last month's winners and this month's live race. The Coaching page is your workflow: the loop tiles, the ranked roster, and the outcome loop that files a verdict on every conversation (section 9). On the queue page, set your Lens to Coaching and the app opens to your lane every time.
The Compliance page tracks DVIR (unrepaired critical defects surface in Priorities, not in a folder), ELD duty status at a glance, and inspection pace. Every table exports to CSV for the audit binder. Coaching session logs - date, behaviors, commitment, owner, outcome - are your documented corrective-action record; print any driver's record from their coaching card.
Asset Health Rank scores every vehicle from fault pressure and open defects; Highest Risk Asset names today's leader. Fault codes read in plain English, with check-engine, stop, and protect lamps called out. Set your queue Lens to Faults (and check DVIR) - the shop opens the same app as everyone else and sees only its own lane.
Three surfaces read the same snapshot of your data - they can never disagree on facts. They differ in what selects the content and what question each answers.
| Today's Priorities | Fleet Operations Briefing | AI Briefing | |
|---|---|---|---|
| Question | "What do I do first?" | "How is my fleet right now?" | "What's the story behind the numbers?" |
| What decides | Deterministic rules + memory - items open, hold state, recur, and auto-close from your data | Deterministic rules, re-derived fresh on every pull - no memory, no model | An AI model, on demand only - but it may only annotate numbers already on the dashboard |
| Shape | A ranked queue with a lifecycle (Open → Actioned → Closed, Recurred) | The delta ribbon + one lane per domain: Critical / Safety / Win / Compliance / Savings | Trajectory, connections between domains, and what acting is worth |
| Cost | Free, automatic | Free, automatic | Runs only when you press Generate - never an auto-spend |
The doctrine behind all three: deterministic selection, AI annotation. Engines pick what appears; the model is only ever allowed to narrate on top. Nothing on your screen was chosen by an AI.
The AI reads the rules before the severity. Since v1.73 the AI briefing is required to check applicability before drawing conclusions - a driver at zero because they haven't driven is a data note ("unranked below the mileage floor"), never a "worsening safety" finding. An acknowledged rule makes the real signal sharper.
The delta ribbon (top of the briefing card): Fleet score, Critical fault events, DVIR defects open, Idle cost - each against the prior window of the same length. Two honesty details: DVIR is an open-now state, so it has no prior window and says so; and idle cost is priced with the same rate and fuel price on both sides, so the delta is a like-for-like comparison. A stat with no prior goes grey and says "prior - not held," never a fake zero.
The briefing lanes. CRITICAL names the single worst critical-fault vehicle and its leading code. SAFETY shows your biggest rising behavior in amber with who owns it ("72% comes from 4 drivers…") - or, when nothing is rising, your largest behavior in blue with "not rising" said out loud. The WIN lane appears only when a behavior genuinely fell (prior ≥ 10 events, drop ≥ 15%) and names how many of last window's top offenders came back clean - recognition with receipts. COMPLIANCE is the DVIR gap; SAVINGS is idle dollars with the formula in the sentence.
Concentration respects the ranking floor. "Who owns it" counts only drivers above the qualifying-mileage floor; anyone below it is disclosed by count ("2 more with events sit below the 50-mi floor - unranked, not hidden") rather than headlining on 12 miles of driving.
Start Here / Your Fleet - Right Now. The impact tiles are picked by fixed rules from the pillars (never by a model, never duplicating the Right Now strip); every tile shows its number, its composition line, and clicks through to the page where you act. Customize tiles (top right) hides what you don't use.
The AI Briefing card sits below the ops briefing with the same collapse behavior. It runs only when you press Generate, reads only the digest of numbers already on screen, and writes trajectory · connections · the arithmetic of acting.
The pillars ride every module page, aimed at that page. The gold Priorities pillar and green Fleet Wins pillar appear on Dashboard, Safety, Coaching, Compliance, and Maintenance. On a module page the pillar carries that page's lane only - Maintenance shows fault / device / DVIR items, Compliance shows DVIR / fuel, Safety and Coaching show the coaching lane - and the drawer's title says so. Items in other lanes are counted with a road to the Dashboard pillar, which keeps the fleet-wide view. Same ledger, aimed at the desk you're sitting at.
The Dashboard drawer's top five is a fleet view, not the loudest generator. The day's five holds at most two items per lane (Fault / DVIR / Coaching / Device / Fuel - the tag on each item names the lane). On a fleet with forty open faults, weight-only ranking would show five faults today, tomorrow, and forever; the cap keeps the shop's two worst faults on top while the unrepaired DVIR, the declining driver, and the idle burn keep their seats. Whatever the cap benches is disclosed - "+ N more open items in the full queue →" - never silently dropped. On a day when only one lane has items, the cap yields: five real items beat four diverse ones.
Ranking. Items rank by exposure weight (critical fault 100 · DVIR 88-92 · declining driver 80 · offline device 70 · idle burn 50). An item that recurred after you closed it gets a +15 boost - enough to jump its lane, never enough to let a $52 idle note outrank a fresh critical brake fault.
Fault cards show their work. The detail line is the asset's real risk picture - breakdown rank in the fleet, the critical/other mix, distinct codes, check-engine lamp, the leading code by name, and any open DVIR defect on the same asset riding along. "View the N codes" expands the outline: every code with its count, criticality, and when it last fired - straight from your platform's own fault readings, no AI in the loop. The Dossier adds context instead of repeating the card: asset health with its band, breakdown rank, and one-click jumps to Asset Health (search pre-filled) and the full fault list.
Two moves, two colors. Mark actioned (blue - dispatched, in flight) and Close with note (green - done; the note becomes the record). The recommended next move sits above the buttons; the buttons live in the card's bottom-right corner with the item's history holding the bottom-left. The four state tiles carry their colors as a tinted fill, so Open / Actioned / Closed / Recurred read at a glance from across the room.
The Lens row (queue page): All / Faults / DVIR / Coaching / Devices / Fuel, with open counts. One click filters the queue to the lane you own, and the pick sticks on your browser - the shop foreman opens the same app and sees six items, not four hundred. The lanes are the roles: shop = Faults + DVIR, safety = Coaching, ops = Devices, finance = Fuel.
Lifecycle. Open - needs a decision. Actioned - you dispatched something and are awaiting the result. Closed - done; your note becomes the permanent record. Recurred - you closed it and the condition came back; your original note is preserved, because "the last fix didn't hold" is the most valuable sentence in maintenance. The Dossier (right rail) shows every item ever opened on a subject and how each was resolved - click any vehicle or driver name to load it.
What counts as a "pull." A pull is any refresh that brings new data: signing in, pressing ↻ Refresh, or the hourly auto-refresh. Changing the group filter or time range recomputes numbers but is not a pull - the queue only reconciles against genuinely new data. If a condition clears and stays clear across two consecutive pulls, the item closes itself with credit. One clean pull can be a data gap; two is a trend. Two honesty rules ride on top for faults: the close note reads "no longer reported - not verified repaired" (a record claim, never a repair claim), and silence is not a clean pull - a fault can only earn credit from a vehicle that is actually reporting. A truck that goes dark with an open fault flips to a grey State unknown chip and its clock freezes until the vehicle talks again; it is never auto-closed by silence.
In the demo, the "Scripted data arc" control (top right) replays a three-day sequence through the real engine: Day 2 brings a truck back online and climbs a fault count on the same card; Day 3 auto-closes the cleared conditions. Close the fault with a note before pulling and it returns flagged Recurred, note intact.
The score starts at 100 and decays with weighted events per 100 miles - harsher behaviors weigh more, and the weights live in Settings where you can see and tune them. The curve auto-calibrates so your fleet's median qualifying driver lands near 75 (or set the sensitivity manually).
Qualifying miles: a driver must log enough miles in the window to be ranked - a 12-mile hero can't take the podium, and a driver who hasn't driven can't be "worst." The floor pro-rates for short windows. Unranked drivers aren't hidden: the search box above the stack rank finds any driver, and a below-floor match appears under the table with their mileage and a click-through to their full profile. Percentiles ("top 8%") are computed on the full ranked population and hold steady while you search.
The Honor Roll shows this window's live race and (behind the toggle) the prior window's winners - same qualifying gate as the rank.
Why doesn't the score match my Samsara/Motive score?
Different formulas - theirs live inside their platform, ours is the same math on every vendor. That's the point: one comparable number across your whole operation, whichever systems it runs on. Use whichever you trust; the briefing is the product.
The loop tiles across the top - Completed (green) / Awaiting result (amber) / Scheduled (blue) / Overdue (red) - are clickable filters over the whole coaching loop. Overdue means a session was scheduled and its date passed without being held; that's a red state because a stalled loop is a silent one.
The roster (left) ranks drivers by urgency (score band + decline), each with a chip showing where they sit: Up next → Scheduled → Awaiting → Done / Overdue. Click a driver: the highlight (center) shows their banded score, daily sparkline, direction, the Focus box (their top behavior, pre-written as the conversation opener), what improving earns (fuel, risk, and score arithmetic), and their session history.
Log a session on the right - date, behavior, notes, the driver's commitment, owner, follow-up date. The score at that moment is captured automatically as the before-picture. When the follow-up date passes, Envision computes the delta and files a verdict - Improved (+3 or better) or No change - never silently. Improved sessions feed Fleet Wins with attribution: coached and improved, named, dated. 🖨 Print record on any driver's card produces the printable session history for the personnel file.
Coach from anywhere: the 🎯 button in any driver profile, the coach link on the map's driver block, or a "Coach {driver}" priority.
The page opens with three scores: Fleet Ready (the maintenance score - vehicles with no open critical fault, no past-due service, and a reporting device; each not-ready vehicle counted once at its worst reason - and each reason is now a clickable chip: click the critical count and the tile lists exactly those assets with their fault, its status, and its age), the Maintenance tile (upcoming / past due / check engine), and Service Outlook - what's due before it's late: past due · due within 14 days · on schedule, with the next services named ("B-214 · oil & filter - ~420 mi, ≈3 days at that truck's own 30-day pace"). Mile rules convert to days at each truck's real pace; day rules are exact; engine-hour rules are counted but never day-projected - drive-time is not an honest engine-hours pace, and the tile's tooltip says so. The Visit Bundler joins the fault ledger, driver-flagged defects, and service due per vehicle - "3 items · one visit" - with a low-utilization hint when the truck is sitting anyway. DVIR record tiles now live on the Compliance page; the Asset Health Rank runs open by default and remembers your collapse choice.
Faults are tracked as episodes - one record per vehicle-plus-code, from first fire to resolution, surviving window changes. The Faults: DTC tile counts open episodes; Highest Risk Asset names the worst vehicle with the management sentence ("open 6 days - first fired Jul 20, 4 re-fires · not yet dispatched"); the Fault Center (its own section on the Maintenance page, and behind every fault click-through) lists each episode with its status: New (first fire on record) · Active (firing in the freshest data) · Inactive - unresolved (stopped firing, no repair on record) · Unknown - not reporting (the truck went dark with the fault open - never auto-resolved by silence) · Returned ×N (came back after closure) · No longer reported (two reporting pulls clean - a record claim, never a repair claim). Each row shows first fired, Active time (first fire until resolution, red when over the time bar: critical 1 day, other 3, with the last fire beneath it), and re-fires; the purple Fault Insights box on the page explains all of it in place. The same episode line appears in the map's asset panel fault tiles. An episode closes as "no longer reported" after two pulls where the vehicle was actually reporting - a record claim, never a repair claim.
Every asset name is the same door. Click a vehicle anywhere - the priorities queue, Fleet Ready's chips, Service Outlook, the Bundler, the Fault Center table, Asset Health Rank, or the map's asset panel - and the same Asset Detail card opens: odometer, engine hours, that truck's own mi/day pace, reporting status, every open fault episode in the ledger's language, its service rules with due states, and any driver-flagged defect connected from the Compliance page. The ✨ Explain this asset button is the card's only AI line - on demand, grounded in that card's records alone, one click = one spend. No cost figures appear until a cost source connects - dark by name.
Every denominator in Envision - scores, health, compliance pace, offline counts - is built on two laws:
Identity: only vehicles. Your platform's own vehicle/trailer labels are decisive in both directions - if your database labels an asset a trailer, it never appears in a vehicle count, and Envision never guesses a class the platform has already ruled on. Trailers, equipment, and unpowered assets are out of scope until asset reporting ships; that's the sign-up expectation, stated plainly.
Activation: only units that have ever communicated. A device that has never reported - still in the box, awaiting install - leaves every denominator and is disclosed as "awaiting install/activation" on Fleet Reporting. A unit that communicated once and then went dark stays in the population: that's what "offline" means, and hiding it would hide the problem.
Windows. 24H / 48H / 7D / 30D / MTD, anchored to local midnight (MTD means "since the 1st at 00:00," not "the last N×24 hours"). Every window's fetch spans twice the range, so "vs prior" comparisons are computed from data already in hand - never estimated.
Pull vs recompute. Sign-in, ↻ Refresh, and the hourly auto-refresh are pulls (new data; the queue reconciles). Group and range clicks are recomputes (same data, new lens; nothing "happens" to the queue).
Rate limiting (Geotab). Geotab admits ~650 API queries per user per minute. Envision budgets itself to 550, paces large pulls to stay under it, and backs off and retries automatically if the server ever says otherwise - a big-fleet pull may take a little longer, but it completes instead of erroring.
What each platform feeds, honestly: Geotab is richest (trips, behavior rules, fault history, DVIR, ELD, measured fuel, IFTA). Motive covers trips, events, fault history, DVIR, ELD. Samsara v1 shows currently-active fault codes and lacks measured fuel and ELD mapping. Lytx v1 is strongest on safety events; its feed reports motion but not ignition, so idle metrics stay dark. Azuga covers the core trip/event/status set. Wherever a platform doesn't provide something, the tile says so - a gap named is a gap you can plan around.
Five laws govern every figure on every page. They exist because a briefing you can't trust is worse than no briefing.
Telematics platforms register fault events - the same code re-fires on every ignition cycle and polling interval, so one failing oil-pressure sensor can log a thousand "events" in a month. Envision counts conditions: distinct vehicle-plus-code pairs, the number a mechanic would count. Every headline speaks conditions; the raw event volume stays visible as evidence ("re-fired 1,113×") and in tooltips, because re-fire volume tells you the condition is still active - it just isn't a thousand separate problems. A condition with no event in the prior window is tagged new - that's the news. Breakdown risk and asset health score conditions too, so one cycling sensor can't outrank a truck with three real problems or peg an asset's health at the floor forever.
Possibly both - and now the card says why. A defect's platform record can be marked repaired, marked not repaired, or carry no repair status at all (common until someone logs a repair record). Blank-status defects used to vanish from both counts; they now appear as their own disclosed line - "N defects carry no repair status - not counted as repaired or awaiting." A blank record is not a repair, and Envision won't claim it either way.
No - the color is the claim, and the claim changes with the comparison. Each range compares against its own prior window of equal length, so a behavior can be rising week-over-week (amber on 7D) and flat day-over-day (blue on 24H). Every trend sentence names its window so you can see which comparison it's making. If the prior window has too few events to support a trend either way, the lane says "too few prior events to call a trend."
Geotab, Lytx, and Azuga: no - each user sees what their own account is cleared to see, enforced by the vendor. Motive and Samsara: yes - access follows the shared token your admin created. The group dropdown is a viewing convenience, not access control.
Envision pulls twice the selected window's history on first login (that's what makes every "vs prior" honest) - a 650-vehicle fleet takes a couple of minutes, then the session cache takes over. On Geotab, Envision also paces itself under the API quota; the progress lines narrate what's arriving.
They're below the qualifying-mileage floor - unranked, not hidden. Type their name in the rank search box; they appear under the table with their mileage, one click from their full profile. The average and the bottom-5 list only ever include qualified drivers, so a driver who barely drove can't drag the fleet number.
Yes - vs Prior on Asset Health is now stated in health points on the 0-100 scale, bounded ±45. The old display leaked the raw event-count swing.
Say so - that's the most valuable report you can send. Sometimes it's a different formula (scores), sometimes a different window, occasionally our bug - and occasionally the vendor's: our IFTA reconciliation found phantom miles from corrupted odometer feeds in a vendor report. The formula is always visible so the difference can be traced.
No. Read-only is architectural, not a setting - and on Samsara your admin can enforce it cryptographically with a read-only token.
Export them first (Coaching page → Export sessions), import on the new machine. Automatic team sync is the Phase-2 upgrade.
Reload; the amber banner describes what happened. Screenshot it and send it via Feedback - it tells us exactly what to fix.
Envision · Continuum Fleet Services · Read-only, always · Guide revised for v1.98 · Questions this guide didn't answer? The Feedback button reaches a human.