How fleet peer confidence is rated: high, directional, low, hidden

What this article covers

Every fleet metric on a franchisee's detail slider carries a peer comparison: a median, a percentile, and a confidence label. The confidence label (High confidence, Directional, Low confidence, or Hidden) tells leadership how much weight that comparison can bear before it should change a decision. This article explains exactly how that label is computed, how the median and percentile behind it are calculated, and how Verinode decides whether a higher or lower number is the good outcome for each fleet metric. It is the mechanics reference for the peer numbers that appear inside the Fleet section; if you have not read how the Fleet section itself is laid out, start with the Assets overview first.

Nothing here is unique to Fleet. Facilities, Equipment, Commercial, and Reputation sliders share the same cohort-size floors, the same median and percentile math, and the same confidence labels. This article documents Fleet's implementation as the concrete example; the mechanics generalize.

Where to find it

Open Assets from the HQ sidebar, then the Fleet tab at hq.verinode.ai/fleet (the tab strip along the top of Assets reads Facilities · Fleet · Equipment). Click any franchisee's tile in the Fleet by Franchisee row (or any Insurance Risk, Compliance Risk, or Accident Risk row) to open that franchisee's Fleet detail slider. The confidence label sits directly under each metric's median line, and a summary version sits in the footer at the bottom of the slider.

The three-scope comparison

At the top of every franchisee's Fleet slider is a scope switcher with three pills:

• Group, the franchisee's own network (its actual peers inside your group).
• Regional, franchisees in the same state.
• National, the full cross-network industry comparison.

Only one scope is active at a time, and every metric tile below the switcher re-renders against whichever scope you have selected. A pill that does not yet have enough peers to compare against is still shown, but grayed out and disabled, with a hover hint explaining what it needs (for example, more active peers in the network, or more peers in the same state). Verinode shows the disabled dimension rather than hiding it so leadership can see that a Regional or National view exists and will light up once the cohort grows, instead of assuming the feature doesn't exist.

Group scope is fully live today: it is computed from your own network's active franchisees. Regional and National scope compare against the broader industry via Verinode's intelligence layer, and light up as that data seeds in; until then those pills show a pending message rather than a fabricated number (see Empty states below).

The cohort-size to confidence mapping

Every peer comparison starts by counting the cohort: how many other franchisees have a usable value for that exact metric, at that exact scope, excluding the franchisee you're viewing. That count (the cohort size) is the only input to the confidence tier. Verinode does not weight cohort quality, recency, or variance into this label, it is a pure floor on sample size, because the risk being managed is anonymity and noise, not data freshness.

The mapping has four tiers:

| Cohort size | Confidence tier | What it means | |---|---|---| | A large cohort | High confidence | Enough peers that the median and percentile are statistically stable and no individual franchisee's data could be reverse-engineered from the aggregate. Read this number as you would any solid benchmark. | | A medium cohort | Directional | Enough peers to publish safely, but small enough that one unusual franchisee can swing the median. Use it to spot a pattern worth investigating, not to set a hard target. | | A small cohort | Low confidence | The minimum floor Verinode will publish a comparison at all. Treat this as a rough signal only, weight it lightly against everything else you know about that franchisee. | | Below the minimum floor | Hidden | Too few peers to publish without risking that the "median" is really just one or two identifiable operators. The comparison does not render; see the tile's own empty state instead. |

Verinode does not publish the exact peer-count thresholds between these tiers in the product, and this article does not either. The reason is the same anonymity boundary that governs the whole comparison: the tiers exist specifically so that a leadership user cannot work backward from "how many peers would it take to move this label" to "how many peers are actually in this cohort." What you can rely on is the ordering: High confidence always has more peers behind it than Directional, Directional always has more than Low confidence, and Hidden means the floor was not met at all.

Two different floors doing two different jobs

The scope switcher's disabled state and the per-metric confidence tier are governed by related but distinct rules, and it helps to keep them separate:

• The scope switcher decides whether a scope is offered at all. A scope with too few peers across the board (using the network's overall Vehicles metric as the bellwether) is disabled outright, with a hint like needing more active peers in the network, or more peers in the same state.
• The per-metric confidence tier applies once a scope is selected, and can differ metric by metric even within the same scope. A franchisee's Group scope might read High confidence on Vehicles (a metric every active peer reports) but Low confidence or Hidden on Accidents (36mo) or At-fault (36mo), because fewer peers have reportable accident history than have a reportable fleet count.

This is why you can be looking at the Group scope for one franchisee and see one tile read "High confidence" while a tile two rows down reads "Cohort too small," both true at the same time, for the same peer network, at the same moment.

How the median is computed

For each metric, Verinode collects every peer's value in the active cohort (again, excluding the franchisee you're viewing), sorts them, and takes the middle value: the standard statistical median, not an average. An even-sized cohort averages the two middle values and rounds to the nearest whole number.

The median is deliberately the median rather than the mean because a mean is far more sensitive to one outlier franchisee (a single office with a much larger or older fleet, or an unusually high premium) pulling the whole comparison off center. The median is the number your franchisee is actually being measured against, and it is the number shown on the tile as "Median [value]" (for example, "Median 4k mi," "Median $1.2M," "Median 78% owned").

If the cohort has too few peers with a usable value (below the Hidden floor described above), no median is computed at all: the tile shows "Cohort too small" instead of a median.

How the percentile rank is computed

The percentile is where the viewed franchisee's own value lands relative to the same peer cohort, expressed as a whole number from 0 to 100. Verinode counts how many peers in the cohort have a value strictly lower than the franchisee's own, divides by the total cohort size, and rounds to a whole percent. A franchisee sitting at the very bottom of its cohort reads a low percentile (close to p0); a franchisee at the very top reads a high percentile (close to p100). The percentile is appended after the median on the tile, separated by a middle dot, for example "Median $1.2M · p62."

Two things follow directly from that mechanic that are worth knowing when you read a number:

• The percentile always describes rank inside the peer cohort only. It does not, on its own, say whether ranking high is good news. A high percentile on Annual commercial-auto premium is not a compliment, see Direction of good below.
• If the franchisee's own value for that metric is missing (not yet reported), the tile still shows the peer median where the cohort supports it, but no percentile, since there is nothing to rank.

Direction of good: why a high number isn't always the good number

Every fleet metric has a fixed direction of good, a lookup that tells Verinode whether a higher number or a lower number is the better outcome for that specific metric. This is what turns a raw percentile into the tone-colored delta line under each median ("+18% vs median" in green, or "-12% vs median" in red), and it is set once per metric, not decided case by case.

The Fleet section's metrics and their directions:

| Metric | Direction of good | Why | |---|---|---| | Vehicles | Higher is better | A larger fleet reflects more operating capacity. | | Avg vehicle age | Lower is better | A newer fleet carries lower mechanical and safety risk. | | Annual commercial-auto premium | Lower is better | A lower premium reflects a better insurance posture for the same coverage. | | Owned vs leased (percent owned) | Higher is better | A higher owned share means fewer ongoing lease obligations. | | Accidents (36mo) | Lower is better | Fewer accidents is a safer fleet, full stop. | | At-fault (36mo) | Lower is better | At-fault accidents are the least ambiguous fleet-safety metric there is. | | Policies < 90d (policies expiring inside 90 days) | Lower is better | A large near-term expiration count is renewal risk, not achievement. |

Two metrics that appear on the Compliance watch panel, Drivers with expired license and Drivers with stale MVR, follow the same lower-is-better logic (an expired license or a stale motor-vehicle-record check is a compliance gap), but they surface as direct counts on the Compliance watch panel rather than as peer-compared tiles.

How direction of good turns into the colored delta line

Once a franchisee's raw value and the peer median are both known, Verinode computes the percent gap between them. If that gap is small (under roughly 5%), the tile reads a neutral "On par with peers" rather than manufacturing a false signal out of noise. Otherwise:

• If the direction of good favors the franchisee's position (higher value with "higher is better," or lower value with "lower is better"), the line renders in Deere Green, matching the platform's Expand signal color.
• If the direction works against the franchisee and the gap is large (25% or more off median), the line renders in Ember Red, matching the platform's Analyse signal color, a flag worth a leadership look.
• If the direction works against the franchisee but the gap is more moderate, the line renders in Hard Hat Yellow, matching the Maintain signal color, worth watching.

These are the same three signal colors used everywhere else on the platform (Analyse, Maintain, Expand). A red delta on a fleet tile carries the same weight as a red delta anywhere else in HQ: it is not automatically a crisis, but it is the platform telling you this specific number, in this specific direction, is worth a look before you move on.

Network targets layer on top, not instead of

Some fleet metrics can also carry a network target if your group has an active program declaring one (for example, a target average vehicle age or a target owned-vehicle percentage). When a target is active for a metric, the tile shows a second line under the peer comparison: a green "Meets target" or red "Below target" tag, with the target value in parentheses and the declaring program's name available on hover. This is a separate mechanism from peer confidence: the target line reflects your own network's stated standard, while the median and percentile above it reflect where the franchisee actually sits against its peer cohort. A franchisee can meet a network target while still sitting at a weak percentile, or vice versa, and both facts are shown side by side rather than collapsed into one number.

Reading the slider's footer line

At the bottom of every Fleet slider is a summary line that restates the active scope's overall standing (using Vehicles as the anchor metric):

• On Group scope with enough active peers, it reads "Within-network comparison" followed by the confidence label and the cohort size, for example "Within-network comparison · Directional · n=14."
• On Group scope with too few active peers, it reads "Peer cohort too small, need more active franchisees in the network."
• On Regional or National scope, before the industry benchmark data has seeded in for fleet metrics, it reads a pending message: industry benchmarks for that scope are still seeding in and will populate as more of the network's intelligence data comes online.

Empty states, verbatim

• A metric tile with no computable median (cohort below the Hidden floor): the tile shows "Cohort too small" in place of the median line, and no percentile, no delta.
• A metric tile at a scope with no data seeded yet (Regional or National, before industry benchmarks are live for that metric): the tile shows "Pending benchmark seed" in place of the median line.
• A metric tile with neither reason recorded: the tile falls back to a plain dash (", ").
• The whole slider failing to load: "Couldn't load this franchisee's fleet data." with a Close link, if the underlying fetch errors out.
• A disabled scope pill (Regional or National without enough peers): the pill still renders, grayed out, with a hover hint such as needing more peers in the same state for Regional, or a cohort too small for cross-network comparison on National.

None of these are broken screens. They are Verinode being honest about what it can and cannot respectably compare a franchisee against right now, rather than inventing a peer number out of too few data points.

Best-practice example

Say you open a franchisee's Fleet slider on Group scope and see Avg vehicle age reading "3.2y," "Median 4.1y · p28," with a green "-22% vs median" delta. Direction of good for this metric is lower is better, so a low percentile here is the favorable outcome: this franchisee runs a newer fleet than most of its Group peers, and the green delta confirms it. Two rows down, Accidents (36mo) reads "Cohort too small," because too few Group peers have a comparable value for that specific metric yet, even though Vehicles and Avg vehicle age both cleared the floor comfortably in the same cohort. That is expected: cohort size is computed per metric, not once per scope, so a franchisee's fleet-shape numbers can be solidly benchmarked while its accident numbers stay Hidden until more peers have reportable accident history. Switching to National scope, both tiles instead show "Pending benchmark seed," because fleet metrics have not yet seeded into Verinode's industry benchmarks at that scope. None of this is a data problem with this one franchisee. It is the confidence system doing its job: publish what the cohort can support, and say plainly when it can't.

Related help articles

• Fleet overview for what the Fleet section is and how it's laid out
• Fleet franchisee detail slider for the full per-franchisee slider these confidence tiles live inside
• Fleet peer scope switcher for the Group / Regional / National switcher this article builds on
• Fleet by franchisee row for the roster row whose tiles open the slider
• Assets overview for how HQ differs from IQ
• Network Health for the composite Network Health score these same signal colors feed into
• Benchmarks for how the same percentile and cohort-confidence mechanics apply at the network-vs-industry level
• Programs for how network targets are declared and become the "Meets target / Below target" tag on a metric tile
• Compliance for the compliance-count metrics (expired license, stale MVR, open events) that sit alongside these peer-compared fleet metrics

Data sources