A Per-Company Growth Score Built From Real-World Behaviour

A card and payments provider sees its customers' activity in fine detail, but that activity is a lagging indicator: the operational change that drives it happens outside the relationship, and earlier. Veridion reads that change across a company's entire digital footprint over time, every observable move a growing or contracting business makes, and compresses it into a single explainable growth score, from −10 to +10, with the reasoning and a confidence attached to every one.

1. 1
   Pulled the whole footprint over time

   locations, headcount, web presence and technology for every company, from the company knowledge graph across rolling 12-month windows

2. 2
   Read every operational change in it

   physical expansion and contraction, hiring, digital and product reinvestment, observed not self-reported

3. 3
   Combined them into one score

   a single explainable −10 to +10 growth score per company, with a written reason and a High/Medium/Low confidence

4. 4
   Separated the book across the full range

   strong decline through strong growth, most companies corroborated by multiple independent signals at once

5. 5
   Saw past a flattering single signal

   a rebuilt website looks like growth; weighed against falling headcount, the combined score still reads contraction

6. 6
   Routed each score to an action

   strong risers, quiet decliners and early-warning cases each separated out, before the numbers

Spend data sees growth only after it happens

A card and payments provider already sees its customers in fine detail: every transaction, every line of spend. But that view is a lagging indicator. By the time growth or distress reaches the spend data, the decision window has narrowed.

The change that drives the spend, a company opening locations, hiring, entering or exiting markets, rebuilding its website, happens earlier and outside the relationship, and it is visible from the outside if you know where to look.

The provider wanted to read that change ahead of the transaction: which businesses are actively growing, which are softening, who is accelerating and who is sliding toward distress, before any of it surfaced in their own data. The hard part is doing it at portfolio scale, from observed evidence rather than self-reported numbers, and in a form a commercial team can actually act on rather than a black-box propensity flag.

Four observed signals, combined into one score

Veridion built a single growth score, from −10 to +10, for every company in the portfolio. It draws on the company knowledge graph: the live, continuously refreshed map of operating companies assembled from real-world digital and operational footprints. The score reads operational change across a company's entire digital footprint over rolling twelve-month windows, anything a growing or contracting business does that leaves a trace.

In this build four behaviours carried it, each scored on the same scale: net change in active physical locations for physical expansion, employee-count trajectory for hiring, an AI reading of the live website against itself a year ago for strategy and momentum, and a verified move onto a more modern web stack for digital reinvestment. The composite is their weighted average, clamped to the band, and every score arrives with a written one-line reason and a High, Medium or Low confidence. The footprint holds more signals than any one engagement weights, so the same machinery extends as new behaviours prove predictive.

The reading is observed, not declared, and guarded against its own failure modes. A newly seen location only counts once it is verified open, so backfilled data cannot fake an expansion. Employee history passes three anomaly checks, filtering glitches, single-identity duplicates and impossible overnight doublings. The website read is treated as calibration-grade, one input of four, never ground truth on its own.

The point of combining independent signals is that no single feed can be gamed or misread into the wrong answer: a company can rebuild its website and look like it is growing while its headcount and client list quietly fall, and only the composite, weighing that against the other signals, reads it correctly as contraction. Every score decomposes back to the dated location, headcount, tech-stack and website-change records that produced it, so a number can be defended row by row rather than taken on trust.

Act on a company's trajectory before the spend

The provider gets a forward-looking, explainable read on the whole book. Across a worked sample the score separated companies cleanly from strong decline through steady to strong growth, with most cases corroborated by three or four independent signals firing at once rather than a lone reading.

Because the same number maps to a play, it is operational on day one: strong risers surfaced for proactive cross-sell, steady growers for upsell and expansion-readiness, softening accounts for retention and exposure review, and sharp decliners as early warnings for churn and credit attention, with the confidence flag deciding where a human looks first.

The result is that the provider can act on a company's trajectory before it reaches the spend, and defend every decision against the evidence underneath it.