> For the complete documentation index, see [llms.txt](https://docs.sumble.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.sumble.com/enterprise-services/account-scoring.md).

# Account scoring

The Sumble Score ranks every organization in your account list so you can focus on the accounts most likely to be genuine buyers — not just the ones that look busy.

{% hint style="info" %}
**Sumble Score v2** — We've updated the scoring model. Scores now reflect structured signals — size, growth, and market concentration — adjusted for company profile. Accounts that frequently mention lots of technologies but rarely buy software (such as IT services firms) now rank lower, so your highest-priority accounts are more likely to convert.
{% endhint %}

## How the score is built

Every account is scored across three dimensions. Each dimension gets a fixed share of the 100-point budget, and within each, individual signals are weighted by importance — not by how many technologies or job functions your configuration happens to include.

| Dimension     | Budget share | What it measures                                                                                                                                                                        |
| ------------- | ------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Size          | 50%          | How much relevant headcount does this account have? Split evenly between the count of people in your target persona and the count of people on technical teams.                         |
| Growth        | 30%          | Is the account hiring in your target persona? Growth signals how quickly relevant headcount is expanding, which often precedes new tool adoption and budget allocation.                 |
| Concentration | 20%          | What share of the company is your target persona or tech team? A small company where your persona is 30% of headcount often signals a better fit than a large company where they're 3%. |

## Why technology categories, not individual technologies

Many of the signals that feed into your score come from job post activity — including which technologies a company is actively hiring around. When Sumble evaluates those technology signals, it scores at the category level rather than by individual tool.

Job posts frequently name several technologies in passing — "experience with tools such as X, Y, and Z." Scoring each technology individually would give an account a separate bonus for every one of those mentions, inflating the score for what is really a single signal. The bonus applies once per category instead, regardless of how many technologies from that category appear in an account's job posts. This keeps scores grounded in genuine adoption signals rather than the breadth of a job description's technology list.

## Profile adjustment

After the base score is calculated, it is multiplied by a profile factor derived from the account's organization tags — the attributes Sumble uses to describe what kind of company an organization is, such as its industry, business model, or customer type.

Org tags that indicate a strong buyer profile increase the score; tags that suggest a poor fit reduce it. An IT services firm, for example, may score high on raw size and technology mentions — but if its org tags signal it is more likely a reseller or partner than a buyer, its final score is adjusted down accordingly.

Accounts with no relevant org tags receive a multiplier of 1 — their score is unchanged.

{% hint style="info" %}
The org tag adjustments applied to your account list are configured by Sumble based on your ICP. Contact your account team to review or update your profile configuration.
{% endhint %}

## What changed from v1

|                    | v1                                                                                                                                   | v2                                                                                                                                                                                                     |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Signal weights     | Divided a fixed budget by however many technologies and job functions the domain tracked. More entities = smaller weight per entity. | Fixed budget split by dimension (size / growth / concentration), then split evenly within each dimension, with secondary entities weighted below primary ones. Entity count no longer dilutes weights. |
| Profile adjustment | Not applied                                                                                                                          | Final score multiplied by an org-tag factor. Company type is accounted for — not just what's in the job posts.                                                                                         |
| IT services firms  | Often ranked high — lots of technology mentions drove up the raw score.                                                              | Ranked more accurately — profile adjustment corrects for companies that mention many technologies but rarely buy.                                                                                      |
| Score range        | 0–100 (same formula, unbounded in edge cases)                                                                                        | 0–100, tighter distribution                                                                                                                                                                            |

## The math

Each raw signal value is normalized to a 0–1 scale, then weighted and summed. Counts use a log transform; growth rates and shares use raw values.

**Step 1 — Normalize**

```
// p99 = 99th percentile of x across all organizations
// x   = ln(1 + raw) for counts; raw for growth / shares

normalized = min( (1 − e^(−x / p99)) / (1 − e^−1) , 1 )
```

An organization at the 99th percentile scores approximately 1.0 on that signal.

**Step 2 — Weight and sum**

```
// Weights come from a fixed budget: size 50 | growth 30 | concentration 20
// Within each dimension, signals share the budget evenly within a tier:
//   w = 1 per key-tier signal, 0.6 per other-tier signal

base_score = Σ ( normalized × effective_weight )
```

**Step 3 — Apply profile adjustment**

```
// adj_t = per-tag adjustment fraction (e.g. +0.15 or −0.20)
// Boosts apply to base first; penalties apply to the boosted subtotal

multiplier  = ∏ ( 1 + adj_t )   // over matching boost tags
boosted     = base_score × multiplier
final_score = boosted × ∏ ( 1 + adj_t )   // over matching penalty tags
```

In one line: **v1** was `score = Σ (normalized × heuristic_weight)`. **v2** is `score = Σ (normalized × budgeted_weight) × tag_multipliers`.

## Common questions

**Will my scores change?**\
In the Sumble web app, yes — scores will reflect the v2 model once it rolls out. The relative ranking of accounts is more accurate, but individual values will differ from v1.

For CRM enrichments, the transition is controlled. Each enrichment config has a `version` field (`v1` or `v2`, defaulting to `v1`), so existing CRM integrations keep receiving v1 scores until you explicitly switch. You can run v1 and v2 side by side to compare before cutting over. If you have automations or thresholds that depend on specific score values, review them before switching your enrichment config to v2.

**Can I customize the weights?**\
Coming soon. Enterprise admins will be able to adjust signal weights and profile tag multipliers directly in the Admin portal — changes will apply to your next scoring run. Reach out to your account team if you'd like to be notified when it's available.

**How often are scores updated?**\
Scores are recalculated daily as new job posts, people data, and CRM enrichments are processed.

**What if a score seems wrong?**\
Open the account's score drawer to see a full breakdown by dimension — you can see exactly which signals are contributing and by how much. If something looks off, contact your account team.

## Appendix: Score breakdown details

The following explains how points are attributed per tag in the score breakdown UI. This is included for transparency — it does not affect how the final score is calculated.

### How points are shown per tag

When multiple tags apply, their multipliers stack — the combined effect is larger than any single tag on its own. To show a meaningful number per tag in the score breakdown, the combined uplift is split across tags in proportion to each tag's own percentage.

For example:

* Two +15% tags multiply to 1.15 × 1.15 = +32.25% combined. Since both tags are equal, the uplift is split evenly — +16.13% each.
* A +5% and a +10% tag: 1.05 × 1.1 = +15.5% combined, split 5:10 — so +5.17% and +10.33% respectively.

The percentage shown next to a tag in the web app is that tag's own multiplier. The points shown are its share of the combined uplift.

### Boosts and penalties are applied separately

Boosts and penalties are kept in separate piles rather than pooled together. This ensures that the sign of each row in the breakdown always matches what the tag actually is — a boost always shows a positive contribution, a penalty always shows a negative one.

In practice: boosts apply to the base score first, producing a boosted subtotal. Penalties then apply to that subtotal — not to the original base. This order means penalties are calculated on a larger number, which keeps boosts from being artificially inflated by doing it the other way around.