# How Analyze picks and ranks your competitors

> The math behind the Analyze wizard: how project types weight each layer, how the ten-peer set is built, how wins and the scorecard are decided.

## What Analyze is doing

Analyze answers one question: of every place at your level, which ten are
most like you for the kind of project you chase? It does that in four moves:

1. Turn your project type into a set of **weights** over the five
   competitiveness layers.
2. Measure the **distance** between your layer scores and every other
   place's, using those weights.
3. Keep the **ten nearest**, optionally screened to a similar population.
4. Compare you against them on the **key measures** that project type cares
   about.

Nothing here is AI-generated and nothing is personalised to your account —
two people running Analyze on the same community with the same project type
get the same set. For how the five layer scores themselves are built, see
[How scores are calculated](https://kb.incenti.co/insights/how-scores-are-calculated.md).

## Project-type weights

Each of the five project types weights the layers differently:

| Project type             | Cost | Workforce | Quality of life | Regulatory | Risk |
| ------------------------ | ---- | --------- | --------------- | ---------- | ---- |
| Manufacturing            | 0.35 | 0.25      | 0.05            | 0.20       | 0.10 |
| Tech & innovation        | 0.10 | 0.35      | 0.25            | 0.10       | 0.05 |
| Distribution & logistics | 0.30 | 0.15      | 0.05            | 0.20       | 0.05 |
| A bit of everything      | 0.20 | 0.20      | 0.15            | 0.15       | 0.15 |
| Headquarters & offices   | 0.15 | 0.30      | 0.30            | 0.10       | 0.05 |

These are the product's raw layer weights, which is why the rows do not each
add to 1. A sixth weight covering site screening carries no per-place score,
so Analyze drops it and rescales the five that remain to add to 1 before
measuring similarity. That rescaling changes nothing about the ordering —
only the relative emphasis between the five layers matters.

**A bit of everything** is selected by default.

## How similarity and the peer set work

For every candidate place, Analyze takes the difference between your score
and theirs on each of the five layers, squares it, multiplies it by that
project type's weight, adds the five up and takes the square root. That
number is the distance. **How similar** is simply `100 − distance`, floored
at 0 and rounded, so 100% means identical layer scores under that weighting.

The candidates are then sorted by distance and the **ten nearest** become
your competitive set. Geography plays no part: a place on the other side of
the country with a matching economic profile will outrank a neighbour whose
profile diverges.

### The population band

**Only show similar-sized places** is on by default. While it is on,
candidates must sit between **0.3×** and **3.5×** your population before
they are ranked at all. Turn it off and the whole level is eligible, which
often changes the set completely — closer-profile places of very different
size displace the similar-sized ones that were there before.

Places that carry no usable population are excluded by the band, so the set
can be shorter than ten in sparsely covered levels.

## Which measures each project type compares

The **At a glance** win counts, the **Where you win** / **Where they beat
you** callouts, the versus table and the scorecard all read the same short
list of **key measures**. The list depends on the project type and on the
level, because some fields only exist at some levels.

| Project type             | Key measures at State level                                                                                                         |
| ------------------------ | ----------------------------------------------------------------------------------------------------------------------------------- |
| Manufacturing            | Electricity Rate, Workers' Comp, Tax Climate Rank, Unemployment, Labor Force Part., Freight Volume, Median HH Income, Right-to-Work |
| Tech & innovation        | Bachelor's+, Median HH Income, Broadband Access, Home/Income Ratio, Quality of Life                                                 |
| Distribution & logistics | Electricity Rate, Median HH Income, Tax Climate Rank, Avg Commute, Freight Volume, Pop. Growth (5yr)                                |
| A bit of everything      | Unemployment, Median HH Income, Bachelor's+, Electricity Rate, Home/Income Ratio, Tax Climate Rank                                  |
| Headquarters & offices   | Bachelor's+, Quality of Life, Avg Commute, Home/Income Ratio, Broadband Access, Violent Crime                                       |

At **Metro** and **County** level the same lists apply with four
substitutions, which is why a metro or county head-to-head can have fewer
rows than the state one:

| Change                      | State                                  | Metro and County                  |
| --------------------------- | -------------------------------------- | --------------------------------- |
| Wage measure                | Median HH Income                       | Median Wage                       |
| Freight Volume              | Included (Manufacturing, Distribution) | Not available — dropped           |
| Housing measure for HQ      | Home/Income Ratio                      | Median Home Value                 |
| Extra Tech measure          | —                                      | Median Rent (2BR) added           |
| Distribution growth measure | Pop. Growth (5yr)                      | Counties only; dropped for metros |

## How wins and the scorecard are decided

### Winner per measure

Each measure knows which direction is better — lower electricity rates win,
higher broadband wins, a tax climate **rank** of #1 is best. A row is **You**
or **Them** accordingly, and an exact tie reads **Even**. **You lead X of Y**
counts your wins against every comparable measure, including the ties in the
denominator; the plain-language tone of the callouts uses only the measures
that were actually decided, so an even matchup never reads as a loss.

Callouts are ordered by how large the gap is relative to the whole dataset's
range for that measure, so the biggest real differences come first. A
right-to-work split is treated as a fixed, mid-sized gap rather than being
scaled, since there is no range to normalise against.

### Step 5: Your scorecard vs. this set

The scorecard ranks you against **all ten peers at once** rather than against
one rival. For each key measure it shows your value, the competitors' median,
and your position as **3rd of 11** — you plus every peer that has a usable
value for that measure.

Your rank is 1 plus the number of places in that group that beat you on that
measure. Ranks are then split into thirds by position within the group, not
by raw rank, so being 1st of 2 counts as the top third rather than the
middle:

- **Lead with (across your full peer set)** lists every measure where you sit
  in the top third or are outright first.
- **Watch (across your full peer set)** lists every measure where you sit in
  the bottom third or are outright last.

Neither pill is padded. If nothing is genuinely strong, **Lead with** does
not appear at all — that is the honest answer, not a bug. A right-to-work row
has no rank; it reads **N of M competitors are right-to-work** instead.

## Incentive firepower

The head-to-head shows an **Incentive firepower** card at **State level
only**. It reads the same per-state incentive aggregates as the
**Incentives** screen: an Incentive Competitiveness Score badge and the total
tracked award value for each side, above a sentence reading "You lead /
trail / match \<rival> on incentive competitiveness — ICS X vs Y."

The card hides only when one of the two states has **no incentive aggregate
at all** or no usable score. A state with **limited disclosure** still
appears — its ICS badge is shown in grey to mark that the score rests on
national fallbacks rather than on the state's own record. See
[How Incentive Intelligence figures are computed](https://kb.incenti.co/insights/incentive-methodology.md).

## The compete report and its PDF

**Generate Report** opens the **Competitive Position Report**: KPIs (Overall
score, Rank in set, Closest match, Peers analyzed), **Your Competitive Set**,
**How You Stack Up** (a layer radar against the peer median plus a
composite-score bar chart), one or two **Head-to-Head** tables, and **The
Bottom Line**. Its footnote names the five layers and the sources behind
them.

**Download Report** saves a PDF that mirrors the screen — nothing is
recalculated — but with four caps:

| Section                     | Cap |
| --------------------------- | --- |
| Your Competitive Set rows   | 25  |
| Head-to-head tables         | 2   |
| Rows per head-to-head table | 15  |
| Bottom Line points          | 4   |

Because the report is capped at ten peers anyway, the 25-row cap never bites
in practice; the 15-row and 2-table caps can.

### The print fallback

Opening the report also records it in your account's report history in the
background. **Download Report** can only produce a stored PDF once that
record has landed. Click it in the first moment after the report opens and
it falls back to your browser's print dialog instead, where you can still
"Save as PDF" — the file just won't be attached to the history row. Wait a
moment and click again to get the saved PDF, confirmed by the toast **Report
PDF saved**; the PDF opens in a new tab and a **PDF** link appears on the
row in **Recent reports**.

## Common questions

**Q: What are the actual weights behind each project type?**
A: They're in [Project-type weights](#project-type-weights) above. The short
version: Manufacturing leans hardest on cost, Tech and Headquarters on
workforce and quality of life, Distribution on cost and regulatory, and "A
bit of everything" spreads almost evenly.

**Q: Which measures count as "key measures" for my project type?**
A: See [Which measures each project type compares](#which-measures-each-project-type-compares).
The list is fixed per project type and trimmed at metro and county level
where a field doesn't exist, which is the usual reason a metro head-to-head
is shorter than the state one.

**Q: Why does my competitive set show a state marked "Limited disclosure" in the incentive firepower card?**
A: Because the card only hides when a state has no incentive aggregate at
all. A limited-disclosure state does have one — its score just leans on
national fallbacks, so the badge is greyed to tell you to weight it less.

**Q: What is "Your scorecard vs. this set" and how is "Your rank" computed?**
A: It's the fifth and last step of the wizard, and it compares you with all
ten competitors at once rather than with one rival. Your rank on a measure is
1 plus the number of places in the group that beat you on it, shown as
"3rd of 11". The top third of positions becomes **Lead with**, the bottom
third **Watch**.

**Q: Can I add my own competitor to the head-to-head?**
A: No. The head-to-head only runs against a row in the set of ten Analyze
built. If you already know who you want to line up, use **Compare** inside
Explore instead — see [Compare locations](https://kb.incenti.co/insights/compare-locations.md).

**Q: Does the compete report PDF match the screen exactly?**
A: Yes, with caps. It captures the on-screen report rather than recalculating
anything, but trims to 25 competitive-set rows, at most two head-to-head
tables of 15 rows each, and four Bottom Line points.

**Q: Why did Download Report open my browser's print dialog?**
A: You clicked it before the report's history record had finished saving. The
print dialog is the fallback so you are never stuck without a copy. Go back,
wait a couple of seconds and click **Download Report** again — you should get
the **Report PDF saved** toast and a **PDF** link on the history row.

## Related

- [Find who you compete with (Analyze)](https://kb.incenti.co/insights/analyze-competitors.md)
- [How scores are calculated](https://kb.incenti.co/insights/how-scores-are-calculated.md)
- [How Incentive Intelligence figures are computed](https://kb.incenti.co/insights/incentive-methodology.md)
- [Compare locations](https://kb.incenti.co/insights/compare-locations.md)
