Lesson 03 · outer fringe

targetingCriteria

It looks like fiddly nested JSON. It is conjunctive normal form with a handful of illegal combinations. Once you see the CNF you will never misread it again.

≈13 min · pinned to li-lms-2026-08

The shape

One field on the campaign — the level that owns targeting, as you correctly guessed in your calibration. Its value is a boolean expression over facets:

{
  "targetingCriteria": {
    "include": {
      "and": [                                  // ← clauses, AND'ed together
        { "or": {                               // ← one clause
            "urn:li:adTargetingFacet:locations": ["urn:li:country:de"]
        }},
        { "or": {
            "urn:li:adTargetingFacet:titles":       ["urn:li:title:4"],
            "urn:li:adTargetingFacet:seniorities":  ["urn:li:seniority:7"]
        }}
      ]
    },
    "exclude": {
      "or": {                                   // ← no "and" allowed here
        "urn:li:adTargetingFacet:employers": ["urn:li:organization:1035"]
      }
    }
  }
}

Read it as one sentence: (location = Germany) AND (title = X OR seniority = Y), minus (employer = 1035).

The one rule people get wrong

Inside a single or object you may list several different facets. They are OR'ed with each other, not AND'ed.

Two facets that must both hold go in two separate clauses of the and array:

"and": [
  { "or": {
    "…:titles":    ["…:title:4"],
    "…:locations": ["…:country:de"]
  }}
]

"and": [
  { "or": { "…:titles":    ["…:title:4"] }},
  { "or": { "…:locations": ["…:country:de"] }}
]

So the structure is exactly CNF — an AND of ORs. Every clause must be satisfied; within a clause anything will do. Values inside one facet's array are also OR'ed, which is the only part that behaves the way people guess.

And the asymmetry

exclude takes a single or — no and array. You cannot express "exclude people who are both X and Y". Everything you list is OR'ed and subtracted wholesale.

Order of evaluation is fixed: include first (and it may not be empty), then exclude is subtracted from the result. And there is a floor — the resulting audience must exceed 300 members or the campaign will not serve. That surfaces later as CAMPAIGN_AUDIENCE_COUNT_HOLD in the servingHoldReasons you met in Lesson 01.

Facets are typed

A facet URN is a key; its values are URNs of a facet-specific type. Getting the value type wrong is a common error and the API will tell you, unhelpfully.

Note the range syntax: urn:li:staffCountRange:(51,200). Parentheses and a comma inside the URN id. Any URN parser that splits naively on : and takes the last field will mangle these. The ranges also use 2147483647 — INT_MAX — to mean "no upper limit".

The last two rows are the hook for Lesson 04, and they are strange: two different facets, identical value type. Hold that thought.

Discovering facets and values

Never hardcode value lists. Three endpoints exist for this, and a real integration needs all three:

That third one is easy to forget and impossible to skip. You will store urn:li:title:4; your UI must show "Chief Technology Officer", in the customer's locale. There is no name in the targeting payload — only URNs.

Illegal combinations

A short list of pairs that cannot be AND'ed, plus a short list of facets that only work in include. These are not arbitrary — they mostly prevent contradictory or privacy-sensitive intersections.

Practice

"include": { "and": [  { "or": {      "urn:li:adTargetingFacet:seniorities": ["urn:li:seniority:7"],      "urn:li:adTargetingFacet:locations":   ["urn:li:country:us"]  }}]}

{  "include": { "and": [ { "or": {      "urn:li:adTargetingFacet:locations": ["urn:li:country:de"] }} ]},  "exclude": { "and": [ { "or": {      "urn:li:adTargetingFacet:employers": ["urn:li:organization:1", "urn:li:organization:2"] }} ]}}

"include": { "and": [  { "or": { "urn:li:adTargetingFacet:titles":      ["urn:li:title:4", "urn:li:title:19"] }},  { "or": { "urn:li:adTargetingFacet:locations":      ["urn:li:country:de", "urn:li:country:at", "urn:li:country:ch"] }}]}

{
  "include": { "and": [
    { "or": { "urn:li:adTargetingFacet:seniorities":
        ["urn:li:seniority:7", "urn:li:seniority:8"] }},
    { "or": { "urn:li:adTargetingFacet:staffCountRanges":
        ["urn:li:staffCountRange:(201,500)",
         "urn:li:staffCountRange:(501,1000)"] }},
    { "or": { "urn:li:adTargetingFacet:locations":
        ["urn:li:country:nl", "urn:li:country:be"] }}
  ]},
  "exclude": { "or": {
    "urn:li:adTargetingFacet:followedCompanies":
        ["urn:li:organization:{theirOrgId}"]
  }}
}

Retaining this

1. Say "AND of ORs" once, out loud. That single phrase compresses the whole structure. If you retain nothing else, retain the CNF framing — you can re-derive the JSON shape from it.
2. Write the failure mode down where your team will see it: two facets in one or silently widen the audience instead of narrowing it. No error, real money.
3. Tomorrow, from memory: how do you express "A and B"? How do you express "not (A and B)"? The second question has an interesting answer — you cannot, and knowing why tells you that you have understood the asymmetry.

Sources: targetingCriteria Object, Targeting Criteria Facet URNs, Ad Targeting.