> ## Documentation Index
> Fetch the complete documentation index at: https://docs.interchange.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Selling doctrine
> Decide when your storefront pitches, counter-pitches, or passes on a brief — and test the change before it ships
A storefront that answers everything answers nothing.
Before your Merchandising Agent composes anything, it makes the judgment every
good sales organisation makes first: **is this brief ours?** Your **selling
doctrine** is how you answer that. It rides your [Playbook
instructions](/v2/storefront/operating-instructions/overview) — same versions,
same activation, same history — because it is those instructions grown, not a
separate setting to keep in sync.
## The three answers
| Disposition | What it means | What you get |
| ----------------- | --------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Pitch** | The brief fits. Argue it in full. | Products ship as they do today; the composed pitch lands on the proposal pass |
| **Counter-pitch** | The literal ask doesn't fit, but a credible reframe does. | Products ship as they do today; the pitch carries the reframe in its honest counter, with the gap named from your own catalogue |
| **Pass** | Neither fits, and no reframe is credible. | The composed proposal is withheld. Passthrough inventory and already-finalized proposals still ship. A short, branded decline naming the mismatch and leaving the door open is recorded for you |
A pass is a composed outcome, not a failure. Passing well is
relationship-building, and buyer agents that run managed evaluation score
sellers on response *quality*, not response volume — so a doomed bid costs you
standing that a graceful decline protects.
**A pass genuinely withholds your answer, and it looks like a no-fit.** When
your doctrine declines a brief, the buyer's agent gets the same response your
storefront already returns when nothing matches: no composed proposal, and no
explanation. It cannot tell "declined" from "nothing matched" — that is
deliberate. AdCP has no first-class "no bid, with reasons" message today, so
rather than invent one we reuse the answer buyers already understand.
Only the *composed* proposal is withheld. Passthrough inventory and proposals
you finalized on an earlier turn are a different lane your doctrine never
judged, so they still ship — withholding them would make a decline detectable
by what went missing.
The decline itself never leaves your side. It is written for *you*: your record
of why your storefront walked away, readable on the proposal pass alongside
every other composed outcome, with a receipt back to the doctrine version that
decided.
## What decides it
Two numbers and a posture, set alongside your Playbook prose:
| Setting | What it means |
| -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `minimumCategoryFit` | The share of the brief's stated verticals and channels your composed catalogue must actually serve before you pitch in full. `0` pitches everything; `1` pitches only a brief you cover completely. |
| `counterAppetite` | How far below that bar you still answer, with a counter-pitch instead of a pass. `0` never counters — a brief that misses the bar is declined. `1` counters anything your catalogue touches at all. |
| `floorPosture` | What to do when the buyer states a budget and none of your composed lines could be priced. `hold_floor` declines rather than negotiate from nothing; `flex_floor` counters and invites a re-brief. |
**Category fit is measured against this pass's own output** — the products the
agent actually composed and the inventory that fed them — never your whole
catalogue. Inventory you did not offer for this brief did not serve it.
A brief that names no vertical or channel we can compare is not a poor fit; it
is an unmeasured one, and it pitches. That is the common shape of a short
brief, and it is why turning doctrine on changes nothing for most traffic.
Because a pass withholds the composed proposal, a doctrine set too tight will
stop your storefront answering briefs it was winning. Run the briefs you care about
through a brief test before you activate a version — and if answers dry up,
activating the previous version puts them back.
Your Playbook **prose** keeps doing what it always did: it shapes the argument,
the register, and the judgment inside a proposal. The numbers above decide only
whether a proposal happens at all — because whether a buyer gets an answer is
the one question that should never be decided differently on two identical
runs.
## Named starting points
Adopt one, edit it, or write your own.
| Variant | The strategy |
| ------------------------ | ----------------------------------------------------------------------------------------------------- |
| `premium_scarcity_house` | Qualify hard, pass often, hold price. Pitches at 0.75 fit, counters down to 0.60, declines below. |
| `volume_partner` | Counter nearly everything, lead with packaging math. Pitches at 0.20 fit, counters down to 0.01. |
| `consultative` | Counter first, lean on the honest counter, invite a re-brief. Pitches at 0.60, counters down to 0.06. |
**A storefront that has never set a doctrine has no qualification gate at all.**
It pitches every brief it can compose products for, exactly as it did before
doctrine existed — nothing is passed or countered on your behalf. Doctrine only
starts deciding once you set one.
Such a version reads back as `house_default` with `isDefault: true`, and the
numbers shown for it are display and starting values, not active thresholds:
they are what you would begin from if you adopted them, not a bar your briefs
are being measured against today. You cannot adopt `house_default` by name, so
"no doctrine set" has exactly one spelling. To return to it, clear the doctrine
or activate a version that never had one.
Adopt one by name:
```bash theme={null}
curl -X POST https://api.interchange.io/api/v2/storefront/operating-instructions \
-H "Authorization: Bearer $SCOPE3_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"content": "Lead with the homepage takeover for launch briefs.",
"doctrine": { "variant": "premium_scarcity_house" },
"activate": true
}'
```
Or state your own numbers:
```bash theme={null}
curl -X POST https://api.interchange.io/api/v2/storefront/operating-instructions \
-H "Authorization: Bearer $SCOPE3_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"content": "Lead with the homepage takeover for launch briefs.",
"doctrine": {
"thresholds": {
"minimumCategoryFit": 0.7,
"counterAppetite": 0.5,
"floorPosture": "hold_floor"
}
},
"activate": true
}'
```
Thresholds that match no named variant read back as `custom`. So do a named
variant's numbers after you edit one of them — the label always describes the
numbers actually in force, never the strategy you started from.
## Test it before it ships
Doctrine is versioned and activated exactly like the rest of your Playbook, so
the loop is: draft a version, test a brief against it, then activate it once it
reads right.
1. Author a new version with the doctrine you are considering. Leave
`activate` off to keep it out of production while you look at it.
2. Test that version by name — pass its number as
`operatingInstructionsVersion` to `test_storefront_brief`, or ask Murph to
test a brief against it. Omit it and the active version is used, as before.
3. Read the result. It carries a `bidDecision` — the disposition, which rule
decided it, the measured category fit, and the doctrine version behind it —
and `composedPitch`, the argument or the branded decline the buyer would
actually have received. The version that was evaluated is reported back, so
you are never guessing which one you just tested.
4. Activate when it reads right.
You can only test a version your own storefront authored; the version is
re-read and re-authorized before anything composes under it. Nothing is
activated by testing — your live active version is untouched until you
activate it yourself.
Changed your mind? **Activating the previous version restores the previous
behaviour**, doctrine included. Nothing is overwritten; every version stays
readable.
Brief tests are dry runs: they create no media buy, persist no products, never
turn on transacting, and leave nothing on your exchange history. The narrative
is composed for you to read and then discarded.
## Reading a decision after the fact
Every composed pass records the disposition it reached, which rule reached it,
the measured fit, and the doctrine version that decided — and the disposition
carries a receipt back to the Playbook version behind it, exactly like the
receipts behind every sentence of a pitch.
That is what makes "why did we pass on this brief?" answerable months later,
without re-deriving it against a doctrine you have since edited.
## What doctrine cannot do
Doctrine governs judgment, not truth. It changes what your agent argues, how
hard, and when it walks away. It cannot mint a fact: every sentence in a pitch
or a decline still sources from the brief, your inventory, or your Playbook,
and a sentence with no source is dropped before anyone reads it.
So a doctrine that says "counter everything" does not produce a counter-pitch
claiming inventory you do not have. It produces a counter-pitch built from the
inventory you do.