bidsmith drift

plan compares the settings bidsmith models. Everything else on a campaign or ad group isn’t merely left alone — it’s never fetched and never compared. So when a plan says 1447 unchanged, the honest reading is “unchanged on the fields bidsmith knows about,” not “your repo matches Google Ads.”

drift reports the difference. It asks Google which settings each resource has, subtracts the ones plan reads, and then looks at your account to see which of the remainder are actually set. Read-only — it never changes anything.

Why this exists

A max CPV of EUR 0.05 once ran for weeks on a campaign whose own notes said EUR 0.10. Every plan was green, because the field holding that bid wasn’t one bidsmith read. drift is how you find that class of problem without hand-writing GAQL queries and comparing by eye.

Synopsis

bidsmith drift [PATH] [--all] [--format text|markdown] [--detailed-exitcode] [--refresh-state] [--refresh-catalog] [--verbose]

Arguments

Argument	Default	Description
PATH	.	File or directory containing .bid files. Names the account to audit, and scopes the audit to the resources those files manage.

Flags

Flag	Description
--all	Also list unmodelled settings that are unset everywhere. Off by default: most of them belong to campaign types you don’t run, and they bury the ones that matter.
--format <text|markdown>	Output format. markdown renders a table suited to posting as a pull-request comment.
--detailed-exitcode	Exit 2 (instead of 0) when an unmodelled setting carries a value, keeping 1 for errors. Lets CI surface the gap without treating it as a failure.
--refresh-state	Ignore any cached live state and refetch from the API.
--refresh-catalog	Refetch the list of settings Google Ads exposes instead of using the cached copy. The cache lasts a week.
--verbose	Print each outgoing query.

Exit codes

Code	Meaning
0	The audit ran. Without --detailed-exitcode, this covers both “nothing set outside the modelled fields” and “some settings are.”
1	Local validation failed, authentication failed, or the field catalog could not be read.
2	Only with --detailed-exitcode: an unmodelled setting carries a value.

Reading the report

campaign — 16 of 106 settable field(s) modelled
  set on managed resources, never compared:
    campaign.tracking_url_template  11 resource(s)  e.g. google_ads_campaign.brand = "{lpurl}?src=g"
    campaign.ad_serving_optimization_status   3 resource(s)  e.g. google_ads_campaign.gh_yt = "OPTIMIZE"

ad_group — 13 of 36 settable field(s) modelled
  nothing set outside the modelled fields.

`unchanged` in a plan means unchanged on the 88 field(s) bidsmith models, out of 541 the API would let a .bid set.
2 unmodelled field(s) carry a value bidsmith never compares — a plan calling these resources unchanged is not speaking about them.

Rows are grouped by setting, not by campaign, because the question drift answers is which settings fall outside the guarantee. Each row gives the setting’s Google Ads name, how many of your resources have it set, and one example so you can go look.

A setting appearing here doesn’t mean anything is wrong. It means bidsmith isn’t watching it: if someone changes it in the Google Ads UI, no plan will tell you. Whether that matters depends on the setting — campaign.tracking_url_template probably does, a hotel-campaign field on a Search account doesn’t.

Blocks read in part

One line comes before the settings list and is worth more attention than any single row:

campaign — 16 of 106 settable field(s) modelled
  4 of 6 field(s) on 'campaign.network_settings' — the block reads as the whole setting; missing: target_youtube, target_google_tv_network

A setting bidsmith doesn’t model at all is invisible, which at least looks like what it is. A block bidsmith models part of is worse: the .bid carries a network_settings { … } that reads as a complete statement of where ads serve, and for the missing fields it isn’t one. This line is printed whether or not anything on the account has those fields set, because the gap is in the file’s apparent meaning rather than in today’s values.

What “settable” means

The count on each heading excludes fields you couldn’t declare anyway — report-only values like primary_status, effective_cpc_bid_micros, or performance metrics. bidsmith doesn’t guess at these from their names; it reads Google’s own published API description, which marks every output-only field as such.

What to do about a finding

1. Ignore it if the setting is irrelevant to how you run the account.
2. Set it in the Google Ads UI and treat it as out of scope for the repo — but know that nothing will catch a change to it.
3. Open an issue asking for the field to be modelled. That’s the point of the report: schema gaps used to be findable only by accident.

Cost

drift is a heavier read than plan: it makes one pass over your account per batch of unmodelled settings, on top of the live state plan already fetches. It’s a command you run deliberately — before a quarterly review, when adopting an account you didn’t build, or when a number doesn’t match what you expected — not one you put on every pull request.

The list of settings Google Ads exposes is cached for a week, so repeated runs only pay for reading your account.

See also

• bidsmith plan — the diff drift describes the edges of
• bidsmith query — read any Google Ads field directly with GAQL
• bidsmith refresh — pull live values of modelled fields back into .bid files