Daily optimization digest

Each morning, for the clients you've opted in, Adsidian reviews the last 7 and 28 days of performance and puts together a short list of concrete changes worth making. This article explains what it looks for, how to turn it on, and what happens when you click Apply.

What it is

Every day, Adsidian looks at recent performance for each opted-in client — pacing, cost per result, conversions, ad review status, and more — and produces a handful of specific, described recommendations: raise this budget, pause that ad set, rewrite this ad, or just a note explaining why a campaign was left alone.

Nothing is applied automatically. Every recommendation sits there until someone reviews it and clicks Apply or Dismiss — a second set of eyes that never sleeps, not an autopilot.

Turning it on

The digest is off everywhere until you turn it on, at three levels:

1. Agency Settings → Automation — a master switch for your whole agency. Off by default.
2. The client's Automation tab — a per-client switch, off even when the agency switch is on.
3. Campaign scope — once a client is enabled, pick "all active campaigns" or check the specific campaigns you want reviewed.

All three have to say yes for a campaign to be reviewed. Enable a client but check no campaigns, and that client simply gets no digest — nothing runs by chance.

What it looks for

The digest checks each in-scope campaign against a set of plain rules, then writes up the results in plain English:

• Budget-limited winners. A campaign performing well that's losing opportunity to its budget cap gets a recommendation to raise the budget, up to the daily limit you've set.
• Expensive campaigns. A campaign running well above its cost target gets a recommendation to lower the budget.
• Spend with nothing to show for it. An ad set or ad that's spent meaningfully without a single conversion, while similar ones are converting, gets flagged to pause.
• Disapproved or limited ads. Anything a platform has flagged gets surfaced with the policy issue, so you know to fix it.
• Tired creative. An ad whose click-through rate has dropped off, or that's simply been running a long time, gets flagged for a refresh.
• "Watch" notes. For campaigns deliberately left alone — still in a platform's learning phase, or changed recently — a short note explains why, so a quiet day doesn't read as "nothing was checked."

On Google, "budget-limited" now uses Google's own impression share lost to budget, rather than an internal pacing estimate. On Meta, the learning-phase check reads the ad set's own learning status directly, so a campaign still learning is left alone rather than judged on early, noisy numbers.

Every one of these checks can be tuned for your agency and your clients — see Tuning the digest below.

Tuning the digest

The digest ships with sensible defaults, but almost everything about how it behaves can be adjusted at three layers, each narrower than the last:

1. Built-in defaults — what the digest uses if nothing else is set.
2. Agency Settings → Digest defaults — your agency's defaults, applied to every client unless overridden.
3. Client → Automation tab — per-client overrides. Any field can be set to "Inherit" to fall back to the agency default, and the tab always shows where each value is currently coming from.

Sensitivity, per recommendation type. Budget increases, budget decreases, pauses, resumes, ad set pauses, ad pauses, creative refresh, disapproval fixes, structure notes, and watch notes can each be set independently to Off, Conservative, Standard, or Aggressive. Conservative asks for stronger evidence before it speaks up; Aggressive flags things earlier. Advanced thresholds are available for exact numbers and override whichever preset is selected.

Budget floors and ceilings, set per client and per campaign, are a hard boundary — the digest will never recommend a budget outside them.

Cadence and windows. Run the digest daily, weekdays only, or weekly on a day you choose. For lower-volume accounts, widen the lookback window the digest reasons over — 7, 14, or 28 days for budget decisions, 14, 28, or 56 days for pause decisions — so a quiet week doesn't read as a trend.

Minimum conversions. Set a threshold below which a campaign is left alone with a watch note rather than judged — there's rarely enough signal in a handful of conversions to act on.

Levels. Choose whether the digest reasons about campaigns only, or drills into ad sets and ads too.

Plain-language rules from your playbook and brief. Sentences already in your agency playbook or a client's brief are recognized and applied automatically, no extra setup needed — things like "Never pause anything before 14 days on Meta," "Max budget change of 10%," "Don't recommend budget increases," "At least 10 conversions before judging," "No ad-level recommendations," "Never raise Brand Search above $150," or "Target CPA $40." The Automation tab lists which sentences were picked up. Playbook rules apply agency-wide; brief rules apply to that one client; an explicit setting in the Automation tab always wins over a rule pulled from text.

Suggested tweaks. Dismiss the same type of recommendation as "disagree" several times for a client, and the digest notices the pattern and proposes a specific settings change — for example, raising the cost-per-result ratio that triggers a budget cut. Apply or ignore the suggestion in one click, right from the Automation tab.

Automatic changes (auto-apply)

For a narrow set of low-risk recommendation types, Adsidian can apply the change itself instead of waiting for a click. It's off everywhere by default, and turning it on takes three separate switches, all at once:

1. Agency Settings → Automation → Allow automatic changes — the agency-wide master switch. Off by default, and nothing below matters until it's on.
2. The client's Automatic changes card — a per-client switch, off even when the agency switch is on.
3. A per-campaign Auto toggle, alongside the recommendation types it's allowed to apply for that campaign: ad pauses, ad-set pauses, budget decreases, and budget increases, and campaign pauses. Resumes and creative changes are never automatic — those always wait for a click, no matter how the toggle is set.

Earning the toggle. A recommendation type doesn't become available for a client the day auto-apply is enabled — it has to earn it first. Once 10 recommendations of that type have been applied for that client and measured with a positive result, the type unlocks. An admin can also override the wait and turn a type on directly, with an explicit acknowledgement that they're skipping the track record.

The same limits, plus its own. An automatic change runs through every guardrail, floor, and ceiling a manual Apply would — the daily budget-swing cap, the 7-day cooldown, the learning-phase exclusion, all of it (see Guardrails below). On top, auto-apply adds two limits of its own: at most one automatic change per campaign per day, and a per-client daily cap overall — 3 by default, adjustable in the Automatic changes card.

What you'll see. An automatic item shows up wherever recommendations normally do, tagged Applied automatically instead of Apply/Dismiss, and it's listed first in the morning email rather than buried under items still waiting on a decision. Every automatic change carries an Undo button for 24 hours, and is logged with who enabled auto-apply for that client and campaign — never a mystery to trace back.

The kill switch. Flipping either the agency switch or the client switch off stops automatic changes immediately — nothing already queued slips through.

Bids and bidding strategies are never touched, automatically or otherwise — that stays out of scope entirely, same as everywhere else in the digest.

Guardrails

The digest is built to be conservative by design:

• Budget changes are capped at a maximum swing per day (20% by default, adjustable per client).
• Nothing is recommended for a campaign changed in the last 7 days — give a change time to show results first.
• Campaigns still in a platform's learning phase are left alone.
• At most 8 recommendations per client per day, so the list stays something you can actually read.
• It never touches bids or bidding strategy — there's no bid-management feature here, on purpose.
• It never recommends turning a disapproved ad back on.
• Any change to ad copy always goes through the AI assistant and Adsidian's compliance check, never a one-click apply.

Applying and dismissing

Each recommendation gets one of a few actions:

• Apply — a single click that runs the change through the same approval and audit trail as any change you'd make through the assistant. Nothing skips review it would otherwise require.
• Apply all safe — a batch action for pauses and budget decreases, the lowest-risk kinds, so you don't click through them one at a time.
• Dismiss — pick a quick reason (not now, disagree, already handled, bad data, other). Dismissals help tune future recommendations.
• Open in assistant — for creative refreshes and structural suggestions, which need a conversation rather than a single click.

Where you'll see it

• Client Overview — a "Today's recommendations" card with a summary and the top items.
• Recommendations inbox — an agency-wide view across all your clients, filterable by client, kind, or status.
• Morning email — optional, per client, if you'd rather get it in your inbox.
• The AI assistant — the assistant knows about today's digest for a client, so you can ask it "why did you recommend this?" in chat.

Track record

Every applied recommendation is checked again a week later to see whether it helped, broken down by type — how budget increases have performed versus pauses, for example. That track record is also what unlocks automatic apply for a recommendation type, once it's built up enough of a positive history.

What it doesn't do

• It doesn't make changes on its own, unless you turn that on. Everything here requires a human click by default. See Automatic changes (auto-apply) above for the opt-in that lets a narrow set of low-risk changes apply themselves.
• It doesn't manage bids. Bid and bidding-strategy changes are out of scope entirely.
• It isn't a replacement for your own review. It's a daily prompt toward the campaigns most worth your attention — the judgment calls are still yours.