For agencies that grow D2C brands
What should a decision ledger record for every brand?
A decision ledger should record every change made for a brand: what was tried, why, the one metric it should move, that metric's baseline frozen before the change, a confidence grade and the outcome once measured. Written before the change goes live, it shows the brand what worked and keeps the reasoning when people leave.
24 of 1,424 product items invalid · Moolwan's first ledger entry · baseline frozen 26 Sep 2026, outcome gatheringSearch Console · Product snippets · 26 Sep 2026 · real data
What you'll see in 10 minutes
- Your first brand's report, built from its domain alone.
- Its ad accounts and GA4 connected, one secure sign-in each.
- Every change your team approves logged with what it should move and a frozen baseline.
Why can't an agency say what worked six months ago?
Quarterly review. The founder asks why budget moved from Meta to Google in July. The account manager who made the call left in August, and the reason sits in a chat thread nobody can find. The agency starts the conversation from zero.
A decision without its reason is a cost twice over. The brand can't see what its fee bought, and the agency can't tell a lesson from a lucky week, so it repeats tests that already failed. Seven fields make a decision checkable, now and a year from now.
| Field | What it records | Why it matters |
|---|---|---|
| The decision | What changed, where and when | The brand sees exactly what was done |
| Why | The evidence that prompted it | Every change ties back to a record |
| Expected to move | One metric and its direction, written first | Stops stories written after the result |
| Baseline | The metric's value, frozen on the day | The result is measured from a fixed point |
| Confidence | High, medium or low | Sets how much to read into the result |
| Review date | When the result will be read | Long enough for the metric to settle |
| Outcome | As expected, no clear change, or the opposite | The agency learns, and the brand sees it |
How do you keep a decision ledger that holds up?
Write first, measure later, keep the misses. A ledger is only useful if the expectation is fixed before the result is known. These six habits keep every entry honest enough to show the brand.
Write it before it goes live
Record the change, the reason and the metric it should move before anyone approves it. A hypothesis written afterwards always fits the result.
Name one metric and a direction
Pick the single metric the change should move and whether it should rise or fall. "Improve performance" can't be checked.
Freeze the baseline
Note the metric's value and the date on the day the change goes live, and don't revise it later.
Set a fair review date
Give the metric time to settle: about a week for click-through rate, two weeks or more for conversions, four weeks for rankings. Read too early and you grade noise.
Grade the outcome honestly
Moved as expected, no clear change, or moved the opposite way. Record all three; the misses teach the most.
Share it with the brand
Show the ledger in the report and the portal, so the brand sees what was tried and what was learned.
Each entry's "why" should point at a real record, so anyone can trace every number to its source. Brands who want to follow along between reports can see the same entries in a live brand portal under your agency's name.
How do you write a ledger entry and grade it later?
Log a decision in two minutes. Fill in the change, the reason, the metric and its baseline. The worksheet writes the entry as a pre-registered line, warns if anything would make it uncheckable, and grades the outcome once you add the result.
Write a ledger entry, then grade the outcome
Before the change goes live
One change, one metric, one direction.
- Ready to log: one metric, a direction, a frozen baseline and a review 21 days out, enough for this kind of metric.
ledger entryDewleaf Skincare · 1 Sep 2026. Decision: Replace the top Meta ad with the new UGC ad. Why: the top ad's frequency hit 4.8 and click-through rate fell 31% in 7 days. Expected to move: click-through rate, up from a baseline of 1.4 frozen on 1 Sep 2026. Confidence: medium. Review on 22 Sep 2026, after 21 days.
On the review date
Enter what the metric reads now. Changes within the tolerance count as no clear change.
outcomeOutcome on 22 Sep 2026: click-through rate moved from 1.4 to 1.9 (+36%), as expected.
Free, no login. The starting entry is an illustrative example. Review windows are rules of thumb, not platform rules.
(login required) · ai-dm writes this entry for every change your team approves, and grades it on the review date.
How does ai-dm keep a decision ledger for every brand?
Every approved change is logged before it runs. When your team approves an ad, a post, an article, a review reply or a budget move, ai-dm records what it is, why it was suggested, the metric it should move, the baseline frozen that day and a confidence grade. On the review date it reads the metric again and grades the outcome, and the entry stays in the ledger permanently, whoever is on the account.
5 decisions · 2 as expected · 1 no clear change · 1 opposite · 1 gathering
In this illustrative example, Dewleaf Skincare's ledger shows five decisions from the last 90 days. A new UGC ad lifted click-through rate from 1.4% to 1.9% and negative keywords cut wasted spend from ₹6,420 to ₹1,180, both as expected. A budget move to Performance Max made no clear change, a switch to carousels lowered reach and was reverted, and a landing page change is still gathering results.
The ledger is what makes the report worth reading. The decisions, what they were expected to move and what happened flow straight into the brand's report, so you can send white-label reports that show the misses as plainly as the wins. Nothing changes in a brand's account until your team approves it, and you can check how brand data is protected.
On our own brand. Moolwan's ledger holds 3 decisions so far, with accuracy still gathering. The first: fix flagged product rich-result items, because 24 of 1,424 items were invalid in Search Console; expected to move valid rich-result items; baseline frozen 26 Sep 2026; outcome gathering (Search Console · Product snippets · 26 Sep 2026). We show it before the result is in, because writing it down first is the point. Entry wording is shortened. You can inspect Moolwan's real numbers yourself.
Related pages
Questions agencies ask about decision ledgers
What is a decision ledger in marketing?
A decision ledger is a permanent record of every change an agency makes for a brand: what was tried, why, what it was expected to move, the baseline before it ran and what happened. It turns a year of campaign changes into evidence the brand can read and the agency can learn from.
What should each decision ledger entry include?
The change and where it was made, the evidence that prompted it, the one metric it should move and in which direction, that metric's baseline frozen on the day, a confidence grade, a review date and, later, the outcome. If any of these is missing, the entry can't be checked.
Why write the hypothesis before the change goes live?
Because a reason written afterwards always fits the result. Writing down what a change should move, and by when, before it runs is what makes the outcome honest. It also stops the team from claiming credit for a lift the change never aimed at.
How long should we wait before grading an outcome?
Long enough for the metric to settle. Click-through rate moves within about a week, conversions and cost per conversion need two weeks or more because late conversions keep arriving, and rankings or organic traffic need around four weeks. The worksheet on this page warns when a review date is too early.
Can the brand see the decision ledger?
Yes, if you choose. Ledger entries can appear in white-label reports and in the brand portal under your agency's name, so the brand sees what was tried, why and what it moved. Nothing in them names ai-dm.
Does a good outcome prove the change caused it?
No. Other things change on the same dates, from the season to a rival's offer. That's why each entry carries a confidence grade and why ai-dm lists what else moved in the same period. ai-dm doesn't build attribution models; it records what was expected and what happened.
What does the decision ledger in ai-dm cost?
It's included on every plan, with no plan to pick. Sign up, pay as you go, and every change your team approves is logged from the first day. Billed in USD and INR. Per-brand plans or pay as you go. A brand's data is deletable on request.
No plan to pick. Billed in USD and INR. A brand's data is deletable on request.