For agencies that grow D2C brands

What is an evidence trail, and why do brands ask for one?

An evidence trail links every number in a brand's report to the raw record it came from: the platform, the account, the date range and how the metric was counted. Anyone can click a figure and check it. It turns "trust us" into "check it", and it shows when two platforms disagree and why.

344k impressions in Google's AI results · Moolwan · one click from the record behind itGoogle Search Console · Generative AI features · 25 Jun–23 Sep 2026 · real data

What you'll see in 10 minutes

  1. Your first brand's report, built from its domain alone.
  2. Its ad accounts and GA4 connected, one secure sign-in each.
  3. Every number in the report with its source, date range and a link to the record behind it.

(login required) · No plan to pick. Billed in USD and INR. Nothing changes without your approval.

Why do brands stop trusting the numbers in agency reports?

Month-end review. The report says ROAS rose to 3.4×. The founder opens Ads Manager on her phone and sees 2.9×. Nobody in the room can say where 3.4× came from, which dates it covers or how it was counted.

One untraceable number taints the rest. After that meeting, every figure in every report gets questioned, and the brand starts asking for logins to check things itself. A number that carries its own source answers the question before it's asked. These are the parts a traceable number carries.

What a traceable number carries, with an example and why each part matters
PartExampleWhy it matters
Source and accountSearch Console · the brand's propertySays who counted it
What the metric countsImpressions in the Generative AI features reportThe same word means different things on different platforms
Date range25 Jun–23 Sep 2026Stops two people comparing different weeks
Counting setting7-day click, 1-day viewExplains most gaps between platforms
Record linkThe report or record ID it came fromLets anyone check it without asking
When it was capturedCaptured 26 Sep 2026, never overwrittenShows the number wasn't changed later

How do you build an evidence trail into a brand's report?

Attach the source to the number, not to the appendix. A sources page at the back of a deck rarely gets read. A short source line under each figure gets read every time. Work through these six habits and every number in the report can be checked on the spot.

  1. Name the source under every number

    Write the platform and account each figure came from directly beneath it, such as Meta Ads Manager for the brand's ad account or the brand's Search Console property. A number without a source is an opinion.

  2. State the date range and time zone

    Give the exact dates, not "last month", and note the account's time zone. Most disagreements about a number turn out to be about which days it covers.

  3. Define each metric once

    Say what the metric counts: purchases or all conversions, users or sessions, impressions in which report. The same word means different things on different platforms.

  4. Write down the counting setting

    Note the attribution setting or conversion window behind any ad platform figure. It explains most of the gap when two platforms disagree.

  5. Keep the raw export, unchanged

    Save the original export or record for every reported figure and never overwrite it. If a number is questioned months later, you can show exactly what the platform said at the time.

  6. Link each number to its record

    Give every figure a link or record ID that opens the source it came from. Anyone at the brand can then check it without asking you.

When two platforms report different figures for the same sales, the evidence trail shows why; to explain the gap to the brand, reconcile GA4, Meta and Google Ads. When a traced number has moved, the records point to the cause, and you can find why the brand's numbers moved.

Could the brand check this number in your report?

Test one figure before the next review. Pick a number from a brand's latest report and fill in what you know about it. The worksheet scores how traceable it is, lists what's missing and writes the source line to put under it.

Check whether a number is traceable

Leave a field empty if the report doesn't say. Each of the six parts counts once.

4 of 6parts of a traceable number
Partlytraceable · the brand can't check it yet
2parts missing
  • counting setting
  • record link or ID

source line to put under the number3.4× · Meta ROAS, purchase value ÷ spend · Meta Ads Manager · Dewleaf Skincare India · 1–30 Sep 2026

what to fixAdd the counting setting and the record link or ID, and anyone at the brand can check this number against the platform themselves.

Free, no login. The starting figure is an illustrative example. "Captured on" is optional and not scored, but it shows the number wasn't changed after the fact.

Trace a brand's numbers

(login required) · In ai-dm, every number carries its source line and record link automatically.

How does ai-dm keep an evidence trail for every brand?

Every record kept as it arrived, every number linked to it. ai-dm stores each signal from a brand's connected accounts exactly as the platform sent it and never overwrites it. Every metric in the workspace, the reports and the portal clicks through to that raw record, with its source, account, date range and counting setting. The evidence trail is included on every plan.

Moolwan · evidence traileach number → its source
344k impressions in Google's AI resultstracing…record linkedSearch Console · Generative AI features · 25 Jun–23 Sep 2026
1,400 products ready for rich resultstracing…record linkedSearch Console · Product snippets · 26 Sep 2026 · 24 of 1,424 items invalid
1,380 valid merchant listingstracing…record linkedSearch Console · Merchant listings · 26 Sep 2026 · 1 invalid
#2 of 17 in the Maps local packtracing…record linkedai-dm · Local presence · geo-grid · Sep 2026
2.13× Meta ROAS on ₹28,064 spendtracing…below 4× benchmarkai-dm · Paid ads · 30 days to 26 Sep 2026
368 new users from ChatGPTtracing…record linkedGA4 · First user source · 29 Jun–26 Sep 2026

6 numbers · 6 sources · strong and weak shown the same way

Real Moolwan data · bot traffic excluded

These six Moolwan numbers are real, each with its source: 344,000 impressions in Google's AI results (Search Console, 25 June–23 September 2026), 1,400 products ready for rich results with 24 of 1,424 items invalid and 1,380 valid merchant listings (Search Console, 26 September 2026), #2 of 17 in the Maps local pack (ai-dm, September 2026), a 2.13× Meta ROAS on ₹28,064 below a 4× benchmark (ai-dm, 30 days to 26 September 2026) and 368 new users from ChatGPT (GA4, 29 June–26 September 2026).

Decisions get the same treatment as numbers. When your team acts on a figure, the decision, the reason and the frozen baseline it will be judged against go on record, so you can show every decision and its outcome to the brand later. Each brand's records stay private to that brand, nothing changes in its accounts without your approval, and you can check how brand data is protected.

On our own brand. Every Moolwan number on this site carries its source and date, including the ones that aren't flattering, like the 2.13× Meta ROAS. You can inspect Moolwan's real numbers and follow each one to where it came from.

We don't build attribution models. An evidence trail shows where each number came from, not which channel caused a sale. ai-dm never merges platforms into one "true" figure and reads no order or revenue data from Shopify.

(login required) · Rather talk first? WhatsApp the founder

Questions agencies ask about evidence trails

What is an evidence trail in marketing reporting?

An evidence trail links every number in a report to the raw record it came from: the platform, the account, the date range, how the metric was counted and when it was captured. Anyone reading the report can click a figure and check it, so the numbers no longer rest on the agency's word.

Why do brands ask agencies for the source of every number?

Because a number they can't check is a number they can't defend to their own team or investors. When one figure in a report turns out to be wrong or unexplained, every other figure becomes suspect. A source line under each number answers the question before it's asked.

What should the source line under a number include?

The metric's name and what it counts, the platform and account, the exact date range, any attribution or counting setting, and a link or ID for the record behind it. The worksheet on this page writes that line for you from a figure in your own report.

Does ai-dm change or overwrite a brand's raw data?

No. ai-dm stores every record exactly as it arrived and never overwrites it. Derived numbers, baselines and reports are built from those records and can be rebuilt from them, so any figure can be traced back to what the platform reported at the time.

Can a brand see the evidence trail behind its report?

Yes. Every figure in the reports and portal your team shares keeps its source line and record link, under your agency's name. Your team decides what goes into each report; the sources travel with the numbers.

Does ai-dm read the brand's Shopify revenue or orders?

No. ai-dm reads no order, revenue or customer data from Shopify. Its numbers come from the brand's connected marketing accounts, GA4 and its own tracking, each shown with its source.

What does the evidence trail cost?

The evidence trail is included on every plan. There's no plan to pick: sign up, pay as you go, and connect a brand's accounts with one secure sign-in each. Billed in USD and INR. Per-brand plans or pay as you go. A brand's data is deletable on request.

Trace a brand's numbers

No plan to pick. Billed in USD and INR. A brand's data is deletable on request.

Win the why.™

Every number that moves has a reason. Dashboards show what happened; they rarely show why. Without the why, teams guess, act late and lose trust. ai-dm exists to win the why for every brand, every day: to find the reason behind every change, prove it with evidence, recommend the next move and check that it worked. Then it starts again, because the why keeps changing. When you win the why, you stop guessing and start growing. That is why we exist.

Gather. Analyse. Detect. Alert. Recommend. You act. Monitor. Learn and repeat. Always on.

Every number, one click from its source.

Talk to the founder