Attribution Reconciler · Marketing Tool Stack

Two attribution reports never agree, and the meeting stalls arguing about which tool is right. This reconciles the actual numbers you paste, channel by channel, so you walk out knowing what to trust and what to fix.

What this isA prompt that takes two attribution views that disagree and reconciles them line by line instead of lecturing you on attribution theory.

You give itTwo per-channel datasets (for example last-click vs multi-touch, or ad platform vs CRM) and the decision you are about to make.

You get backA per-channel delta table, the likely cause behind each gap, which view to trust for your decision, the gaps that cannot be explained from the data, and a fix list.

Paste this into Claude, ChatGPT, or any capable model. Fill in the bracketed parts and it does the rest.

You are reconciling two attribution datasets that disagree. You are not teaching me attribution theory and you are not recommending an attribution model. Your only job is to explain why these two specific datasets differ and tell me what to do about it.

Hard rules:
- Work ONLY from the two datasets I paste. Do not pull in benchmarks or "typical" figures unless I provide them.
- Quote the real numbers. Every gap you name must show both source values and the actual delta (absolute and percent).
- Never invent, estimate, or "smooth" a conversion count that is not in the data.
- When a gap cannot be explained from what I gave you, say so plainly and name the one field or cut you would need to close it. Do not fill the hole with a guess.
- If the two datasets are not comparable (different date ranges, conversion definitions, currencies, channel groupings), STOP and flag that first. Do not reconcile mismatched data and pretend the deltas are real.

DATASET A
  Source / system: [e.g. Google Ads, last-click, platform-reported]
  Date range: [START to END]
  Conversion event counted: [e.g. form fill, MQL, closed-won]
  Lookback / attribution window: [e.g. 30-day click, 1-day view]
  Per-channel data (channel, conversions, revenue, spend if available):
  [PASTE TABLE A]

DATASET B
  Source / system: [e.g. CRM / HubSpot / Salesforce, multi-touch]
  Date range: [START to END]
  Conversion event counted: [e.g. closed-won opportunity]
  Lookback / attribution window: [e.g. 90-day, first-touch + last-touch]
  Per-channel data (channel, conversions, revenue, spend if available):
  [PASTE TABLE B]

THE DECISION I AM ABOUT TO MAKE: [e.g. cut paid social budget 30%, reallocate to events, defend channel ROI to finance]

Before you reconcile, normalize what you can: map channel names across both datasets (state every mapping), and confirm date ranges and conversion definitions line up. List any channel in one dataset but not the other.

Then return exactly these sections:
1. COMPARABILITY CHECK. One line per dimension (date range, conversion definition, currency, channel mapping, lookback window). Mark each MATCH or MISMATCH. If anything material is a mismatch, say how much trust that removes from the reconciliation.
2. RECONCILIATION TABLE. One row per channel. Columns: Channel | A value | B value | Delta (abs) | Delta (%) | Direction. Sort by absolute delta, largest first. Add a TOTAL row.
3. LIKELY CAUSE PER GAP. For each channel with a material delta, give the single most likely cause and the runner-up, and say which number in the data points to it. Draw from: cookie/signal loss, view-through credit, self-reported platform bias, dedup window difference, lookback mismatch, first-touch vs last-touch split, offline/sales-assisted revenue, tracking gap. If a gap fits none from the data alone, label it UNEXPLAINED and name the field you would need.
4. WHICH VIEW TO TRUST FOR THIS DECISION. Given the decision I named, say which dataset to act on per channel and why. Be specific. Do not hedge into "it depends".
5. FIX LIST. 3 to 6 concrete fixes ranked by impact on closing these gaps. Each: what to change, which gap it closes, roughly how long it takes. No generic "improve your tracking".
Keep it tight. No attribution history lesson.

When to use it

Use it the moment two reports for the same period disagree and someone is about to make a budget or target call on the louder number. Best right before a spend review or a finance reconciliation.

How to use it

Paste both per-channel tables with their source, date range, conversion definition, and lookback window, then name the decision you are about to make. Read the comparability check first; if the data is not comparable, the deltas are noise.