Copy This Prompt to Diagnose Why Revenue Dropped

Revenue falls 14%. You ask ChatGPT, “Why did my revenue drop?” It suggests marketing, pricing, competition, and churn. That is not a diagnosis.

This chatgpt revenue drop analysis prompt takes a different approach. Supply acquisition, conversion, customer-value, retention, and payment data. Identify changes, unknowns, and next checks.

Narrow the investigation before changing the business.

Why Revenue Drops Are Hard to Diagnose

Revenue is an outcome. Transaction revenue can reflect order volume multiplied by order value. Volume may reflect traffic times conversion. Other models differ.

Subscription businesses track new MRR, expansion, contraction, churn, and payment failures. Services track qualified leads, close rate, deal size, and timing.

Define your revenue measure: recognized, invoiced, collected, or recurring. Stripe explains why MRR is distinct from accounting revenue and immediate cash collection.

A weekly AI KPI review can alert you to revenue drift. This prompt identifies the changed driver.

When to Use This Prompt

• Use it after a meaningful decline in revenue, orders, new customers, or recurring subscriptions.
• Use it when several metrics have worsened and you need to prioritize checks.
• Do not use it as a replacement for bookkeeping, a recurring business review, or a revenue forecast.
• Pause when the comparison periods or metric definitions cannot be reconciled.

The TRACE Revenue Diagnosis Framework

TRACE enforces the right order:

1. T — Total Change: Define the metric, current and previous periods, absolute decline, percentage decline, and source.
2. R — Revenue Drivers: Build the relevant acquisition, conversion, customer-value, retention, and payment branches.
3. A — Actual Evidence: Compare verified metrics and check missing or inconsistent records.
4. C — Causes to Check: Propose hypotheses only after identifying the observed changes.
5. E — Experiment / Next Check: Choose a small verification step or reversible test.

For example, $18,400 to $15,900 across equal periods means -$2,500, or -13.6%. The cause remains unknown.

What Data You Need Before Asking ChatGPT

Use billing, accounting, or verified sales records as your revenue source of truth.

• Periods: Start and end dates, equal duration, comparable weekdays, and relevant seasonal events.
• Revenue: Both totals, the same definition, currency, and source.
• Acquisition: Visitors, leads, trials, or qualified opportunities.
• Conversion: Orders, closed deals, or trial-to-paid outcomes with denominators.
• Customer value: AOV, account value, deal size, plan mix, or discounts.
• Retention and payments: Churn, cancellations, refunds, downgrades, and payment failures.
• Known events: Confirmed campaigns, outages, launches, price changes, or stock problems.

Prefer equal 28-day windows. Normalize different lengths. Consider year-over-year or cohort comparisons for seasonality. Google Analytics documents period and weekday-matched comparisons.

Check duplicates, delays, tracking changes, and one-offs. Share only aggregated or redacted data, never identities, card details, or confidential contracts.

Revenue Drivers by Business Model

Business Model	Volume Driver	Conversion Driver	Revenue per Customer	Retention Driver	Other Key Checks
SaaS / subscription	Trials or signups	Trial-to-paid	ARPA and plan mix	Churn, contraction	Expansion, failed payments
Ecommerce	Sessions	Purchase conversion	Average order value	Repeat purchases	Refunds, product mix
Consulting / agency	Qualified leads	Proposal close rate	Average deal size	Client renewals	Capacity, project delays
Creator / digital products	Qualified audience	Checkout conversion	Revenue per sale	Renewals, repeat buyers	Launch timing, refunds
Local service	Bookings or inquiries	Booked-to-completed	Revenue per job	Repeat customers	Availability, no-shows

These branches are not universal equations. Separate MRR, recognized revenue, and cash. Shopify also separates sales, order value, sessions, and conversion metrics in its analytics reporting.

The Complete ChatGPT Revenue Drop Analysis Prompt

Copy the prompt. Replace fields with verified numbers; mark gaps “Unknown.”

ROLE: Act as an evidence-first small-business revenue analyst. Identify measurable drivers, not imagined root causes.

BUSINESS CONTEXT: Business type [SaaS / ecommerce / services / agency / creator / course / local service / other]. Business model [subscriptions / transactions / projects / other]. Main offer [offer]. Revenue metric and definition [MRR / recognized / invoiced / collected / net sales]. Currency [currency].

PERIODS AND SOURCE: Current period [start–end]. Comparison period [start–end]. Current revenue [amount]. Previous revenue [amount]. Source of truth [system and report]. Metric definitions [definitions]. Known calendar or reporting differences [facts or unknown].

DRIVER DATA: For each available metric, provide previous value, current value, unit, definition, and source: traffic [values]; leads or trials [values]; qualified leads [values]; conversion numerator and denominator [values]; new customers or paid signups [values]; average order/deal/account value [values]; churn or cancellations [values]; refunds [values]; failed payments [values]; expansions/downgrades [values]; channel or product mix [values]. Mark missing metrics “Unknown.”

KNOWN BUSINESS CHANGES: Confirmed pricing, offer, product, channel, marketing, operational, or tracking changes [dated facts]. Known seasonal factors [facts]. Main diagnostic question [question].

RULES:

1. Identify the business model first. Adapt the revenue-driver tree to it; do not impose SaaS metrics on other models.
2. Ask up to five questions if revenue definitions, sources, or comparable periods are missing.
3. Calculate absolute and percentage changes. Show formulas, denominators, units, and any normalization. Never invent inputs.
4. Check gaps, duplicates, outages, definition changes, delays, and source conflicts.
5. Label every statement OBSERVED, CALCULATED, HYPOTHESIS, or UNKNOWN. Treat simultaneous events as context, not proof of causation.
6. Quantify contributions only with valid decomposition. Otherwise label major, partial, minor, or insufficient evidence.
7. Distinguish first-order drivers from unverified root causes.
8. Rank by evidence, magnitude, revenue proximity, verifiability, and impact. Confidence: High, Medium, or Low.
9. Suggest no more than three next checks. Recommend action only where evidence supports it; favor small reversible tests. Do not make pricing, staffing, channel, or spending decisions for me.
10. Use redacted, aggregated data only.

OUTPUT — STRUCTURED DIAGNOSTIC REPORT:

1. Revenue Change: exact comparison, source, formula, and period caveats.
2. Data Quality Check: reliable, inconsistent, missing, and disputed inputs.
3. Revenue Driver Breakdown: table with driver, previous, current, change, evidence class, and contribution assessment.
4. Strongest Observed Changes: three to five findings at most.
5. What the Data Does Not Show: unknowns and unsupported explanations.
6. Ranked Hypotheses: table with supporting evidence, contradicting evidence, missing data, qualitative confidence, and next check.
7. Next Checks: maximum three, each able to support or reject a hypothesis.
8. Possible Actions: only proportionate, reversible actions justified by evidence.
9. Decision Summary: Investigate, Test, Hold, Escalate, or Monitor; give one reason and identify the human decision owner.

FINAL QUALITY CHECK: Recheck arithmetic, units, period alignment, metric definitions, evidence labels, attribution claims, and proposed checks. If the data cannot explain the revenue movement, say so explicitly.

How to Personalize the Prompt

Prompt Field	What to Enter	Example
Business model	How money is generated	Monthly SaaS subscriptions
Current period	Exact start and end	Most recent 30 complete days
Comparison period	Equivalent dates or justified baseline	Previous 30 complete days
Revenue	Two verified totals, same definition	MRR: $14,800 to $12,900
Volume	Relevant activity with counts	510 to 418 trial signups
Conversion	Both numerators and denominators	82/510 and 67/418
Customer value	AOV, ARPA, or deal size	$92 to $91 ARPA
Retention	Churn and downgrade data	3.1% to 3.8% churn
Payments	Failed payment and refund counts	7 to 9 failed payments
Business changes	Verified events with dates	One acquisition channel changed
Data source	Report or system for each metric	Billing export and analytics
Missing metrics	Explicit gaps, not estimates	Expansion MRR unknown

Include denominators and consistent reporting cutoffs.

Example: Solo SaaS Founder With Signups Down 18%

Illustrative example only. A solo founder sells a monthly subscription product. MRR and trials fell. One channel changed, with no supporting performance detail. Prices stayed unchanged.

Verified Input in This Illustration	Previous 30 Days	Current 30 Days	Movement
MRR	$14,800	$12,900	-$1,900 (-12.8%)
Website visitors	9,600	7,850	-18.2%
Trial signups	510	418	-18.0%
New paid signups	82	67	-18.3%
Trial-to-paid rate	16.1%	16.0%	-0.1 percentage point
ARPA	$92	$91	-1.1%
Customer churn	3.1%	3.8%	+0.7 percentage point
Failed payments	7	9	+2

Customized prompt excerpt: “Monthly SaaS. Compare the two equal 30-day periods above. MRR fell from $14,800 to $12,900. Diagnose only from these supplied figures. Separate observed acquisition changes from hypotheses about the channel shift. Do not claim an exact MRR contribution without the full MRR bridge.”

Sample AI output — OBSERVED: Visitors dropped 18.2%, trials dropped 18.0%, and new paid signups dropped 18.3%. Trial-to-paid conversion was nearly unchanged. ARPA was also nearly stable.

CALCULATED: The signup decline is (418 − 510) / 510 = -18.0%. The MRR movement is -12.8%. Do not add percentages across unlike measures.

HYPOTHESIS: Lower acquisition volume is the strongest observed first-order signal. Churn deterioration is a secondary concern. The channel change is unverified as a cause.

UNKNOWN: Which channel lost visitors, cohort-level churn amounts, new MRR, expansion, contraction, and failed-payment recovery. The $1,900 MRR drop cannot yet be attributed precisely.

Next checks: First, compare visitors and trials by channel across both periods. Second, reconcile beginning MRR plus new and expansion MRR minus contraction and churned MRR to ending MRR. Third, inspect churned cohorts and failed payments. Investigate before changing prices or offers.

Second Example: Ecommerce Revenue Down With Stable Traffic

Illustrative example only. A store receives 20,000 sessions in both periods. It completes 500 orders each time. Purchase conversion stays at 2.5%. Yet average order value falls from $80 to $70.40.

Previous revenue: 500 × $80 = $40,000. Current revenue: 500 × $70.40 = $35,200. The decline is 12%.

OBSERVED: Order count and traffic are flat. CALCULATED: Lower basket value accounts for the change in this simple order-value decomposition. UNKNOWN: Whether discounts, cheaper product mix, returns, or another factor lowered AOV. Check product-level sales and discount history before proposing an offer change.

How to Review the AI Diagnosis

• Math: Recalculate each percentage from the actual numerator and denominator.
• Periods: Check equal days, weekdays, timezones, and reporting maturity.
• Definitions: Confirm that “revenue,” “customer,” and “conversion” stayed consistent.
• Evidence: Require a source for every factual claim and calculation.
• Causality: Reject claims that an event caused a decline solely because it came first.
• Unknowns: Ensure gaps are labeled instead of replaced with plausible numbers.
• Hypotheses: Look for supporting and contradicting evidence, not only confidence labels.
• Next checks: Each check should confirm or weaken a specific hypothesis.
• Actions: Assess reversibility, effort, and risk before acting.

Correlation vs. Causation

Suppose revenue declined after a price increase. The sequence raises a pricing hypothesis. It does not prove pricing caused the loss. Check conversion by cohort, plan mix, objections, acquisition quality, and the timing of changes.

Likewise, lower traffic after a campaign ends does not prove causation. Check channel reports, seasonality, referrals, and tracking.

Common Revenue Diagnosis Mistakes

1. Providing only a revenue number. It invites generic answers. Add volume, conversion, value, retention, and collection measures.
2. Comparing unequal periods. A 31-day month is not a 28-day month. Normalize or select equal windows.
3. Asking for “the reason.” It invites false certainty. Request ranked hypotheses with evidence gaps.
4. Treating a business change as proof. An ad pause is an event, not proven causation. Compare channel-level results.
5. Ignoring churn. New signups can hide recurring-revenue losses. Inspect cancellations and contraction.
6. Ignoring AOV or ARPA. Customer counts can stay flat as revenue per customer falls. Compare values and mix.
7. Overreading small samples. A drop from five sales to three looks large. Show counts beside percentages.
8. Changing strategy too soon. A weak signal does not justify a total funnel rebuild. Verify the driver first.

Turn the Diagnosis Into the Next Check

Make the strongest finding testable: “Which channel lost visitors?” or “Which cohort cancelled?”

Log the driver, evidence, unknowns, next check, and review date. If warranted, test one reversible change: restore a paused campaign or examine one landing page. Do not make major decisions from one AI output.

For routine sales reporting across a full period, keep using the ChatGPT monthly sales analysis prompt. Use this diagnostic workflow only when a specific decline needs investigation.

Once the cause is better supported, AI financial planning tools can help evaluate future cash or scenario implications. Diagnosis itself is not a forecast.

7-Day Implementation Plan

Day	Task	Time	Expected Output
1	Define revenue measure and equal periods.	20 min	Reconciled comparison
2	Collect business-model driver metrics.	30–45 min	Verified input dataset
3	Check source quality and missing fields.	20 min	Known / unknown / inconsistent list
4	Run the TRACE prompt.	20–30 min	Structured diagnosis
5	Verify the top hypotheses in source reports.	30–45 min	Evidence notes
6	Choose one small reversible test if justified.	20 min	One controlled action
7	Record the result and next review point.	15 min	Decision and evidence log

FAQ

Can ChatGPT tell me exactly why revenue dropped?

Only with sufficient evidence. Traffic down 16% with stable conversion and AOV points first to acquisition, not its underlying cause.

What data should I give ChatGPT?

Provide comparable revenue and model-specific metrics: ecommerce sessions, orders and AOV; SaaS MRR movements and churn.

How do I distinguish traffic problems from conversion problems?

If sessions fall from 10,000 to 8,000 at 2% conversion, orders fall from 200 to 160. Acquisition volume declined.

Should I change prices when revenue falls?

Not automatically. Check price changes, conversion, and plan mix. The owner decides whether to test.

What if several metrics worsened together?

Rank by evidence and revenue proximity. Never add percentages from different metrics.

What if important metrics are missing?

Mark Unknown / Data Needed. Request the smallest useful export; never invent missing values.

Conclusion

A chatgpt revenue drop analysis prompt separates measured changes from possible explanations. TRACE narrows uncertainty.

Compare equivalent periods, verify definitions, and run the prompt. You own the decision. A useful output may be: “Acquisition volume is the strongest observed signal. Why it fell remains unknown. Here are three checks.”