Is It Working? How to Report on Your Marketing Efforts

A marketing report arrives in a business owner’s inbox. It includes website traffic, social media impressions, engagement rates, advertising spend, conversions, and revenue. There are charts showing which numbers increased and which declined.

But after looking through the report, the owner still has a question:

Is the marketing actually working?

It’s a reasonable question, and one that a collection of metrics doesn’t necessarily answer.

Reporting has become easier in many respects. Ecommerce platforms, advertising tools, and social networks all provide detailed performance data. But understanding what that information means—and how much confidence to place in it—remains difficult.

In Influencer Marketing Hub’s 2025 social media marketing research, 25.8% of respondents identified measuring ROI as a challenge when scaling influencer campaigns. The difficulty isn’t always a lack of data. It’s often understanding how different activities contribute to the business.

AI tools like ChatGPT and Claude can make some of the mechanical work easier. They can help organize information, check calculations, identify patterns, and raise questions worth investigating.

But knowing whether marketing worked requires more than processing information. Someone still needs to decide what success means, assess the strength of the evidence, and understand the business context.

The purpose of using AI in reporting isn’t to hand over those decisions. It’s to spend less time assembling information and more time making sense of it.

A useful marketing report should answer three questions:

1. Did it work? Did marketing achieve what the business set out to accomplish?
2. What worked? Which activities have credible evidence of contributing to those results?
3. Is it working? Do current results suggest the strategy should continue, change, or receive closer examination?

These questions provide a practical way to measure marketing performance without turning every report into an exercise in collecting more numbers.

Start with the questions, not the numbers

Marketing performance should be evaluated against what the business was trying to accomplish.

That sounds straightforward, but it’s easy to begin with an activity instead of a business objective.

A brand might decide to run an influencer campaign, increase its advertising budget, or publish more content. Those are activities. They describe what the team intends to do, not necessarily what it hopes to achieve.

Before selecting metrics, clarify four things:

The distinction matters because a marketing activity can perform well by one measure while failing by another.

The same results can tell different stories

Consider an illustrative direct-to-consumer brand working with five creators.

The campaign generates 80,000 content views, 1,200 website visits, 40 tracked purchases, and eight pieces of reusable content.

Whether that represents success depends on the original objective.

If the goal was customer acquisition, the most important questions concern how many purchases came from new customers and how much those customers cost to acquire.

If the goal was awareness, reach, relevant audience exposure, and indications of brand discovery become more important.

If the goal was immediate revenue, the brand needs to examine sales alongside campaign expenses and product margins.

If the goal was content production, eight usable pieces of content might represent meaningful progress, even if immediate sales were limited. Usage rights, quality, and suitability for future campaigns would also matter.

The metrics haven’t changed. Their significance has.

Decide what success looks like before launching

A useful campaign plan defines an expected result, a timeframe, a budget, and the evidence required to evaluate success.

For example:

Generate 30 first-time customer orders associated with the creator campaign over six weeks, while keeping acquisition cost within the brand’s acceptable range.

This gives the team a concrete target without pretending that all possible marketing effects will be captured.

It’s also worth recording what could challenge the original assumption. Perhaps creators reach the intended audience but generate little interest. Perhaps visits are strong but the product page fails to convert. Or perhaps the campaign generates revenue primarily from existing customers.

A report written without predefined success criteria risks becoming a search for whichever numbers look favorable.

Set Yourself Up for Success: How to Write Marketing Proposals explains how to agree on those objectives, costs, and measurement responsibilities before work begins.

AI can help a marketer think through possible metrics or identify ambiguities in a campaign plan. But choosing the objective is a business decision. It requires understanding priorities, constraints, and what the business considers worthwhile.

The goal isn’t to predict performance perfectly. It’s to establish an honest basis for evaluating what happened.

Collect the evidence

Once the objectives are clear, the next step is to gather the information needed to evaluate them.

For small ecommerce teams, this information rarely lives in one place.

Shopify may contain orders, discounts, refunds, and customer purchase history. Meta and Google may report advertising spend and attributed conversions. An email platform may show sends, clicks, and attributed sales.

Creator marketing adds another set of records: who was invited, who accepted, what was agreed to, which products were gifted, which content was published, and which discount codes or affiliate links were assigned.

A campaign might have produced measurable sales while also missing half of its agreed deliverables. Both facts matter, but they answer different questions.

Three types of marketing evidence

Operational records describe what was planned, agreed to, and delivered. These include campaign objectives, creator agreements, gifting, content, deliverable dates, and compensation commitments.

Transactional records describe business events. These include orders, revenue, refunds, discounts, commissions, and expenses.

Marketing-platform data describes reported exposure, engagement, traffic, and attributed conversions from advertising, social, email, and affiliate systems.

Each category provides part of the picture.

Operational records can establish that a creator published an agreed video. Platform data might establish that the video received 15,000 views. Transactional data might show orders using that creator’s code.

None of those facts, by itself, establishes the full financial impact of the partnership.

This is why consistent campaign and creator records matter. Keeping track of relationships, campaign participation, agreed deliverables, and published content gives later analysis something reliable to work from.

For more on organizing this information, see How to Manage Creators with ChatGPT and Claude.

Keep definitions and reporting periods consistent

Before combining information, establish basic reporting rules.

A practical reporting record should identify the date range, timezone, data source, export date, and definition of each important metric.

For example, Shopify distinguishes between gross sales, net sales, and total sales. Its financial reporting documentation explains that net sales account for discounts and sales reversals, while total sales also incorporate items such as shipping and taxes.

A report comparing Shopify total sales with an advertising platform’s reported purchase value may therefore be comparing different measures.

Refund timing also matters. A campaign may end before every return has been processed. Expenses can also be recorded at different times. A creator fee committed during one month might not be paid until the next, but it may still belong in the campaign’s cost assessment.

Missing information should be identified rather than silently treated as zero. An unknown creator fee, unavailable advertising cost, or untracked purchase is not the same thing as a confirmed absence of activity.

Reliable reporting begins with knowing what the records can—and cannot—tell you.

Bring your data together

Traditionally, assembling a marketing report involves exporting CSV files, copying information into spreadsheets, checking formulas, matching records, and creating summary tables.

For a small team, that process can consume much of the time available for analysis.

Spreadsheets remain useful. They provide a familiar way to inspect calculations, maintain references, and correct errors. A carefully maintained workbook is often more valuable than a sophisticated dashboard nobody fully understands.

But AI tools can reduce some of the manual work involved in exploring that information.

OpenAI documents how ChatGPT can analyze uploaded data, including examining tables, identifying inconsistencies, calculating summaries, and combining datasets through shared identifiers.

Anthropic similarly documents Claude’s ability to analyze uploaded CSV and spreadsheet data.

The opportunity isn’t to hand over reporting responsibility. It’s to make the information easier to inspect and understand.

A practical process for organizing marketing exports

A straightforward approach is to:

1. Export the relevant orders, marketing performance, creator records, and campaign expenses.
2. Confirm the reporting periods, currencies, and metric definitions.
3. Remove unnecessary personal information and upload only approved data.
4. Ask the AI tool to describe the structure and meaning of each dataset.
5. Identify reliable matching fields, missing values, and potential discrepancies.
6. Reconcile records using trustworthy identifiers, such as a shared order ID.
7. Check important totals against the original source reports.
8. Preserve the original files and document any transformations or unresolved differences.

The fifth and sixth steps deserve particular attention.

Suppose an ecommerce export contains an order identifier and discount code, while a separate creator report includes the same order identifier and a creator name. Those records may be linked confidently if the identifiers and definitions agree.

But two records containing a similar customer name and purchase amount are not necessarily the same transaction.

AI can suggest relationships that look reasonable without having enough evidence to establish them. Uncertain matches should remain unresolved until verified.

AI prompt: Inspect and organize marketing data

Use this prompt with actual exports from your business:

An initial inspection like this is often more useful than immediately asking an AI assistant to generate a dashboard.

It’s also worth inspecting the results of the AI-assisted reconciliation itself. Ask how records were joined, what was excluded, and which calculations were performed. Verify important totals against the original reports before using the combined data to evaluate performance.

A note on data privacy: Customer exports can contain names, email addresses, shipping information, and other personal data that may be unnecessary for reporting. Remove or aggregate that information wherever possible. Review company policies, client agreements, and the AI provider’s data-handling settings before uploading files. OpenAI’s data controls and Anthropic’s privacy documentation are useful starting points; available protections depend on the account and workspace.

Dig into the results

With organized data, reporting becomes less about assembling numbers and more about understanding relationships.

A first pass might establish that revenue increased, website traffic declined, and advertising spend remained steady.

Those observations are useful, but they raise more questions than they answer.

Why did revenue increase despite fewer visits? Did conversion rate improve? Did average order value rise? Were returning customers responsible for more purchases? Did a discount temporarily increase demand?

Good analysis moves from describing results to examining the conditions behind them.

Compare results with an appropriate baseline

A comparison requires a meaningful reference point.

Depending on the objective, that might be the original campaign target, a previous period, the same period in a prior year, or another reasonably comparable group.

Not every comparison is equally informative.

Comparing a six-week creator campaign with the preceding six weeks may be misleading if one period included a major product launch or a substantial change in advertising spending.

Likewise, a campaign producing more revenue after its budget doubled hasn’t necessarily become more efficient.

The comparison should account for meaningful differences in spending, promotions, seasonality, product availability, and customer mix wherever those factors can be measured.

An illustrative campaign review

Suppose a brand planned a creator campaign around acquiring 30 new customers. At the end of six weeks, its records show 42 purchases associated with creator discount codes.

The initial result sounds positive.

But closer inspection shows that only 17 of those orders came from first-time customers.

If the original goal was new customer acquisition, the campaign fell short of its target—even though the total number of tracked purchases exceeded 30.

That doesn’t mean the campaign had no value. Returning-customer purchases may have contributed revenue, while the creators may have produced useful content.

It means those benefits should be reported separately rather than substituted for the original objective.

Separate calculation, investigation, and judgment

There are three related but different parts of performance analysis.

Calculation asks what the records show. How many first-time orders were recorded? What was the total spend? How did conversion rate change?

Investigation asks what might explain the results. Did a discount influence purchases? Was the audience different? Were there changes in advertising spend or product availability?

Judgment asks what conclusions are justified. Did the campaign achieve its objective? Is the evidence strong enough to justify another investment? What matters most to the business?

AI can be particularly useful with calculation and investigation. It can process large exports, compare periods, identify unusual changes, and suggest possible explanations.

But possible explanations aren’t established causes.

An AI assistant might correctly calculate that a creator generated fewer tracked purchases while incorrectly concluding that the audience was poorly matched to the product. The calculation may be sound, but the explanation could overlook tracking gaps, content objectives, or other factors.

This is an important distinction: AI-generated errors aren’t limited to fabricated numbers. A plausible interpretation can also exceed what the evidence supports.

AI prompt: Investigate campaign performance

A response to this prompt should provide a shortlist of observations and questions worth investigating, not a ready-made verdict.

When a surprising result appears, inspect the underlying records. Check formulas, date boundaries, filters, and source definitions before treating the finding as meaningful.

Understand attribution and conflicting numbers

One of the most confusing parts of marketing reporting is attribution.

Attribution is the process of assigning credit for a conversion, such as a purchase, to one or more marketing interactions.

Different systems can assign credit differently, even when they’re observing the same customer journey.

Consider this illustrative example:

A customer discovers a product through a creator’s video. A few days later, the customer encounters a paid social advertisement for the product. After searching for the brand, the customer visits the store and purchases using the creator’s discount code.

The merchant has one purchase.

But several marketing activities may have contributed to it.

The creator introduced the product. Advertising provided another exposure. Search helped the customer return. The discount code was used at checkout.

Depending on their tracking coverage and rules, different platforms might each report some credit for that purchase.

This is not necessarily a reporting error.

Four differences that matter

Duplicate transactional records occur when the same underlying transaction is included more than once in a dataset or calculation. For example, joining an order to multiple line items and then summing the order’s full value on every row could inflate reported revenue.

Overlapping attribution claims occur when different marketing systems take credit for the same underlying purchase. Those reports may accurately reflect their individual attribution rules without representing separate purchases.

Attribution models determine how credit is assigned. A first-click model emphasizes discovery; a last-click model emphasizes the final recorded interaction. A linear model divides credit among recorded interactions.

Commission crediting rules determine which partner receives compensation under an agreed arrangement. Those rules do not necessarily establish which partner caused the sale.

A fifth concept is especially important: incrementality.

Incrementality asks whether marketing generated additional business that would not otherwise have occurred.

An order associated with a creator code is evidence that the code was used. It doesn’t establish that the creator caused the customer to buy. The customer might have purchased anyway, or may have discovered the code after deciding to buy.

Incremental impact is more difficult to establish and often requires a structured comparison, such as a reasonably controlled holdout group or geographic test.

Why attribution numbers disagree

Shopify’s marketing reports documentation supports several attribution models, including first click, last click, last non-direct click, any click, and linear attribution.

Notably, Shopify explains that its any-click model can allocate full credit to multiple channels involved in a purchase. As a result, attributed totals across channels can exceed the actual value of purchases.

That distinction matters when someone attempts to add independently reported revenue from Shopify, Meta, Google, and an affiliate program.

The combined figure could count the same purchase’s attributed value multiple times.

Attributed revenue from independently reporting platforms should not simply be added together and presented as total sales.

The merchant’s order records provide a useful reference for the underlying transactions, while marketing-platform reports help explain how those transactions were attributed.

Even then, reconciliation depends on the tracking available, the attribution windows, and whether records can be reliably connected.

What affiliate research adds to the picture

In its 2025 State of Affiliate Marketing research, impact.com reported that 94% of surveyed brands were experimenting with, or planning to adopt, alternative attribution models within the following year.

That is a finding about reported intentions and experimentation—not proof that an alternative model is more accurate.

The underlying issue is important: different partners may contribute at different moments in a customer’s journey.

impact.com’s Contribution Report documentation distinguishes partners that introduce, influence, or close conversions.

Its deduplication guidance also explains how overlapping claims across affiliate networks can create duplicate payout risks.

These are related issues, but they shouldn’t be confused. Understanding contribution, preventing duplicate payouts, and establishing incremental sales require different evidence and decisions.

For a small brand, the practical starting point is simpler than building a complex attribution model: reconcile actual orders, identify overlapping reported credit where possible, document commission rules, and describe attribution limitations openly.

Creator content may also provide value through product discovery, audience exposure, or reusable assets. Those outcomes deserve attention when they were part of the campaign objective, but they shouldn’t automatically be converted into claims of financial return.

Answer the questions that matter

Once the records are organized and attribution limitations are understood, the original questions become easier to answer.

These questions should guide the marketer’s assessment, rather than simply become another prompt asking an AI assistant to decide what happened.

Did it work?

Return to the business objective.

If the goal was acquiring customers, compare new customer orders and acquisition costs with the established target.

If the goal was revenue, compare sales with the campaign’s financial expectations.

If the goal was producing content, evaluate the agreed deliverables, quality, rights, and usability of the resulting assets.

For revenue-focused work, cost matters as much as sales.

A simple return on ad spend (ROAS) calculation divides attributed revenue by advertising spend. That can help compare advertising efficiency, but it does not establish profitability.

For creator marketing, a fuller assessment may need creator fees, products gifted at their relevant costs, shipping, commissions, paid amplification, production expenses, and product costs.

Even a positive calculated return on marketing investment is not necessarily proof of incremental profit unless the additional business caused by marketing has been credibly established.

Whether a result was worthwhile ultimately depends on the original objective, the available evidence, and the economics of the business.

What worked?

Look for activities with meaningful evidence of contribution.

Which creators generated tracked visits or purchases? Which content formats held attention or produced engagement consistent with the objective? Which channels attracted first-time customers? Were some results more efficient after accounting for costs?

A creator with fewer tracked purchases might still have produced valuable licensed content. Another might have generated strong awareness but little measurable revenue.

These distinctions become more useful when the evidence is considered against each partnership’s purpose.

The key is to avoid turning every association into a causal claim.

A creator publishing shortly before sales increased does not, by itself, establish that the creator caused the increase.

Is it working?

A completed campaign asks what happened. Ongoing reporting asks whether the current approach is producing acceptable results and remains worth pursuing.

That means examining pacing, repeated performance, emerging problems, and changes over time.

For example, an active creator campaign might be meeting its content-delivery targets while spending more quickly than expected and generating fewer first-time orders than planned.

Waiting until the campaign ends would make the report less useful.

Regular reporting provides an opportunity to adjust the work while those decisions can still matter.

Separate facts, interpretations, and uncertainty

One practical way to improve a report is to distinguish three levels of confidence.

This approach is more credible than claiming that creators generated 42 entirely new sales.

It is also more useful. It tells decision-makers what is known, what can reasonably be inferred, and what would require additional investigation.

Use AI to challenge your conclusions

Once a marketer has reviewed the evidence and formed preliminary conclusions, AI can serve another useful purpose: acting as a critical reviewer.

Instead of asking ChatGPT or Claude to decide whether the campaign succeeded, provide the proposed conclusions and ask where the reasoning might be weak.

This changes the role of AI from producing an answer to helping evaluate the quality of one.

It can be particularly useful when the results appear promising. Positive performance can make it tempting to overlook alternative explanations, attribution limitations, or costs that weren’t included in the original calculation.

A critical review won’t guarantee that every issue is discovered. But it can help surface questions that deserve consideration before conclusions are presented to a client or business owner.

AI prompt: Review the strength of your conclusions

The aim isn’t to eliminate every uncertainty. Marketing decisions rarely come with perfect information.

It’s to understand which conclusions are supported, which are tentative, and which claims should be avoided.

Turn findings into recommendations

Reporting has limited value if it doesn’t affect what happens next.

Once results are understood, most marketing decisions fit into four practical categories:

• Continue: Keep an activity that has reasonably strong evidence of meeting its objectives.
• Change: Adjust an activity when the evidence identifies a plausible improvement.
• Stop: Discontinue an activity that repeatedly underperforms against relevant goals and costs.
• Investigate: Gather more evidence when the result is promising, conflicting, or inconclusive.

Consider an illustrative creator campaign involving three partners.

The first creator delivered all agreed content, generated several first-time purchases, and performed consistently across two campaigns.

The second produced attractive content but limited tracked sales. The brand also has permission to use the content in advertising.

The third missed agreed deliverables and generated little measurable engagement.

The recommendations shouldn’t necessarily be identical.

For the first creator, the observation might support continuing the partnership while evaluating whether additional investment remains worthwhile.

For the second, the recommendation might be to evaluate the content against the original creative objective before deciding whether to renew the relationship.

For the third, the immediate action may be to resolve the outstanding deliverables and reconsider future commitments.

Each recommendation follows the same progression:

Observation → Interpretation → Recommendation

For instance:

Observation: A creator delivered two approved videos, but the tracked purchase count was below the campaign target.

Interpretation: The partnership has not yet demonstrated the expected direct-response performance, although the resulting content may have separate value.

Recommendation: Assess the licensed content’s usefulness and total partnership cost before renewing a sales-focused agreement.

The recommendation reflects the evidence without claiming more than the records support.

The evidence informs the decision. It doesn’t make it.

A recommendation involves more than interpreting performance data. It also reflects available budgets, business priorities, relationships, opportunity costs, and the risks a team is willing to accept.

Two businesses can examine the same campaign results and make different reasonable decisions.

One might continue investing in creator awareness because it supports a longer-term objective. Another might stop because it needs measurable short-term customer acquisition.

Neither decision can be evaluated from campaign metrics alone.

This is also where experienced marketers contribute something that reporting tools cannot reliably supply: an understanding of why certain tradeoffs matter to a particular business.

AI can help explore possible courses of action, compare costs, or identify factors that may have been overlooked. But those outputs are inputs to a decision, not instructions that should be followed automatically.

Strong evidence can justify clearer decisions. Weak evidence calls for smaller, reversible changes or additional investigation.

A useful report should make those tradeoffs easier to understand, not create the impression that the numbers automatically determine the correct course of action.

Create a report people can understand

After the analysis is complete, the final challenge is communication.

A client or business owner generally doesn’t need to see every exported field, calculation, or exploratory chart. Those records should remain available for verification, but the report itself should explain the important findings.

A useful structure is:

1. Original objectives: What the campaign intended to accomplish.
2. Executive summary: The main conclusion in plain language.
3. Important results: The few metrics that matter most to the objective.
4. Interpretation and context: What contributed to the outcome and which comparisons are meaningful.
5. Uncertainties and limitations: What cannot be established confidently.
6. Recommendations and next actions: What should happen next, and why.

An executive summary should be understandable without reading the entire report.

An illustrative executive summary

Returning to the six-week example with a target of 30 new customers:

The creator campaign recorded 42 purchases using creator discount codes, including 17 first-time customer orders against a goal of 30. The recorded acquisition result fell short. Most code-associated purchases came from returning customers.

These records do not establish incremental revenue or explain why acquisition fell short. Before deciding whether to repeat the campaign, review campaign costs, audience fit, and the available evidence from visits through purchase. Assess content delivery and usage rights separately against the original agreement.

This is more useful than a collection of positive and negative percentages without an explanation.

Let AI help with clarity, not conclusions

After the findings and recommendations have been reviewed, AI can help organize the report and improve its readability.

This is a different task from asking AI to interpret raw data.

The marketer has already established the objectives, reviewed the evidence, and decided which conclusions are justified. The writing tool’s job is to communicate those conclusions accurately.

For example, an AI assistant might identify repetitive explanations, suggest clearer section headings, or shorten an executive summary without changing its meaning.

A practical prompt might be:

Good reporting also makes disappointing results understandable. A missed target, explained accurately, is more valuable than a favorable metric selected to avoid an uncomfortable conversation.

For additional guidance on that part of the process, see Making Your Case: How to Present Marketing Results with Confidence.

Better reporting supports better decisions

Knowing whether marketing is working requires more than access to performance metrics.

It begins with clear objectives, continues through reliable records, and depends on distinguishing what happened from what marketing may have contributed.

Spreadsheets, analytics platforms, and AI tools can all make that process easier. AI is particularly useful for organizing information, investigating patterns, checking calculations, and challenging preliminary interpretations.

But the responsibility for drawing conclusions remains with the people who understand the business, its objectives, and the decisions those conclusions will inform.

For creator marketing in particular, keeping relationships, agreements, gifting, content, discount codes, commissions, and campaign activity organized makes later reporting easier to explain. That’s part of the purpose of CreatorMarketing.app: maintaining understandable records of the work behind creator partnerships.

A visually impressive dashboard can show hundreds of metrics without resolving the question that matters.

A good report may be much simpler. It explains what the business hoped to accomplish, what the evidence shows, what remains uncertain, and what should happen next.

The value of reporting isn’t having more numbers—or having AI interpret them for you. It’s developing a clearer understanding of the business so that better decisions can be made.