Skip to main content
CognizonYour team’s AI analyst
← All use cases

Use case · TikTok Shop owners

Know which creators and products actually make money.

Cognizon joins orders, affiliate commissions, samples, campaign spend, refunds, and settlement into contribution by creator — then answers the follow-up and drafts the memo.

Start with your own dataFollow the analysis

You bring the seller-center exports and your own cost sheet. Cognizon does the joining, the arithmetic, and the explanation.

Sample shop · synthetic data

QuestionWhich affiliate creators actually made us money in August?

Creator B brings 33.5% of affiliate GMV but only 14.2% of the contribution.

Where each creator’s GMV wentAffiliate orders, August 2026 · IDR millions
  • Contribution
  • Samples and campaign spend
  • Refunds
  • Commission and platform fees
  • Product cost

  1. Creator BGMV 158.6 · contribution 13.6 (8.6%)

  2. Creator DGMV 112.2 · contribution 38.2 (34.0%)

  3. Creator AGMV 96.4 · contribution 24.9 (25.8%)

  4. Creator CGMV 64.8 · contribution 18.6 (28.7%)

  5. Creator EGMV 41.5 · contribution 0.6 (1.4%)

Show the numbers
GMV and its cost stack by creator, IDR millions, August 2026
CreatorGMVContributionSamples and campaign spendRefundsCommission and platform feesProduct cost
Creator B158.613.624.838.633.648.0
Creator D112.238.25.82.222.044.0
Creator A96.424.98.46.420.736.0
Creator C64.818.63.64.813.824.0
Creator E41.50.614.21.59.216.0
All creators473.595.956.853.599.3168.0

One pass

From the exports to the memo, on the same rows.

The joined rows become the chart. It stays beside the chat while you ask why, keep a definition for your team to review, get the memo, and ask the next question.

  1. Read the exports
  2. Clean and check the keys
  3. Join on order, creator and SKU
  4. Chart what each creator kept
  5. Ask why Creator B is low
  6. Save a definition for review
  7. Memo and dashboard
  8. Ask the next question

Sample analysis

  1. Order exports from the Order page

    1. all_orders.csvorder_id · skuorder lines, SKU, value, refunds and the refund reasonrefund value and reason kept on each order line

  2. Affiliator exports from the Affiliator page

    1. affiliate_creator_list.xlsxorder_id · creator_idwhich creator each order came from, commission rateeach order_id matched to one creator_id
    2. sample_analysis_by_creator.xlsxcreator_id · skusample units sent to each creatorsample units priced from the cost sheet

  3. Internal data your own files, such as a cost sheet

    1. cost_sheet.xlsxskuproduct cost per SKU, kept by handunit_cost read per SKU

  4. Also exported from other Seller Center pages

    1. income_statement.xlsxorder_idcommission and platform fee actually deducted3 lines with no order_id, listed not absorbed
    2. ad_creative_data.xlsxcreator_idad spend on each creator’s videosad spend summed per creator_id
    3. product_list.xlsxskuthe SKUs the orders refer toSKU names matched to the order lines
Joined creator modelWhere each creator’s GMV wentCreator profitabilityAffiliate orders · IDR millionsDashboard · shared with your organization
  • Contribution
  • Samples and campaign spend
  • Refunds
  • Commission and platform fees
  • Product cost

  1. Creator BGMV 158.6 · contribution 13.6 (8.6%)

  2. Creator DGMV 112.2 · contribution 38.2 (34.0%)

  3. Creator AGMV 96.4 · contribution 24.9 (25.8%)

  4. Creator CGMV 64.8 · contribution 18.6 (28.7%)

  5. Creator EGMV 41.5 · contribution 0.6 (1.4%)

order_id · creator_id · sku · 3 unmatched settlement lines listed

Contribution, all creators
IDR 95.9M
Creator B refund rate
24.3%
Unmatched settlement lines
3 listed

YouWhy is Creator B so low when it sells the most?

Cognizon

Mostly refunds. Creator B’s orders refunded IDR 38.6M, 24.3% of its GMV, against 4.7% across the other four creators. 58% of that value came back as “not as described”. Samples and campaign spend took another IDR 24.8M.

Creator B refunds by reason · IDR 38.6M
  1. Not as described58%
  2. Wrong size or variant25%
  3. Damaged in transit11%
  4. Changed mind6%

Usedall_orders.csv + affiliate_creator_list.xlsx, joined on order_id

YouCommission is paid on orders net of refunds. Keep that as our definition.

Cognizon

That is how this chart already counts it, so no figure changes. I saved it for your team to review before later answers use it.

Business definitionSample shop · creator profitability

Affiliate commission is charged on orders net of refunds, not on GMV.

  1. Said in this chat
  2. Saved for team review
  3. Approved by a teammate

YouWrite that up for Monday’s ops review, with what we should change.

Cognizon

Drafted for your review. The chart is saved to the dashboard, and the memo is attached as PDF and PowerPoint.

Memo · Creator profitability · August 2026Draft for your review

Creator B drives 33.5% of affiliate GMV and 14.2% of contribution. Refunds are the gap.

Recommendation

Pause new samples and campaign spend for Creator B until “not as described” refunds come down. Give September’s first sample allocation to Creator D and Creator C.

Attached to this memoPDFPowerPoint

YouWhich creators get September’s samples?

Cognizon

Using your business context

  • Commission on orders net of refunds · approved definition
  • August 2026 memo on Creator B

Start with Creator D and Creator C: contribution margins of 34.0% and 28.7%. Creator E returned IDR 0.6M after IDR 5.2M in samples, so it is the first to cut. Creator B stays paused, as the memo recommends.

Contribution margin · August 2026
  1. Creator D34.0%
  2. Creator C28.7%
  3. Creator E1.4%

YouDid Creator B’s GMV keep growing through August?

Cognizon

Using your business context

  • Same joined data as the memo

Yes, every week: from IDR 30.2M in the week of Aug 3 to IDR 48.0M in the week of Aug 24. These exports carry no refund dates, so they cannot show whether refunds grew with it.

Creator B GMV by week · IDR millions
  1. Week of Aug 330.2
  2. Week of Aug 1036.4
  3. Week of Aug 1744.0
  4. Week of Aug 2448.0

What it read

Seller-center exports and your own cost sheet.

Contribution needs every deduction charged back to the creator that caused it. That lives in seven files, joined on order and creator, with the rows that do not match listed rather than dropped.

all_orders.csvorder lines, SKU, value, refunds and the refund reason

affiliate_creator_list.xlsxwhich creator each order came from, commission rate

sample_analysis_by_creator.xlsxsample units sent to each creator

ad_creative_data.xlsxad spend on each creator’s videos

income_statement.xlsxcommission and platform fee actually deducted

product_list.xlsxthe SKUs the orders refer to

cost_sheet.xlsxproduct cost per SKU, kept by hand

7 FILES · ORDER_ID AND CREATOR_ID KEYS · 3 UNMATCHED SETTLEMENT LINES LISTED

Ask the next question

The chart raises a question. Ask it in the same thread.

The follow-up reuses the same joined data, so the answer arrives with its chart and the files it used.

YouWhy is Creator B so low when it sells the most?

Cognizon

Refunds, first. Creator B’s orders refunded IDR 38.6M, 24.3% of its GMV, against 4.7% across the other four creators, and 58% of that value came back as “not as described”. Samples and campaign spend on its videos took another IDR 24.8M.

Creator B refunds by reasonIDR 38.6M refunded · share of value
Not as described 58%, Wrong size or variant 25%, Damaged in transit 11%, Changed mind 6%
  • Not as described58%IDR 22.4M
  • Wrong size or variant25%IDR 9.8M
  • Damaged in transit11%IDR 4.2M
  • Changed mind6%IDR 2.2M

Usedall_orders.csv + affiliate_creator_list.xlsx, joined on order_id

YouWrite that up for Monday’s ops review, with what we should change.

What you send

A memo for the meeting. A dashboard for every week after.

You review the memo before it is shared. The dashboard is assembled from charts you approved.

Memo · Creator profitability · August 2026Reviewed · approved

Creator B drives 33.5% of affiliate GMV and 14.2% of contribution. Refunds are the gap.

Recommendation

Pause new samples and campaign spend for Creator B until “not as described” refunds come down. Give September’s first sample allocation to Creator D and Creator C, the two highest contribution margins.

Contribution, all creators
IDR 95.9M
Creator B refund rate
24.3%
Creator E after samples and campaign
IDR 0.6M

Behind it: Creator D earns 34.0% of GMV as contribution; Creator B earns 8.6%.

7 sources · calculation and rows attachedPDFPowerPoint

Creator profitability

Dashboard · 3 widgets from approved charts

Shared with your organization

Contribution margin by creator
  1. Creator D34.0%
  2. Creator C28.7%
  3. Creator A25.8%
  4. Creator B8.6%
  5. Creator E1.4%
Weekly GMV, IDR millions
  • Creator B
  • Other creators
  1. 103.7Aug 3
  2. 114.6Aug 10
  3. 124.4Aug 17
  4. 130.8Aug 24
Refund rate by creator
  1. Creator B24.3%
  2. Creator C7.4%
  3. Creator A6.6%
  4. Creator E3.6%
  5. Creator D2.0%

Next month

The definitions you agreed on are there next time.

Context kept from the chat and the approved memo is listed in Memory, where anyone on the team can check or correct it.

Kept in Memory

  • Business definition

    Contribution counts samples at product cost and the campaign spend on each creator’s videos.

    Said in chat

  • Correction

    Affiliate commission is charged on orders net of refunds, not on GMV.

    Corrected on the memo

  • Past finding

    August 2026: Creator B was 33.5% of affiliate GMV and 14.2% of contribution, refunds at 24.3%.

    From the approved memo

A later conversation

Which creators get September’s samples?

Start with Creator D and Creator C: contribution margins of 34.0% and 28.7% in August 2026, with refund rates of 2.0% and 7.4%. Creator E returned IDR 0.6M after IDR 5.2M in samples, so it is the first to cut. Creator B stays paused, as the August 2026 memo recommends.

How it got there

The work, in the order it actually happens.

A chart is reviewed before it becomes a widget, and widgets exist before a dashboard does.

  1. 01

    Prepare and join

    Orders, affiliate attribution, sample requests, campaign spend, refunds, settlement lines, and your cost sheet are standardized and joined on order and creator. Rows that fail to match are listed and excluded, never quietly absorbed.

  2. 02

    Analyze and explain

    Cognizon computes contribution per creator — net sales less commission, platform fees, product cost, samples, and campaign spend — and explains what drove the gap, with the query, rows, and caveats attached.

  3. 03

    Review in Data Canvas

    Open the chart-backed answer in Data Canvas to cut it by week, creator, or SKU, refine the chart, inspect the rows under a number, and export what you need.

  4. 04

    Save what you approve

    Save the charts you approve as widgets, then assemble the dashboard from them. Nothing becomes a dashboard until you have approved the chart it came from.

  5. 05

    Write the briefing

    When the decision has to be read in a meeting, Cognizon writes the briefing as a memo: the finding, the evidence, and the recommendation — which creators get samples, which spend pauses.

Your team’s AI analyst

Start with the exports you already have.

Not your exact role? Cognizon works with the files, exports, metrics, and business context your team already uses.

Start with your own data
Other use casesMarketing AgenciesKey Account ManagersExplore all use cases →