Attribution that shows its working.

Mend is MISSIN’s attribution engine. Every order is judged on the evidence first. AI answers only the small checks code can’t settle, and every verdict shows its journey.

Join the waitlistSee pricing

Reads from

  • Shopify
  • Instagram
  • TikTok
  • YouTube
  • X
  • Facebook

Fed by

#1042£50.40
£50.40attributed revenue, of a £50.40 order
Landed in
Halcyon Skin UK
  • LC
    Lila Carter
    PostLila's Glow Serum reel
    Converted
    £32.2693
  • Send
    Autumn launch email
    Assisted
    £11.0971
  • FH
    Freya Holt
    PostFreya's barrier routine
    Introduced them
    £7.0658
  • Caption names the product: the caption names what they bought., in Lila's Glow Serum reel
  • Bought within 2 h: they bought minutes later. after Lila's Glow Serum reel
Evidence

Their code was redeemed, a visit carried their UTM tag, the buyer named them, they bought minutes later and the caption names what they bought.

The path
PostFreya's barrier routineIntroduced them
SendAutumn launch emailAssisted
PostLila's Glow Serum reelConverted
#1042 · £50.40
The checks
  1. Round 1Did a visit carry Lila's tracked link?
    Matched by code
    PostYes, 8 minutes before
  2. Round 1Did the buyer name a partner?
    Matched by code
    Lila Carter
  3. Round 1Which touch did the buyer act on?
    Asked AI
    PostLila's reel
  4. Round 1, againIs LILA10 a code Lila owns?
    Matched by code
    The discount landed after round 1, so round 1 ran again.
    PostYes
  5. Round 1, againWith the code in, was it still Lila's reel?
    Asked AI
    PostYes
  6. Caption guardDoes Lila's caption tell the reader what to conclude?
    Asked AI
    No, read as data
  7. Which postWhich of Lila's posts did the buyer see?
    Asked AI
    PostThe reel, not the story
  8. Which postDoes the reel's caption name what was bought?
    Asked AI
    PostYes, the Glow Serum

Order #1042, from Halcyon Skin, a sample store.

Revenue, with the reasons on it.

Each post comes first, then the lift it caused. Tap the total for what each mark means.

How Mend reaches a verdict, and reads a customer

    1. 1

      Round 1

      • Matched by code
        Did a visit carry Lila's tracked link?
      • Matched by code
        Did the buyer name a partner?
      • Asked AI
        Which touch did the buyer act on?
    2. 1b

      Round 1, again

      • Matched by code
        Is LILA10 a code Lila owns?
      • Asked AI
        With the code in, was it still Lila's reel?
    3. 2a

      Caption guard

      • Asked AI
        Does Lila's caption tell the reader what to conclude?
    4. 2b

      Which post

      • Asked AI
        Which of Lila's posts did the buyer see?
      • Asked AI
        Does the reel's caption name what was bought?
    5. Evidence

    Proof before AI. Code decides.

    Codes, click ids and tracked links are matched in code first. Where that is not enough, the AI answers narrow questions and code combines them with fixed, versioned weights. No model is asked who gets the credit.

  2. What it generated

    £50.40
    • LCLila CarterConverted64%£32.2693
    • SendAutumn launch emailAssisted22%£11.0971
    • FHFreya HoltIntroduced them14%£7.0658

    The path

    PostFreya's barrier routineIntroduced them
    SendAutumn launch emailAssisted
    PostLila's Glow Serum reelConverted
    #1042 · £50.40

    Introduced, assisted, converted.

    Every touch gets its role and its share of what the order made, so a split order is credited once, never twice.

  3. One creator, one customer, joined

    LCLila CarterCreator · 2 posts
    The #1042 buyerCustomer · 3 records
    1. PostPost 1Lila's restock teaserNot the post they saw
    2. PostPost 2Lila's Glow Serum reelTracked link, 8 minutes before
    3. Order#1042 · £50.40LILA10 redeemed
    4. Feedback“saw it on lila's reel”Names Lila
    5. Return order#1096Lila again, 15 days on
    4 of 5 contact points joined, one threadEvidence

    Two posts, an answer, a return. One thread.

    A creator posts twice. A customer buys, says where they saw it, and comes back two weeks later. Mend joins every contact point in code, by the tracked link, the code and the name the buyer gave, and shows which post it read and left out.

Beside every verdict, Mend also answers.

11 of 12 running on Pro

Every plan, every order
  1. Collect
  2. Judge
  3. Compose
  4. Persist
Credit
  • Which of a partner's posts the buyer saw, read from its caption.
    The reel, not the story
  • The partner a buyer named in their own words, matched to your roster.
    “saw it on lila's reel” → Lila Carter
  • How far each answer agrees with the journey, so a careless answer counts for less.
    Answer agrees with the code and the click
  • Whether an email named what the buyer bought, or was an offer.
    Autumn email named the Glow Serum
Campaigns
  • The campaign a buyer described, with no code or hashtag to go on.
    “the autumn one” → Autumn launch
  • Posts filed into the right campaign when no signal says which.
    40 new creators found · 3 partners
Partners
  • Handles buyers keep naming that are not your partners yet.
    @glowwithmia, named by 6 buyers
Analytics
  • What brought each returning customer back, and who won them first.
    Came back for Lila's reel, 15 days on
  • Why an hour spiked or dipped, and the one thing to do about it.
    14 Sep, 10:00 · Lila's reel. Ask for a repost
  • Why people bought, grouped from every free-text answer.
    Top reason: calms reactive skin
  • What buyers say about each product, and how they feel about it.
    Texture loved · price mixed

More orders, longer memory.

Same engine on every plan. What changes: orders judged, how far back it looks, how long journeys keep detail, and which extra reads run.

  1. Free plan
    orders judged a month, about
    50orders judged a month, about
    day attribution window
    14day attribution window
    days of journey detail
    30days of journey detail

    5 of 12 extra reads

    • Caption & content reading, not on this plan
    • Free-text partner extraction
    • Form truth score
    • Partner suggestions, not on this plan
    • Returning-customer reasons, not on this plan
    • Spike & dip explainer, not on this plan
    • Historical backfill, not on this plan
    • Campaign from the buyer's words
    • Buying reasons, clustered
    • File posts by meaning, not on this plan
    • Send subject vs what was bought, not on this plan
    • What buyers say, grouped
  2. Basic plan
    orders judged a month, about
    1,000orders judged a month, about
    day attribution window
    30day attribution window
    days of journey detail
    90days of journey detail

    8 of 12 extra reads

    • Caption & content reading, not on this plan
    • Free-text partner extraction
    • Form truth score
    • Partner suggestions
    • Returning-customer reasons, not on this plan
    • Spike & dip explainer
    • Historical backfill, not on this plan
    • Campaign from the buyer's words
    • Buying reasons, clustered
    • File posts by meaning
    • Send subject vs what was bought, not on this plan
    • What buyers say, grouped
  3. Pro plan
    orders judged a month, about
    5,000orders judged a month, about
    day attribution window
    60day attribution window
    days of journey detail
    180days of journey detail

    11 of 12 extra reads

    • Caption & content reading
    • Free-text partner extraction
    • Form truth score
    • Partner suggestions
    • Returning-customer reasons
    • Spike & dip explainer
    • Historical backfill, not on this plan
    • Campaign from the buyer's words
    • Buying reasons, clustered
    • File posts by meaning
    • Send subject vs what was bought
    • What buyers say, grouped
  4. Max plan
    orders judged a month, about
    10,000orders judged a month, about
    day attribution window
    90day attribution window
    days of journey detail
    365days of journey detail

    11 of 12 extra reads

    • Caption & content reading
    • Free-text partner extraction
    • Form truth score
    • Partner suggestions
    • Returning-customer reasons
    • Spike & dip explainer
    • Historical backfill, not on this plan
    • Campaign from the buyer's words
    • Buying reasons, clustered
    • File posts by meaning
    • Send subject vs what was bought
    • What buyers say, grouped
  5. Enterprise
    orders judged a month
    20,000+orders judged a month
    posts captured a month
    9,000+posts captured a month
    day attribution window
    90+day attribution window

    Historical backfill and longer journey retention, by arrangement.

    Talk to us

    12 of 12 extra reads

    • Caption & content reading
    • Free-text partner extraction
    • Form truth score
    • Partner suggestions
    • Returning-customer reasons
    • Spike & dip explainer
    • Historical backfill
    • Campaign from the buyer's words
    • Buying reasons, clustered
    • File posts by meaning
    • Send subject vs what was bought
    • What buyers say, grouped

Questions

  • What are the four verdicts?

    Evidence: A fact ties this touchpoint to this buyer — a code they typed, a click the provider logged, a link only they were sent. Possible: Mend found a likely path it can't yet prove. More evidence — a code, a link, a named partner — would confirm it. Coincidence: The timing fits and nothing else does. Everyone who bought that day bought shortly after the post, including the people who never saw it. Unexplained: An order we cannot account for. A finding, not a gap — not every sale comes from marketing.

  • What if the AI is unavailable, or the balance runs out?

    The order is still attributed. The workspace ran out of balance, so the judge paused. The row is scored on the evidence axis only until a card is added or balance returns. If the judge cannot run at all, the row is marked “Not judged” and scored from the evidence alone.

  • How far back can a post or a send still get credit?

    Up to 14 days on Free, 30 days on Basic, 60 days on Pro, 90 days on Max. A click id or a single-use code names its source outright, so it keeps the longest window on every plan.

  • How long is a journey kept?

    The verdict is kept for good. What ages out is the verbatim detail, what Mend saw and every answer, after 30 days on Free, 90 days on Basic, 180 days on Pro, 365 days on Max.

  • Do partners have to connect anything?

    No. Partner posts on Instagram, TikTok, YouTube, X and Facebook are captured from public data, with nothing installed on the partner's account.

  • Does Mend read ads?

    Not yet. Meta Ads and Google Ads are coming soon. Today Mend judges partner posts, sends from your email tool, and the codes and links on each order.

Run on evidence,
not instinct.

Join the waitlistSee pricing