Aria — Your AI Analyst

Getting to Know Aria, Your AI Amazon Analyst

Most reporting tools make you know where the answer lives before you can find it. Aria removes that step: you ask the question, and it goes and gets the answer. Watch the Aria walkthrough Prefer to watch? Eighty seconds on asking a question and saving the answer. What you actually get Three things that a normal report does not give you: 1. The answer, not the data. Aria draws the chart and tells you what it shows. 2. No navigation. You do not need to know which screen holds the number. 3. Reuse. A question worth asking weekly becomes a Studio you never ask again. Asking a question Open Aria → Chat in the sidebar and type into the message box. An Aria answer: chart plus written interpretation Here is what a real answer looks like. Aria restates the question, draws the chart, and then — the part that matters — reads it back to you: Your ROAS has been declining over the past 8 weeks, dropping from a peak of 1.86 in early July to 1.43 in the most recent week. This represents a 23% decrease in return on ad spend, suggesting your advertising efficiency has weakened significantly and may need optimization. That is the difference between a chart and an answer. The chart shows a line going down; the sentence tells you it is 23% and that it needs attention. Every answer carries a Download PDF button, and a Details control that opens the work behind it — useful when you want to check what Aria actually queried before you act on the number. Choosing the shape of the answer The visuals row beneath the message box Beneath the message box, the VISUALS row decides how the answer is presented: | Option | Use when | |---|---| | Auto | you do not care — Aria picks. The right default. | | None | you want the number in a sentence, nothing else. | | Charts | you are looking for a trend or a shape. | | Tables | you want to sort, scan or copy the figures out. | | Mixed | both. | | Dashboard | several related answers laid out together. | | Studio | you want this saved as a permanent, refreshing report. | Most of the time leave it on Auto. Reach for Tables when you intend to export, and Studio when you know you will ask this again. Writing a question that gets a good answer Aria is querying real data, so the more precisely you frame the question, the less it has to guess. A good question names three things: - the metric — ROAS, ACoS, units, sessions - the period — last 8 weeks, this month, since 1 July - the comparison — by week, by product, versus the previous period Weak: how are my ads doing? Strong: show me ROAS by week for the past 8 weeks, and tell me which ad type moved most. The second one produced the answer above. Studios: when a question is worth keeping A Studio is a dashboard Aria built because you asked for it. A Studio: title, date range and headline metrics It keeps a title and description, its own date range, and headline metrics with their trend and change — so opening it next week gives you next week's numbers, not a snapshot of the day you made it. Weekly ad spend broken down by ad type Underneath, the breakdowns: spend over time by ad type, and a comparison table per channel. Ad type performance comparison Studios live under Aria → Studios, and you can rename one from its title. When to make one: any question you have now asked twice. A weekly advertising review, a monthly category check, the three numbers you report upward. Ask it once properly, save it as a Studio, and the asking is done forever. Finding past conversations The conversation list with its filters The left panel keeps every conversation, with a search box and All / Chats / Studio filters. Answers are not throwaway — when you need last month's reasoning, it is still there. Checking Aria's work Aria queries your live data, but it is still interpreting a question. Two habits worth keeping: Opening Details on an Aria answer Details opens the work behind the answer — what Aria actually queried to produce it. Open Details on anything surprising. It shows what was actually queried. A number that looks wrong is usually a question that was read differently from how you meant it. Watch the date range. “Last month” and “the last 30 days” are different questions, and some data has its own cadence — Search Query Performance is weekly, AMC is monthly. If the period in the answer is not the period you meant, ask again with explicit dates. Where to start Ask one real question you actually need answered this week. Name the metric, the period, and what you want it compared against. If the answer is useful, save it as a Studio. Related - Asking Aria a Good Question — the single biggest lever on answer quality. - What Aria Can and Cannot Answer — the boundaries, stated plainly. - Visuals: Auto, None, Charts, Tables, Mixed, Dashboard and Studio — Auto, None, Charts, Tables, Mixed, Dashboard, Studio. - Studios: Building a Dashboard by Asking for It — building a dashboard by asking for it. - Managing Chats: Recent, Search and the All/Chats/Studio Filter — Recent, search, and the Details drawer. - Reading Aria's Plan and Data Queries — the working it shows under every answer. - Checking Aria's Work When an Answer Looks Wrong — when an answer looks wrong. - Aria Prompt Library: 40 Questions Worth Asking — 40 questions worth asking. - Chatting With Aria in Slack — the same Aria, where your team already works.

Chatting With Aria in Slack

The fastest way to get an answer out of TrackIQ is often not to open TrackIQ. Aria connects to Slack, so the same analyst that answers questions in the app will answer them in a DM or a channel — with the same data, the same reasoning, and no login. Watch the Aria in Slack walkthrough Prefer to watch? Fifty seconds on connecting it and pinning a brand per channel. Why this is worth doing Most questions about a brand's performance are asked in conversation, not in a dashboard. Someone in #marketing asks how last week went; someone in #ops asks whether a product is going to stock out. Today those questions get answered by whoever happens to have TrackIQ open. Putting Aria in the channel removes that bottleneck. The person with the question gets the answer, and the answer arrives where the decision is being made rather than in a screenshot pasted twenty minutes later. Connecting it Go to Settings → Aria in Slack. The Aria in Slack settings page The status chip on the right tells you where you stand — NOT CONNECTED until you finish, connected afterwards. Connect in three steps The three steps, in order. The first is a single click. 1. Connect your workspace. Click Add to Slack. This is one click and no Slack admin install is required — it works with any workspace you already belong to, which means you do not have to raise a ticket with whoever owns your Slack. 2. Pick a brand. Aria asks on your first message. If you manage several brands you get a picker, and your choice sticks to that channel until you change it. 3. Start asking. DM the app directly, or bring it into a channel with /invite @TrackIQ and then @TrackIQ it. Type switch at any time to re-pick the brand for that channel. The feature that matters most: per-channel brand pins What you get Each channel keeps its own brand. Brand A lives in #marketing, Brand B lives in #ops, and neither team ever has to remember to switch context before asking. This is the difference between an integration people use and one they abandon. If every question required re-selecting a brand, the first wrong answer — right numbers, wrong brand — would end the habit. Pinning removes that failure entirely. What the answers look like A question and answer in Slack You ask in plain English. Aria replies in the channel, with the figures inline. The one honest limitation: Slack replies are text-only. The reasoning is identical to Aria on the web, but charts and full tables do not render in Slack. When a question deserves a chart, ask in Slack for the answer and open Aria in the app for the picture. That shapes how to use it well. Slack is excellent for specific questions with short answers — which products performed best last month, is anything about to stock out, what did we spend yesterday. Longer analysis is still better in the app. Good first questions to try | Ask | Because | |---|---| | Which products performed best last month? | the standard Monday-morning question, answered without opening anything | | Is anything projected to stock out in the next two weeks? | catches the expensive problem early | | Which campaigns had the worst ACoS last week? | turns a channel into a review meeting | | How did branded search change after our display campaign? | the kind of question nobody bothers to log in for | A note on access Anyone you invite the app into a channel with can ask it questions about that channel's pinned brand. Treat channel membership as the access control — if a channel contains people who should not see a brand's numbers, do not pin that brand to it. For everything Aria can do, see Getting to Know Aria, Your AI Amazon Analyst. Related - Slack Notifications Stopped — when the channel goes quiet. - Get Your Daily Number in Slack Without Opening TrackIQ — the scheduled-summary playbook. - Asking Aria a Good Question — the same rules apply in Slack.

What Aria Can and Cannot Answer

Aria answers questions about your Amazon data, in the brand you have selected, using the same reporting TrackIQ shows on its screens. Knowing where that boundary sits saves a lot of wasted questions — and, more usefully, tells you when the honest answer is “ask a different way.” Watch the walkthrough Prefer to watch? Sixty-nine seconds on where Aria's boundary sits and how to phrase past it. What it can answer Aria answers from your own reporting, with the working shown Anything that lives in your reporting. In practice that means: | Area | Example | |---|---| | Sales and traffic | “Which products sold best last month?” | | Advertising | “Which campaigns had ACoS above 40% last week?” | | Amazon DSP | “How did display perform by device?” | | AMC | “Which products brought in the most new customers?” | | Search | “Which queries do I convert on but have no keyword for?” | | Rank | “Where do I rank for my top ten keywords?” | | Inventory | “What runs out in the next two weeks?” | The genuinely useful class is the one no single screen covers. The product has a screen for campaigns and a screen for inventory; it does not have a screen for “campaigns advertising products that will be out of stock before the click converts.” Aria can join those. What it cannot Anything outside Amazon. Your cost of goods, your other sales channels, your ad spend elsewhere. Aria sees your Amazon reporting and nothing else — so it cannot tell you true profit, only revenue and ad cost. Anything you cannot see. Aria runs under your own permissions and the brand in your top bar. A section your role does not include is not reachable by asking about it. Anything your plan does not include. Search Query Performance, DSP and AMC are not on Starter — see What's Included in Each Plan. Anything Amazon has not delivered yet. SQP publishes weekly, AMC monthly, and the last day or two of anything is still filling in. Changes. Aria reports; it does not act. Bids, budgets and pauses live in Rule-Based Optimization, under your control. The three that look like faults but are not These produce answers that feel wrong, and in each case the answer is right and the question was ambiguous. “Why do my ad types not add up?” Sponsored Products is credited over 7 days, Brands and Display over 14. A combined figure blends two windows — correct for comparing size, misleading for comparing efficiency. “Why is my share not the sum of the weeks?” SQP figures are percentages of that week's market. 20% and 20% is not 40%. See SQP Is Weekly. “Why is this product's rank blank?” Blank means not found within the depth checked — not rank zero. And lower is better throughout. Getting a better answer The fix is almost always in the question. Name the period. “Last month” beats “recently”. Name the thing. A product, a campaign, a keyword — rather than “my ads”. Ask one question. Three bundled together produce a summary of all three rather than an answer to any. Say what you would do with it. “Which campaigns should I cut budget on?” gets you a ranked list with a reason. “How is my advertising?” gets you a description. Check the state you meant. Only enabled campaigns are counted unless you say otherwise — if you want everything, ask for paused and archived too. See Asking Aria a Good Question. When the answer still looks wrong Aria shows its working — the plan it followed and the data it queried. Read that before disputing the number; it usually reveals a different date range or a different brand rather than a different answer. If it still looks wrong, check the same figure on the screen it comes from. Agreement means the question was read differently than you intended. Disagreement is worth reporting, with the brand, the marketplace and the exact date range. See Checking Aria's Work When an Answer Looks Wrong. When to use something else A chart you will re-read every week belongs in a Studio or a Business Report, not a fresh chat each Monday. Monitoring belongs in alerts and rules. Asking repeatedly whether something broke is the expensive way to do what the product already does unprompted. Amazon data inside other work — a plan, a client email, a model with your costs in it — is what MCP is for. See What Is MCP, and Why It Matters. Related - Feed TrackIQ Data into Your Own AI Workflow — questions better answered outside Aria. - Aria Prompt Library: 40 Questions Worth Asking — the ones it answers well.

Asking Aria a Good Question

Aria is good at answering precise questions and merely adequate at answering vague ones. The difference is almost entirely in how you ask — and the fix takes about five extra words. There is no cost to asking. TrackIQ does not cap usage, so this is about getting a better answer, never about spending fewer questions. Watch the walkthrough Prefer to watch? Seventy-four seconds on the four things a good question carries. The four things a good question carries Aria answers from your own reporting, and shows its working A period. “Last month”, “the last 30 days”, “February” — not “recently”. A subject. A product, a campaign, a keyword, a marketplace. “My ads” covers hundreds of things doing different jobs. A measure. ACoS, spend, sessions, days of cover. Say which one you mean, because “performance” means six things. A decision. What you would do with the answer. This is the one people leave out, and it changes the reply more than the other three. The decision is the trick Compare two questions that ask for the same underlying data: “How is my advertising doing?” …returns a description. Accurate, and you still have to decide what to do. “Which campaigns should I cut budget on this month?” …returns a ranked list with a reason against each, because you have said what the answer is for. Ask for the decision, not the data. Aria is doing the analysis either way — you may as well have it point at an action. Before and after | Instead of | Ask | |---|---| | How are my ads? | Which campaigns had ACoS above 40% in the last 30 days? | | Is inventory OK? | Which products have under 21 days of cover? | | How's the new product? | Sessions and conversion for ASIN X, last 14 days vs the 14 before | | What should I do? | Where am I spending most with the least return this month? | | Am I ranking? | Which of my tracked keywords dropped out of the top 10 this week? | The pattern in every right-hand column: a measure, a threshold, a period. Say the state you mean This one silently changes answers, and it catches people out. Only enabled campaigns are counted by default. Ask “how many campaigns do I have?” and you get the active ones. If you want the full picture, say “including paused and archived”. The same applies to a marketplace or brand — every answer is scoped to the brand in your top bar unless you say otherwise. Ask one thing at a time Three questions bundled into one produce a summary of all three rather than an answer to any. “How are my ads and inventory doing and what about rank?” Split it. Ask the first, read the answer, then ask the follow-up — and the follow-up is where the value usually is. The second and third questions are the ones that get specific, and nothing stops you asking them. Follow up rather than restart Aria holds the thread of a conversation. Once you have a list of campaigns, you can say “now show spend and orders for those” without naming them again. That makes a conversation cheaper to steer than a perfect first question. Do not agonise over the opening question — ask something reasonable and refine. Asking about a period is not the same as asking about a trend “What was my ACoS last month?” gets a number. “How has my ACoS moved over the last six months?” gets a direction. If you want a comparison, say both periods: “this month versus last”. Aria will not assume you wanted a comparison. Note the grain limits when you ask for a trend: daily breakdowns cover the last 30 days, Search Query Performance is weekly, and AMC is monthly. See The Full MCP Tool Reference for why. When the answer looks wrong Read the plan and the data queries Aria shows before disputing the figure. Nine times in ten the answer is right and the question was read differently — a different brand, a different range, enabled-only. See Checking Aria's Work When an Answer Looks Wrong and What Aria Can and Cannot Answer. When not to ask at all A question you ask every week belongs in a Studio or a scheduled Business Report, so it is waiting when you open it and consistent week to week. Monitoring belongs in alerts and rules. Asking repeatedly whether something broke is a worse way to find out than being told — alerts run whether you ask or not, and rules act the moment a threshold is crossed. For a starting set of questions worth asking, see Aria Prompt Library: 40 Questions Worth Asking. Related - Reading Aria's Plan and Data Queries — checking it understood you.

Aria Prompt Library: 40 Questions Worth Asking

Forty questions worth asking, grouped by what you are trying to decide. Copy one, swap in your own product or period, and adjust from the answer. They are written the way Asking Aria a Good Question recommends — a measure, a threshold, a period, and a decision. Some need data that starts at PRO; those are marked. Watch the walkthrough Prefer to watch? Sixty-three seconds on how the library is grouped, and the one question worth running weekly. Monday morning: what changed? Aria answers from your own reporting 1. What changed most in the last 7 days versus the 7 before, and why? 2. Which products gained or lost the most revenue week over week? 3. Which campaigns moved most in spend last week, up or down? 4. Did anything go out of stock in the last week? 5. Summarise last week in five bullet points I could send to my team. Advertising: where is the money going? 6. Which campaigns had ACoS above 40% in the last 30 days, ranked by spend? 7. Where am I spending the most with the least return this month? 8. Which campaigns spent money and produced no orders in the last 14 days? 9. What is my ACoS this month versus last, by ad type? 10. Which keywords cost the most per order in the last 30 days? 11. Show my daily spend and ad sales for the last 30 days. 12. Which campaigns should I cut budget on this month, and why? 13. Which campaigns are converting well enough that I should spend more? 14. What is my TACoS trend over the last six months? 15. Which search terms are converting that I have no keyword for? Products: what is working? 16. What were my ten best-selling products last month by revenue? 17. Which products have the highest conversion rate, and the lowest? 18. Which products lost the most revenue this month versus last? 19. Show sessions and conversion for ASIN X, last 14 days versus the 14 before. 20. Which products have sessions but almost no orders? 21. Which of my categories grew fastest last quarter? Inventory: what will stop me? 22. Which products have under 21 days of cover? 23. What is forecast to run out in the next two weeks? 24. Which products are out of stock right now? 25. Which products am I still advertising that will run out this month? 26. Which products have the most stock and the slowest sales? Question 25 is the one worth running weekly. No single screen answers it — it joins advertising to inventory, which is exactly the class of question Aria exists for. Search and rank (SQP needs PRO) 27. Which queries do I convert on but have no keyword for? 28. Where do I rank for my top ten keywords, and what moved this week? 29. Which of my keywords dropped out of the top 10 this week? 30. Which queries have high click share but low purchase share for me? 31. What are the highest-volume queries where my impression share is under 10%? 32. How has my purchase share moved week over week for my top query? Question 30 is a diagnosis in one line: winning the click and losing the sale points at the listing, not the bid. New customers (AMC needs PRO) 33. Which products bring in the most new-to-brand customers? 34. Which campaigns drive the highest share of new-to-brand purchases? 35. How long do customers take to buy after first seeing an ad? 36. Which products do customers buy together? Reconciling and sense-checking 37. Why is my ad sales figure lower than my total sales? 38. How fresh is my data, by source? 39. What is my organic versus paid split this month? 40. Show my ACoS by ad type for a period that closed a month ago. Question 40 exists for a reason: the last fortnight is still accruing, so a closed period is the only fair comparison across ad types. See Why SP, SB and SD Numbers Don't Add Up. Getting more out of these Add your own threshold. “Above 40%” should be your target, not a round number from a list. Say the state. Only enabled campaigns are counted unless you ask for paused and archived too. Follow up rather than restart. Once you have a list, “now show spend and orders for those” works without renaming anything. Pin the ones you repeat. Any question here that becomes a weekly habit belongs in a Studio or a scheduled Business Report — not because asking costs anything, but because a pinned view is waiting for you and stays consistent week to week. Ask in Slack when you are not in the app. The same questions work from a DM or a channel — see Chatting With Aria in Slack. Related - Asking Aria a Good Question — the pattern these are built on. - What Aria Can and Cannot Answer — where the boundary is. - What's Included in Each Plan — which of these your plan can reach.

Visuals: Auto, None, Charts, Tables, Mixed, Dashboard and Studio

Under the question box is a row labelled VISUALS, with seven options. It controls what shape Aria's answer takes — and two of the seven do something more interesting than formatting. Most of the time Auto is right. It is worth knowing the other six for the times it is not. Watch the walkthrough Prefer to watch? Sixty-six seconds on the seven options and when to override Auto. The row The visuals row, under the question box | Option | You get | |---|---| | Auto | Aria picks — the default | | None | prose only, no visual | | Charts | the answer as a chart | | Tables | the answer as a table | | Mixed | chart and table together | | Dashboard | several panels in one answer | | Studio | a saved, reusable dashboard | The setting applies to the question you are about to ask, so change it before pressing Enter. When to override Auto None — when you want a number or a sentence, not a picture. Good for quick checks and for anything you are going to paste into a message. Tables — when you intend to copy the figures, scan a list, or compare rows precisely. A chart is worse than a table for “which twelve campaigns and what did each spend?” Charts — when the shape matters more than the values. Trends, comparisons over time, anything where you are looking for a direction rather than a number. Mixed — when you want both, which is more often than people expect. The chart shows you the pattern; the table lets you check it. Dashboard — when one question genuinely has several parts. “How did last month go?” is not one chart. Studio is different The first six change how an answer is drawn. Studio saves it. Ask with Studio selected and you get a dashboard that persists — titled, described, and listed under Aria → Studios, where you can pin it. It is the difference between an answer you read once and a view you return to. Use it the moment you notice you have asked the same question twice. See Studios: Building a Dashboard by Asking for It. Every answer can be a PDF Download PDF sits under the answer Under any answer is a Download PDF button. It captures the answer as rendered — chart, table and the written interpretation together. This is the fastest route from “I need to explain this to someone” to something you can send, and it works regardless of which visual you chose. Worth knowing before you start screenshotting. The interpretation is the point Whatever visual you pick, Aria writes the sentence underneath it: Your ROAS has been declining over the past 8 weeks, dropping from a peak of 1.86 in early July to 1.43 in the most recent week. This represents a 23% decrease… Anyone can read a chart. The value is the reading, and that arrives with every visual option except None — which is a reason to leave None for the cases where you genuinely only want the number. Two things that catch people Charts do not always suit the question. Ask for a ranked list of thirty campaigns as a chart and you get thirty bars nobody can read. Tables win on lists; charts win on trends. The visual does not change the data. Switching from Charts to Tables re-renders the same answer — it does not re-query or reveal anything hidden. If a figure looks wrong, changing the visual will not fix it. See What Aria Can and Cannot Answer. A sensible default habit Leave it on Auto. Override when you already know the shape you want: - Pasting into Slack or an email → None - Checking specific figures → Tables - Looking for a trend → Charts - Presenting to someone → Mixed, then Download PDF - Asked it twice already → Studio Related - Asking Aria a Good Question — phrasing that gets the visual you wanted. - Managing Chats: Recent, Search and the All/Chats/Studio Filter — keeping the good answers findable.

Studios: Building a Dashboard by Asking for It

A Studio is a dashboard you get by asking for it. No layout editor, no choosing chart types, no dragging panels — you describe what you want to see, and the answer becomes something that persists. It is the one Aria feature that turns a conversation into an asset. Watch the walkthrough Prefer to watch? Sixty-four seconds on asking for a view rather than a figure. Making one Select Studio in the visuals row before you ask, then ask for the whole view rather than a single number: “Week-by-week advertising performance with ad type and product category breakdowns” You get a dashboard, and it is saved. It appears under Aria → Studios with a title, a description and a created date. The Studios list Ask for a view, not a figure This is the difference between a Studio worth keeping and one you never open again. | Weak | Strong | |---|---| | What was my ACoS last month? | Monthly ACoS by ad type, with spend and sales alongside | | Which products sold best? | Top products by revenue, with sessions and conversion for each | | How is inventory? | Days of cover by product, with velocity and inbound | A single number does not need a dashboard. A Studio earns its place when it holds several things you always look at together — which is exactly what you would have built by hand, if building it by hand were quick. Pin the ones you use Pinned, title, description and created date The list has a Pinned column. Pinning is what keeps a Studio in reach once you have more than a handful. Write the description properly. It is the only thing that tells you — or a colleague — what a Studio is for without opening it. “Week-by-week advertising analysis with ad type and product category breakdowns” is useful. “Ads” is not. The list also has Search, Filters, Manage Columns and Export, and a menu on each row. Studios versus Business Reports Both give you something recurring. They are not the same tool. | | Studio | Business Report | |---|---|---| | Made by | asking Aria | Templates or the AI Report Generator | | Best for | a view you return to | a document other people read | | Lives | in Aria | in Reports | | Shape | dashboard | Slides or Dashboard | The practical rule: a Studio is for you, a Business Report is for an audience. If it needs to be forwarded to a client or read without logging in, it wants to be a report. See Understanding Business Reports. Plenty of accounts run both — a Studio the team opens daily, and a Slides report that goes out weekly. When to make one The second time you ask the same question. That is the signal, and it is a reliable one. Before a recurring meeting. A Studio that answers the standing questions turns preparation into opening a page. Per client, if you are an agency. One Studio each, pinned, is a faster route to “how are we doing?” than rebuilding the same answer every time. Note that there is no cost pressure on any of this — asking is not metered, so the reason to build a Studio is convenience and consistency, never saving usage. See Are There Token Limits or Usage Caps? What it does not do It is not a live alert. A Studio shows you data when you open it. If you want to be told when something changes, that is alerts and rules. It does not escape your plan. A Studio can only show data your plan includes — asking for Search Query Performance on Starter will not work. See What's Included in Each Plan. It is not scheduled delivery. Nothing arrives in your inbox, and a Studio will not email itself to anyone — you open it when you want it. Related - Visuals: Auto, None, Charts, Tables, Mixed, Dashboard, Studio — the row that creates one. - Asking Aria a Good Question — the phrasing that produces a good Studio. - Aria Prompt Library — questions worth pinning.

Managing Chats: Recent, Search and the All/Chats/Studio Filter

Aria keeps every conversation. The left panel is how you get back to one — and the filter pills on it are the part worth understanding, because Studios and chats live in the same list. Watch the walkthrough Prefer to watch? Fifty-six seconds on the filter pills, and when to start a new thread. The panel New Chat, search, filter pills and the recent list Four things, top to bottom: + New Chat — starts a fresh thread. Search conversations… — searches across everything you have asked. All / Chats / Studio — the filter pills. RECENT — the list, newest first, each with its type and how long ago. The pills | Pill | Shows | |---|---| | All | everything | | Chats | ordinary conversations | | Studio | saved dashboards only | A Studio appears in the list beside your chats, marked with its own icon and labelled Studio under the title. So Studio is the pill to reach for when you want your saved views rather than your history — and Chats is for when a Studio you made is cluttering the list you are trying to search. See Studios: Building a Dashboard by Asking for It. New chat, or keep going? This decision affects your answers, so it is worth being deliberate. Keep going when the next question builds on the last. Aria holds the thread, so “now show spend and orders for those” works without naming anything again. That follow-up is usually where the real answer is. Start a new chat when you change subject. A long thread about advertising carries context that will colour an inventory question — and a fresh thread is one click. The rough rule: one thread per question you are trying to answer, not one per day and not one per query. Search is better than scrolling The search box covers your conversations, which makes it the fastest way back to “that thing I worked out last month.” Search for the subject, not your phrasing. A product name, a campaign name, a metric. You will not remember how you worded it; you will remember what it was about. That is also an argument for asking specific questions in the first place — a thread that begins “which campaigns had ACoS above 40% in July” is findable. One that begins “how are things” is not. See Asking Aria a Good Question. The Details drawer The conversation drawer, opened from Details Details, top right of any open chat, slides out a Conversation drawer. It is where a chat is managed, and it is easy to miss because nothing else points at it. Title, with a pencil to rename it. Aria names chats from your first question, which is usually serviceable and occasionally not. Renaming is what makes search work later — see below. Messages — how many turns the chat holds. A useful proxy for how much context it is carrying. Status — idle when nothing is running. It changes while Aria is working. Created — the date and time the chat began, in full. The list only shows relative ages ("19h ago"), so this is the only place to get the actual timestamp. Then two actions: Pin conversation — keeps a chat at hand instead of letting it slide down Recent. Delete conversation — permanent, and not undoable. Deleting also removes it from search, so a chat you may want to find again is better renamed than deleted. Rename before you pin. A pinned chat called "Which products performed best last month?" tells you nothing in three weeks; "Monthly product review — standing query" does. Getting an answer out Under any answer is Download PDF, which captures it as rendered — chart, table and written interpretation together. That is usually a better way to share a finding than a screenshot, and it works from any conversation without re-asking. A tidy habit Turn repeated chats into Studios. If you find yourself searching for the same thread every week, it wants to be a saved view rather than something you go looking for. Name what you are doing in the first message. The thread takes its title from how it starts, so the opening question is effectively the filename. Do not ration threads. There is no usage cost to starting a new chat — see Are There Token Limits or Usage Caps? — so keeping threads focused costs nothing and makes them findable. Related - Asking Aria a Good Question — the four things a good question carries. - What Aria Can and Cannot Answer — where the boundary is. - Chatting With Aria in Slack — the same questions, from a DM or channel.

Reading Aria's Plan and Data Queries

Aria shows its working. Every answer carries the plan it formed and the data it actually retrieved, and both are on screen without asking — which turns "do I believe this?" from a judgement call into something you can check. Watch the walkthrough Prefer to watch? A hundred and one seconds on the two parts of an answer nobody opens, and what the row count tells you on its own. The shape of an answer Plan, query and rows, under one answer Four parts, in order: 1. Your question, as you typed it. 2. The plan — an Insights block restating what Aria is about to do. In the example above: "I'll get you a chart of your account's ad performance ROAS over the past eight weeks." 3. The data query — a collapsible row headed Queried your data · 56 rows, with a green tick beside it. Expand it and you get the rows themselves. 4. The answer — the visual, then the written interpretation. Parts 2 and 3 are the ones nobody opens, and they are where the value is. Read the plan first The plan is a restatement, and restatements are where misunderstandings surface. Check the metric. You asked about ROAS; the plan should say ROAS, not ACoS. Check the period. "Past eight weeks" should match what you meant. If you said "recently" and the plan says eight weeks, that is Aria choosing for you — fine, but now you know. Check the scope. "Your account's ad performance" is account-wide. If you meant one product, the plan has just told you it did not understand that. If the plan is wrong, the answer will be wrong. Stop there and re-ask rather than reading on. It costs ten seconds instead of ten minutes. See Asking Aria a Good Question. Then open the query The disclosure expands to a real table of what came back. In the example: WEEK LABEL, WEEK NUM, TOTAL AD SALES, TOTAL AD SPEND, ROAS — the ingredients of the chart above it. This is the part that makes Aria checkable. You are not being asked to trust a summary; you can look at the numbers it was computed from. Spot-check one row against a dashboard. If week of Jul 05 reads 7,930.80 in ad sales, that figure should appear on the Sponsored Ads screen for the same range and scope. Check the shape. Does the table cover the whole period you asked for, or does it stop short? Check the granularity. A question about weeks answered from daily rows is fine; a question about weeks answered from four monthly rows is not. The row count is a signal on its own The header says how many rows came back, and it is worth reading before you expand anything. | What you see | What it usually means | |---|---| | A plausible number for the period | normal | | Suspiciously small (a handful) | the query was narrower than your question | | Zero, or no query at all | Aria answered from context, not data — treat with care | | Very large | fine, but the summary is doing a lot of compression | A small row count with a confident answer is the combination to watch. Ten rows can support a sweeping statement that the data does not. More than one query A question that needs several sources produces several query rows, each with its own count and tick. Aria joins them for you. That is where to look when two figures in one answer disagree — they may have come from different queries with different scopes or periods, and the join is where the mismatch lives. What the plan and query cannot tell you They do not prove the interpretation is right. Correct data can carry a wrong conclusion. The written paragraph is Aria's reading, not a fact. They do not show the transformation. You see what came back and what was produced, not every step between. They are not the MCP call log. That is a separate audit trail for AI tools querying your data from outside. See MCP Call Logs: Seeing What Your AI Tools Asked. A habit worth forming Before you act on an Aria answer, or put it in front of anyone else: Read the plan. Metric, period, scope. Read the row count. Does it look like the right amount of data? Expand the query and check one number against a screen you trust. Three checks, under a minute, and they catch nearly everything. Related - Asking Aria a Good Question — getting the plan right first time. - Checking Aria's Work When an Answer Looks Wrong — the full diagnostic routine. - What Aria Can and Cannot Answer — the boundaries. - MCP Call Logs: Seeing What Your AI Tools Asked — the other audit trail.

Checking Aria's Work When an Answer Looks Wrong

An Aria answer that looks wrong is usually right about something else. The question, the scope or the period drifted, and the number is a correct answer to a question you did not ask. This is how to find out which, in about two minutes, without asking Aria to try again. Watch the walkthrough Prefer to watch? Ninety-nine seconds on separating a wrong question from wrong data, wrong scope, and an interpretation that overreached. Do not start by re-asking The instinct is to rephrase and hope. It is the slowest route, because you learn nothing about why the first answer was what it was. Start with the evidence that is already on screen. 1. Read the plan The plan and the query, under the answer Every answer opens with an Insights block restating what Aria set out to do — "I'll get you a chart of your account's ad performance ROAS over the past eight weeks." Compare it, word by word, against what you meant: Metric — ROAS or ACoS? Sales or ad sales? Orders or units? Period — the range it chose, against the range you intended. Scope — account-wide, one product, one campaign? Most wrong answers are wrong here, and the plan says so in plain English. See Reading Aria's Plan and Data Queries. 2. Expand "Queried your data" The collapsible row under the plan shows the row count and, when opened, the actual rows Aria retrieved. Check the row count first. A handful of rows behind a sweeping claim is the classic tell. Then spot-check one figure against the screen it should have come from — Sponsored Ads, Sales Performance, SQP. Same brand, same marketplace, same dates. Three outcomes: | Result | What it means | |---|---| | The row matches your screen | the data is right — go to step 4 | | The row does not match | scope or period differ — go to step 3 | | There is no query at all | Aria answered from conversation context, not data | 3. Check the scope you were both working in Two scopes have to agree, and neither is Aria's fault when they do not: The brand and marketplace in the top bar. Aria answers for the current scope. If you switched brands mid-conversation, earlier answers were about a different brand. See Switching Brands and Choosing a Marketplace. The date range you had in mind versus the one Aria chose. If you did not state a period, it chose one, and the plan told you which. And check the data itself is complete. An answer built on a brand whose sync is switched off will be confidently wrong. See Turning Brand Sync On and Off. 4. If the data is right, the interpretation is the problem This is the case people miss. The rows check out; the paragraph beneath them overreaches. Correlation stated as cause. "Spend rose and sales fell, so spend caused it" is a sentence, not a finding. A trend from too few points. Three weeks is not a trend. Attribution ignored. Recent days are incomplete — seven days for Sponsored Products, fourteen for the rest — so a "decline" at the right-hand edge of a chart is often just attribution that has not landed. See Time to Conversion: Checking Your Attribution Window. Amazon's publishing lag treated as a drop. SQP is weekly and late; AMC is monthly. A missing recent period is a schedule, not a fall. See SQP Data Is Sparse or Missing. The interpretation is Aria's reading, not a fact. You are entitled to disagree with it while accepting the numbers. 5. Only now, re-ask If it was the question, fix the question rather than repeating it: Name the metric exactly. "ROAS", not "performance". Name the period. "the last 8 complete weeks", not "recently". Name the scope. "for ASIN X", not "for that product". Ask one thing. Compound questions get compound answers, and the weak half is hard to spot. See Asking Aria a Good Question. The two-minute routine | Step | Look at | Rules out | |---|---|---| | 1 | the plan | wrong question | | 2 | the row count and one row | wrong or missing data | | 3 | the top bar and the date range | wrong scope | | 4 | the written paragraph | overreach | | 5 | re-ask precisely | everything else | When it is genuinely wrong It happens. If the plan is right, the rows match your screens, the scope is right, and the interpretation still does not follow — that is worth reporting. Include the chat, the question as you asked it, the plan text, and the row count. See Getting Help from the TrackIQ Team. Do not delete the chat. It is the evidence. Rename and pin it instead — see Managing Chats. Related - Reading Aria's Plan and Data Queries — the evidence, in detail. - Asking Aria a Good Question — preventing the problem. - What Aria Can and Cannot Answer — questions it will never do well. - Why Don't My Numbers Match Seller Central? — when TrackIQ and Amazon disagree, not Aria.