Changelog
Follow up on the latest improvements and updates.
RSS
Overview
Earlier this month we added session limits following a third-party penetration test: a session ends after 30 minutes without activity and after 8 hours regardless. Those limits are meant for people who have stepped away, but they were also catching people in the middle of working in a dashboard. This release puts them where they belong. Working in a dashboard keeps you signed in, and if you are signed out after stepping away, each tab brings you back to the dashboard and filters you had open.
What's fixed
- Working in a dashboard keeps you signed in.Changing filters, drilling into a tile and editing a dashboard now count as activity, so you won't be signed out in the middle of using one. Sessions still end after 30 minutes with no activity, and after 8 hours no matter what.
- Each tab comes back to where it was.When you sign back in, every tab returns to its own dashboard with the filters you had applied, instead of the home page.
- One sign-in covers every tab.Sign in on one tab and the others sign in with it, each reloading to its own dashboard.
- Dashboard filters are in the address bar.Reload a dashboard, or bookmark it, and it opens with the same filters.
What to expect
No setup is needed. Reload any dashboard tabs you already had open so they pick up the change.
The 30-minute and 8-hour limits are part of our security commitments to you. They are the same for every account, and they only affect people who have stepped away.
If you spend more than 30 minutes editing one dashboard without saving, you may be asked to sign in again when you save. The save still goes through, and you return to the dashboard.
Overview
AI Analyst and the Daasity MCP Server now carry the measures your team customized the way you built them. When one of your measures adds to or subtracts from Daasity's standard definition, the AI describes it with your adjustments included. When a measure is built on a split, like spreading a bundle's sales across the products inside it, it gets a view of its own, so it adds up correctly without inflating anything else.
What's new
- Your adjustments are part of the definition.If your version of a measure adds or subtracts something that Daasity's standard version doesn't, the AI's description now says exactly what it adds and what it subtracts, right after the standard formula. Ask "how do you calculate net sales?" and the answer gives the standard formula followed by your adjustments, so your number is never mistaken for the standard one. Available on Snowflake accounts today.
- Measures built on a split get their own view.Some data models split a single order line into several rows on purpose, for example to spread a bundle's sales across the products inside it. Measures built for that split now live on a companion view, named "split by" the table that does the splitting, where they add up correctly across the split's fields. The original view keeps one row per order line, so every other measure stays accurate. The AI picks the right one from your question: ask for net sales by product line and it answers from the split view, ask for total net sales and it answers from the original.
- Joined tables written in multiple steps come through.Custom tables in your model that are written as multi-step SQL and join to a view had been treated as missing, which took every field that depends on them out of the view. They now ship with the view, along with those fields.
What to do
Run
Refresh Views
from the account menu in AI Analyst. No Refresh Semantic Layer is needed for this one. If you haven't run the refresh from today's other updates yet, running Refresh Semantic Layer and then Refresh Views once covers all of them.Then ask the AI how one of your customized measures is calculated and check it against your own definition.
Overview
AI Analyst and the Daasity MCP Server now answer from your dashboards the way you see them in Daasity: with the dashboard's filters, the default scope your data model sets, and any filter you name in your question. Every answer tells you which filters it used, so you can check it against the dashboard at a glance. Refreshes are also much faster on large accounts, more of your measures come through, and personal customer information stays out of answers.
What's new
- Dashboard filters carry through.When an answer comes from one of your dashboard tiles, it now runs with that dashboard's filters, so the number matches what you see on the dashboard. For example, if a dashboard is filtered to your online store, asking "What were my top products last quarter?" answers for the online store instead of adding every channel together. Answers also list the filters that were applied, keep the dashboard's totals row out of ranked results, and tell you when a tile's row limit cut the results short.
- Your filter wins.Ask for a different value than the dashboard's default, and the answer uses yours. For example, "top 10 brands, latest 52 weeks" from a tile that defaults to 12 weeks now runs for 52 weeks and says so. If you ask for something the tile can't filter on, like a region on a tile with no region filter, the AI tells you the tile can't answer that way rather than returning the tile's default numbers under your label.
- Default scope, applied every time and stated.Your data model can say how a view should be filtered unless someone asks otherwise, like ranking retail brands against your competitive set instead of the whole category. Answers now apply that default consistently and say which scope was used. For example, "top 10 brands, last 52 weeks" returns your competitive set with a note that it's scoped that way, and adding "across the whole category" returns every brand. The AI also won't add up markets or time periods that are meant to be picked one at a time. This works now, with no refresh needed.
- More of your measures come through.Averages, medians and measures calculated from other measures in your data model are now part of your views, so the AI uses your definition instead of working out its own version.
- Column names you recognize.Tables in AI Analyst show your field labels, like "Net Sales," instead of the underlying field name.
- Faster refreshes.Refresh Views on accounts with very wide reports used to take hours and now finishes in minutes.
- Personal customer information stays out of answers.A customer's name, email, phone number, street address and IP address are never returned. City, state, zip code and country are, so regional reporting keeps working. Business locations and products aren't treated as personal, so store, warehouse and distribution center addresses and full product names come through. For example, "net sales by shipping zip code last quarter" returns zip codes, while a request for your top customers' emails comes back with that field withheld and a note saying so.
- Know when a measure isn't available.If a measure can't be built on a view, your AI can see which one and why, and tells you instead of substituting a similar measure. For example, ask for lifetime net sales on an order-level view and the AI explains that it lives on your customer views, rather than adding up a different number under that name.
What to do
- Run Refresh Semantic Layer, thenRefresh Views, from the account menu in AI Analyst. Refresh Semantic Layer picks up your dashboard filters, and Refresh Views rebuilds your views with everything above. Refresh Views was previously called Refresh MCP Views.
- Ask a question you already know the answer to from a dashboard, like your top 10 products by net sales last month, and compare the two.
This is the same refresh that brings back the fields in today's fix post, so you only need to run it once.
Overview
A handful of fields, views and dashboards weren't reaching AI Analyst and the Daasity MCP Server. All of them are back with your next refresh, and when a field can't be built on a view, the AI now tells you so instead of leaving it out quietly.
What's fixed
- Your fields stay on your views.Customer details like first order store type, SKU attributes like category or product line, and fields your team customized, like a custom order flag or a custom revenue measure, could go missing from a view after a refresh. They're back. Ask for net sales by product line or by first order store type and the breakdown comes through.
- Number attributes come through as numbers.A SKU attribute that holds a number had been showing up as a date. It now comes through as a number, so you can group and filter by it.
- Custom views written your way are included.Custom views whose SQL includes comments or is written in multiple steps had been skipped. They're now part of every refresh.
- Dashboards in Shared folders are included.Dashboards saved in Shared folders weren't being picked up. Ask about one and the AI can answer from it.
- SPINS Latest and Omni Agg are available.Both views had been built but couldn't be used by your AI. They're available now.
What you may notice
- A customized field may use our standard version on one view.Some customized fields can't be built on every view. A flag that looks at individual order lines, for example, can't be built on Sales Report, which has one row per order. On a view like that, the field uses Daasity's standard definition, and its description says which version it is and why. Views at the order line level use your version. If an answer doesn't match your dashboard, ask the AI which version it used.
- Customer totals come from customer views.Totals calculated per customer, like lifetime net sales, are on your customer views. On order-level views like Sales Report and LTV they're listed as unavailable, because a customer with five orders would be counted five times. Ask "lifetime net sales for customers acquired in Q1" and the AI answers from your customer view.
What to do
- Run Refresh Semantic Layer, thenRefresh Views, from the account menu in AI Analyst. Refresh Semantic Layer picks up Shared folder dashboards, and Refresh Views rebuilds your views with the fields above. Refresh Views was previously called Refresh MCP Views.
- Check a field you use often, like a SKU attribute or first order store type, by asking for net sales broken out by it.
This is the same refresh as today's improvements post, so you only need to run it once. If a field you rely on is still missing after the refresh, contact Support.
Overview
Our new Instacart Ads integration is now in Beta. If you advertise on Instacart, you can bring that spend and performance data into Daasity directly from Instacart, with no more exporting and uploading files. You connect it yourself from your New Integration screen.
What's new
- Instacart ad performance, loaded daily.Spend, impressions, clicks and attributed sales for your Sponsored Product, Display, Video and Inspiration campaigns, broken out by campaign and ad group.
- Product-level results for Sponsored Products.Spend and attributed sales are loaded for each advertised product.
- Self-serve setup.Instacart Ads is on your New Integration screen under Ads & Marketing Spend.
- History on demand.Load past spend reports from the integration's page once it's connected.
What to expect
Instacart requires each advertiser to register their own API application, so setup includes a few steps in Instacart Ads Manager. You'll need a user with API Authentication enabled and Read access to Customer Insights Reporting.
Attributed sales are the numbers Instacart reports, using last-touch attribution. Instacart does not report order counts, so orders are not included.
Instacart Ads Integration Setup walks through each step.
Overview
Our new Walmart Connect integration is now in Beta. If you run Sponsored Search ads on Walmart, you can bring that performance data into Daasity directly from Walmart, with no more exporting and uploading files. You connect it yourself from your New Integration screen.
What's new
- Walmart ad performance, loaded daily.Spend, impressions, clicks, orders and sales for your Sponsored Search campaigns, broken out by campaign, ad group and advertised item.
- Self-serve setup.Walmart Connect is on your New Integration screen under Ads & Marketing Spend. All it asks for is a name and your Walmart Connect Advertiser ID.
- History on demand.Load past performance from the integration's page once it's connected.
What to expect
Walmart requires each advertiser to approve the partners that can read their advertising data. Before you connect, an admin on your Walmart Ad Center account adds Daasity as an API partner with read access. Then choose Walmart Connect in Daasity, enter your Advertiser ID and click Create.
Orders and sales are the numbers Walmart reports. This first release covers Sponsored Search. Display ads are not included.
Walmart Connect Integration Setup walks through each step.
Overview
Notifications in Daasity have been rebuilt. Every notification now tells you whether it's yours to fix or ours, explains what happened in plain language, and gives you one clear next step. A problem that keeps happening shows up once and clears itself when it's fixed.
What's new
- Grouped by who needs to act."Needs you" is what's yours to fix. "We're on it" is what Daasity is already handling. "Recent" covers one-time events, like a file that couldn't be loaded. See every notification type and what it means.
- One notification per problem.Repeats collapse into a single notification ("Seen 5 times") and send one email when the problem first appears, so expect far fewer emails. See when emails are sent and who receives them.
- Notifications clear themselves.When an integration recovers, its notification moves to Resolved with the reason. "Mark as solved" is gone. Use "Got it" to show your team you've seen a problem, and "Dismiss" for one-time events.
- One clear next step.Each notification has a single button for what to do, like Reauthorize, Check connection, Reload history or Review. See what each button does.
- A full Notifications page.See everything in one place, with tabs for Needs you, We're on it, Recent, Older and All.
- Skipped data alerts.If a data refresh finishes with some data skipped, you're told that reports may be incomplete.
What to expect
Your notification list started fresh on October 5. Notifications from the old system were cleared automatically, and you can still find them under Resolved on the All tab for 30 days. Anything still failing has come back as a new notification. If you see one under "Needs you," that integration needs your attention. No other action is required.
Learn more
Overview
AI Analyst and the Daasity MCP Server now answer using the definitions written into your data model, and tell you which ones they used. Ask about "DTC," "net sales" or "refunds," and the answer quotes your definition and names the filter it applied, so you can check the logic at a glance.
What's new
- Your definitions in every answer.Descriptions, dimension definitions, the values a field can take, and what each metric counts now reach the AI. Answers use your vocabulary and show their work.
- Refresh your views yourself.Rebuild your views with your latest definitions fromRefresh MCP Viewsin the AI Analyst account menu. It runs in the background: usually 5 to 15 minutes, up to an hour for larger accounts.
- SKU attributes, found automatically.Ask about scent, category or product line, and the AI finds and uses the right SKU attribute without being pointed to it.
- No more stuck questions.If a question can't finish, you get what's available within about a minute, marked "Partial."
- Gemini CLI support.Connect Gemini CLI alongside Claude, ChatGPT, Codex and Cursor. Setup guide
- A security overview for your IT team.What the connection can and can't reach, how sign-in works, and how access is limited and revoked. Read it
What to do
Definitions appear in views rebuilt since September 28. Run
Refresh MCP Views
once, then ask a question you already know the answer to and compare it with your dashboard.Overview
Loading past Pinterest data now brings in the full date range you ask for, so your Pinterest spend and performance history is complete.
Issue
Loads of past Pinterest data could miss roughly 24 days at the start of the range you requested. Some reporting periods were also dropped when Pinterest was slow to respond. New Pinterest connections could only load two months of history.
Solution
History loads now cover the full range you request and wait for Pinterest to finish before moving on. A new Pinterest connection can now load up to 900 days of history, the most Pinterest makes available.
What to expect
If you loaded Pinterest history before September, run the history load again to fill any gaps. New loads are complete going forward.
Overview
Ad spend for AppLovin, Snapchat and TikTok Ads now matches what you see in each platform, so your marketing efficiency numbers (ROAS, cost per acquisition and similar) are based on the right spend.
Issue
AppLovin spend could be overstated after an ad was renamed, because the renamed ad was counted as a new one. Some Snapchat spend could be counted twice. TikTok Ads spend was being rounded, which dropped the cents. AppLovin history also only reached back 45 days.
Solution
All three are corrected, and AppLovin history now reaches back 88 days, the most AppLovin makes available.
What to expect
Spend from here forward is correct. Earlier AppLovin and Snapchat spend isn't corrected automatically, so if spend on either platform has looked high, contact Support and we'll reload the affected dates.
Load More
→