Overview
AI Analyst and the Daasity MCP Server now carry measures that are built by adding or subtracting other totals, including Total Sales, Average Order Value and Average Unit Retail on Sales Report, and the change, lift and incremental measures on your syndicated and retail views. Order explores built on your own data models get a line-level view for product and unit questions. Year-over-year answers now say how the two periods were lined up, and the AI tells you when a field was left out of a view.
What's new
  • Total Sales, Average Order Value and Average Unit Retail are on Sales Report.
    Total Sales adds together gross sales, discounts, refunds, shipping, tax and duties, and Average Order Value and Average Unit Retail are built on it. These measures had been left out of views because they combine several totals. They now come through with the same definition as your dashboards, and a total with no rows for the period counts as zero, the same way Looker counts it.
  • Change and lift measures on syndicated and retail views.
    On your SPINS, Circana, Nielsen, Kroger and URMS views, measures that compare the latest period with the year-ago period, such as dollar sales change, unit sales change, lift, non-promo sales and incremental sales, come through for the same reason. They appear on views that carry both periods.
  • Year-over-year, lined up the way most brands do.
    A day or a week compares with the same days of the week 52 weeks earlier, so a Monday is compared with a Monday. A month, quarter, year or to-date period compares with the same calendar dates a year earlier. A question asked in your retail or fiscal calendar compares with the same retail or fiscal period last year, when your view carries that calendar. Every answer names the basis it used and gives the date range of both periods. If you want the other basis, ask for it and the AI will use it.
  • Line-level detail for orders built on your own data models.
    If your model has an orders table with its order lines joined to it, a common setup for teams that build their own models in dbt, the AI now gets a second view at the order line level, named for the explore and the order line table. Questions about products, units and line sales answer from that view, and order-level totals stay on the original view so nothing is counted twice.
  • Fields from joins that match one row per order.
    Some explores mark a join as one-to-many even when each order matches only one row in the joined table. Those joins had been left out of views along with their fields, such as an order type. Each one is now checked against your data during the refresh, and when it adds no rows, its fields come through, along with any default filter that relies on them.
  • Know when a field was left out.
    When you ask for a field that couldn't be included in a view on its latest refresh, the AI now says it isn't available on that view and gives the reason. It no longer tells you the field doesn't exist or uses a similarly named field from another view.
What's fixed
  • Views dated only by a hidden field have their date.
    When an explore's only date field is hidden from Looker's field picker, its view had no date at all, so a question about a date range added up every row. The view now includes that date, and its description notes that the field is hidden in your field picker.
What to do
Run Refresh Views from the account menu in AI Analyst. No Refresh Semantic Layer is needed. If your company has several brand accounts, run it on each one. Year-over-year answers and the note on fields left out of a view work now, with or without the refresh.
Then try two things. Ask for last month's Average Order Value and compare it with your dashboard. And ask for last week's net sales against last year; the answer should name the basis it used and give both date ranges.