Currently in alpha. Sign up free. Expect a few rough edges.
Docs

The inbox

The questions Overscore has for you, one at a time, each answered yes or no.

The inbox is where Overscore asks you things. Everything waiting is one list, one row per thing that happened, and every row is one question:

  • The first line says what happened, then asks. "Pipedrive has sent nothing for 3 days. Check the connection?"
  • The second line says what it means for you. "That's longer than usual. The numbers it feeds stay where they were until it sends again."
  • At the right, how long it has waited, then the same three answers on every row: Yes, No and Talk about it.

The three answers

  • Yes does what the sentence says.
  • No answers no to this question and nothing else. The card closes. It does not stop the next one: if the thing fails again, it asks again.
  • Talk about it opens a chat on the card, under its buttons. It already knows what the card knows, with the run and its log when there is one, and for something that broke it starts with what the agent found when it looked into it. Say yes or no there and it presses the button for you, once, as you, and only when you say so.

Whether a thing tells you at all is set on that thing: a source's Settings tab, a model's page, or Settings, Alerts. No button on a card changes it.

A row with something to fill in first, a new key or which person is meant, says Open instead. So does a row with several things on it.

A question nobody answers stays here. Nothing sets it aside, and nothing answers it for you. Its age turns amber once it has waited a week.

Find what needs you

Each row has an icon for its kind, with a red mark on it when something stopped.

The rows come in one order. Something that stopped first. Then the questions your numbers depend on. Then suggestions. Inside each, the one that has waited longest is first.

The two tabs carry their counts: how many questions wait on you, and how many things were handled for you this month. The line beside the title says the second too: "12 were handled for you this month."

The count is of rows. It is the same number beside Inbox in the menu and on your phone, and it hides at zero.

A row can stand for several things. The eighteen things heard in one interview are one row, not eighteen. The five questions one day's work asked are one row.

Click a row and it opens in a panel on the right. Every panel has the same shape. The question. At most two lines of what it means. The three answers. Show how it's worked out, a closed line with the workings and what each answer does. Who was told.

Rows come 25 at a time. A list longer than that has a search box and next and previous pages. On a shorter list the search is behind Search, under the rows.

Check for problems now, also under the rows, looks for a failed sync or a source gone quiet right away. It looks on its own every few minutes.

A row you answer leaves the list at once. If the list cannot load, it says what went wrong and offers Retry.

With nothing waiting the page says so in one line: "Nothing needs you. When something does, it shows up here as a question."

The kinds of questions

Keep this fact?

Something heard in the interview, a chat, a note, or from an agent. The card says the fact, then "Keep this fact?"

Under it, since when, in plain words, with Change beside it. If it replaces something on file, one line says what.

Add a note is a line for your own words, like "that was the price at the time". The note is kept with the fact and read with it.

Yes makes it a confirmed fact with your name on it, from that date. Only a person can confirm. Nothing a model writes is ever confirmed on its own.

No leaves it out of answers. On a fact that would replace another, what is on file stays and the card's words go on its record.

Whom it is about and what it says are behind Show the details, already filled in from the card. Where it will be saved and its key are under Advanced there.

Keep them?

Everything heard in one interview or one chat is one row: "18 things heard in your interview. Keep them?"

Its panel lists each thing with a tick, the first six and a link to the rest. Yes keeps the ticked ones, each exactly as a yes on its own would. Anything unticked stays waiting on the same row. No keeps none of them.

One that needs more than a yes (which Sam is meant, a detail it could not fill in) starts unticked. Click its words to open it on its own, and use the link at the top to come back.

Questions about your numbers

A metric nobody has confirmed is a guess: a definition the AI made in a chat that nobody has signed off. Every answer built on it carries a warning until someone does. Before you confirm one, Overscore looks at its data and asks what the words do not settle: "Some people signed up more than once with the same email. Count each person once?"

The questions one day's work asked are one row: "5 questions about your signups and subscribers. Asked October 24. Until you answer, those numbers use a best guess." Open shows them in the order they were asked, one at a time, each with the same three answers. An answered one shows its answer with a tick and Change. The ones still to come wait under it.

There is no Confirm button apart from the answers. The last answer is the confirm: when you answer a metric's last question and you are an admin or a named confirmer, the metric is confirmed and counted again with your answers. A metric with no question left asks the one that confirms it: "Is this how Signups is worked out?" Yes confirms it. No opens its page, where the definition is changed.

A metric whose model was rewritten after someone confirmed it asks whether it is still right ("The model it reads, westside.daily_sales, changed on Oct 20. Is it still right?"). Nobody is emailed about questions. The metric's own page still has its questions box, for a person who goes there first.

A metric a source proposes

When a source's first sync lands, each of its obvious metrics is a row ("A number to confirm: Mrr. Is this how it's worked out?") whose panel has its definition, last month's figure with Where from, and the choice it makes when it makes one. An admin or a named confirmer can say yes.

Yes makes it a metric with a receipt, and No removes it for good.

Something that stopped

A sync that failed, a key that expired, a source gone quiet. The first line says what happened and the question is the fix: "Customer value failed its last run. Pause it until it's fixed?" Yes does that fix, and the card closes itself when the next run works.

Behind Show how it's worked out is the agent's own message, written from what it found: what failed and when, what it read and what looks wrong. A failure it already ran again says so ("I ran it again and it failed again").

When AI on its own is on (Settings, Company), the AI reads the run's error and its log to say what looks wrong. When it is off, or at its ceiling for the month, a rule of thumb picks the fix and the card says so ("a rule of thumb, since AI on its own is off").

A key that was refused has a box above the answers. Yes tests the new key and saves it. A new key goes in there and never into a chat.

A source that has sent nothing for longer than it usually goes quiet is a question too: "PostHog has sent nothing for 19 hours. Check the connection?" Yes asks the source for anything new. If nothing new comes, it pulls again from the start, in case something was missed. No closes it, and it comes back only once the source has sent again and then gone quiet again.

A model whose checks held its new rows (a month already reported that changed in this week's file, months that no longer add up to the year to date) asks "A check held back 11 new rows of Launches. Are those rows fine?" Yes, for a person who may change that data, puts the new rows in and checks the next file against them. No leaves the table as it was, on the numbers it had.

The card says whom it reached: "Sent to Rachel Moore by email, 1 phone and Slack."

A model still being built

A model that has never run clean does not reach the Inbox. Neither does one whose newest version was saved in the last hour. Whoever is building it hears from the run itself: the answer in their chat or terminal, and the model's page. Once it has run clean and nobody is at it, a failure is a question here.

A query people keep running

The same query run many times in a month is a card: "Keep it as a model?" Yes makes it a model (a view, on Free), so everyone reads the same rows.

Still right?

When a finished answer has used something it only heard and nobody confirmed, it asks once, for that version: "... was used in an answer. Is it still right?" Something it heard over a year ago that no answer has used is asked the same way, ten a night.

Yes confirms it. On Never ask (Settings, Company) it asks nothing.

A page number close to a new metric

A metric defined after a page works something out its own way, close but not the same: one card with both figures, asking "Use the metric for it?" Yes switches the page's number to the metric. No keeps the page's own calculation for now (about that version of the metric only).

To make the page's calculation a metric of its own, press Define them on the dashboard's Definitions tab. The builder's chat reads it back to you and defines it on your yes. For a card that draws a query's rows, it asks which total inside the query comes first, as buttons.

An old answer close to a new metric

A chat answer's own SQL that nearly computes a metric defined later (a filter left out, another date) asks "Count them as the metric's number?" Yes counts those answers as the metric's. No leaves them as they were.

Facts from a document

When a deck, a contract or meeting notes you keep in Warehouse, Files is read, it suggests up to five facts, each said as the document's: "Ben's plan sets Riverfront's target: full run-rate by Q6", "Ben's plan requires...", never a plan's target as if it had already happened. They are one row: "4 facts from Ben's plan. Keep them?"

A price, a count or a date your tables already hold is never suggested. A price is true one day, and the table keeps the day. The document still has it, and a chat that reads the document can find it.

Each fact is a line with a tick, and with its page where the file has pages. Open one on its own for the sentence, the words it rests on, and a link to the file there. Where it is filed is under Advanced.

Only people who can open the file see them. Say yes to one and it's a fact about the company: The company in Context lists it, the chat answers from it, and its source is the file and the page.

Handled for you

The Inbox's second tab has no buttons. It is what asks nothing, so nothing vanishes and nothing waits on you.

What is still true comes first, with its age. It closes itself the moment it stops being true:

  • A number you asked to be watched that crossed its line: "Net sales last week below $60,000. It reads $56,702." You hear about it by email or on your phone the moment it happens. Its panel has what it looked into, a few sentences written by AI about why when AI on its own is on, and Open the alert.
  • A live dashboard the nightly check found no longer answers, with Open the dashboard. It closes itself when the page answers again.
  • A metric whose query no longer runs, with what is wrong in plain words ("Can't be worked out: what it reads has no column called credit_memo_total") and Open the metric.
  • A source or a model that fell behind its schedule, one row however many times it happens.
  • A failure with no fix to offer. You hear about it, and it closes itself when a run works.
  • The plan's month nearly used.

Then what happened this month, newest first:

  • Numbers that changed after the month was over. "8 numbers changed on November 2. Rows dated in October came in late. Orders, Revenue, Customers, Repeat customers and 4 more. Nothing broke." Its panel is a table of what each was, what it is now and the change. Your dashboards already show the new numbers.
  • A metric with no date column has only its whole number to watch, so a change of more than 2% from last time is listed the same way.
  • A number that grows because the business keeps going is not a change. Signups go up every day. A metric with a date column is watched month by month, and a month that changed by less than half a percent is accepted without asking.
  • A sync, alert or email that stopped because the key or person that set it up is gone.
  • Feedback on an answer, for an admin who may read the chats, with Open the chat.
  • A table that fell behind 26 times and caught up each time.
  • A failed run that worked the next time.
  • A dashboard's checks of older versions replaced by a newer one, and a draft's checks.
  • An alert that came back inside its line.
  • A number added up the same way in several places with no metric for it yet. The chat that builds pages reads that one, so a new page adds the number up the same way.

The figures show to people who can read the data behind the metric. Anyone else sees which months changed.

Click a row to see what was said each time, or the months with what each was and is. There is nothing to press.

The count on the Handled for you tab, and the line beside the title, count every time, so a table that fell behind 26 times is 26. The month is your company's own, by its time zone.

Checks from teammates

Answers a teammate sent you to check (Ask someone to check, under any chat answer) sit at the top. A check can be sent to a teammate who has never signed in here or whose current access cannot be verified; known insufficient or revoked access is refused. Opening and answering always require current access to all the original tables, without masking, including those behind its metrics. Open one here or with the email's Open review button, see exactly what it rested on, and say looks right, right with a nuance, or wrong and why.

Improvements

Improvements, a quiet link under the rows, is what people noted and nobody has made yet, grouped by what it's about (tables, models, metrics, dashboards, sources).

Each is a row: what to improve, then where it lives, who noted it and why, with the day it was noted and its status (Open, In a session, Done) at the right.

The assistant and agents add what they spot but aren't fixing now. You can add your own.

Click one to see where it lives, where it came from ("from Sarah's chat, Sep 29"), and what happened to it.

Work on it in a chat starts a chat with it. Only once that chat exists does the item say "in a session". The assistant looks into it and says what it would change before changing anything.

One about a model that failed opens already holding that model: its saved SQL, what its last save changed and the failed run's error. The first reply says what broke and proposes the fix. Nobody is asked to find or paste the SQL.

Come back next week and choose Resume the chat to pick up that conversation. Start a new chat is a separate choice when you want a fresh conversation. Put it back returns the item to open and keeps its chat for later.

With several open in one area, Work on these in a chat starts one new chat for those items.

Mark it done with a line on what changed, and it stays as history. One about a table or a model shows only to people who can read that data.

On a phone, each item stacks what needs doing, who noted it and when, and its full status, so you can read it without sliding across a table.

Cards by email

A card can be addressed to whoever owns that number. They get one email with what it says and a link that opens that card in the app. A metric whose finished months changed emails its owner once, when its row is made, not once for each month.

What to say to it

Cartridge reorders dropped 40% because we moved the reorder button. Keep that.
Is the refund number in the board deck still right?
Ignore this; the test orders aren't real.

All docs · Something wrong on this page? Tell us.