Kinds of insights

Insights are grouped into eight kinds. Each kind can be switched off in Insights settings if it isn't useful to you, and owners and admins can hide a kind straight from any insight's Hide menu.

The last two describe the shape of an insight rather than its topic: a change in the UK's churn is both churn and a change in a group. Switching off either kind hides it.

Leads, trials and new customers

Examples: "Leads below normal for 3 months", "New customers up 38%, led by Business", "Trials started 1.6× normal".

These follow your funnel from the top: leads, trials (if most customers start with one), new customers and new MRR. The insight leads with the first stage that moved and says what happened further down. A drop in leads where new customers held steady means conversion improved; a drop in new customers where leads were normal means conversion fell.

What to do: open the linked report and look at the driver the insight names. If one channel or region explains it, start there. If leads fell but customers didn't, you may be getting fewer but better leads, which is often fine. If you know what caused it, like a campaign you stopped, press We know why and add it to your timeline.

Expansion and reactivation

Examples: "Expansion MRR 2.0× normal, mostly discounts ending", "Reactivation MRR 3.2× normal", "Expansion MRR has fallen for 6 months".

Expansion insights say what kind of expansion drove the change: upgrades to another plan, more seats or quantity, price increases, or discounts ending. A spike that is "mostly discounts ending" is scheduled revenue arriving, not customers choosing to pay more.

What to do: for a drop in real upgrades, look at whether fewer customers reach the point where the next plan makes sense. For a spike from discounts ending, watch churn in the following months: that's when customers who only stayed for the discount leave.

Churn and contraction

Examples: "Churned MRR up 41%", "Churned MRR 2.1× normal, led by the UK", "Contraction MRR 2.2× normal", "Churned customers in DACH 2.6× normal".

Churn insights combine churned MRR and churned customers into one story, and say where it came from: a region, a plan, a billing interval, or failed payments. When your customers give reasons when they cancel, the most common reason is named, and the insight says if it's clearly more common than usual. If one customer is behind most of the churned MRR, the insight says so: that's an account to call, not a trend to fix.

Two related kinds of insight come from cancellation reasons: "Customers now give a reason when they cancel" (from then on, churn insights include reasons), and a reason becoming clearly more or less common than in the six months before.

What to do: open the linked MRR churn report filtered to the driver, and read the cancellation reasons and comments of the customers behind it. Talk to a few of them. A churn spike that's back to normal the next month was often a batch of annual renewals or a billing cleanup; one that continues is a pattern worth acting on.

Net new MRR

Examples: "Net new MRR 64% below normal".

Net new MRR is the sum of every movement: new, expansion, reactivation, contraction and churn. When one movement explains most of a change in net new MRR, the net change is part of that movement's story. It's only told on its own when several movements add up to it and none explains it alone.

What to do: open the MRR movements table for the month and see which movements were off. Usually two or three of them each moved a bit in the wrong direction.

Payments

Examples: "Unpaid is 39% more than usual: 69 customers owe you $10.7k a month", "7 customers have been unpaid for over a year and still count in your MRR", "184 failed payments from September were paid in the end", "Failed payments 1.7× normal", "One-time revenue 3.4× normal", "Stripe fees up to 3.4% of the money that came in, led by Card processing".

• Unpaid customers come from the money at risk right now, not from a close. Every business has some customers in dunning, so this is only told when unpaid MRR, as a share of your MRR, is clearly above its usual level over the last year. The insight splits it by how long customers have been unpaid, says how often invoices that late were never paid in your account, and tells you which group to email first. It updates during the day, and ends when unpaid is back to its usual level.
• Unpaid for over a year, still in your MRR is told when your settings keep past-due customers until they're cancelled in your billing system. Those customers inflate your MRR and every report built on it. See Revenue at risk.
• Failed payments recovered is good news: last month's failed payments that were paid in the end, when clearly more than usual. It uses the same numbers as the Failed payments report.
• Failed payments follow the Failure rate report: the share of invoices with a failed payment in the period. Only more failures than usual are told.
• One-time revenue follows the cash flow report: successful one-time payments, net of the discounts on them.
• Stripe fees, for accounts with Stripe payouts, follow the effective fee rate in Payouts › Fees: the share of the money that came in that went to Stripe. A rise is told at a month-close once it has held for a while, with the fee type behind it and the extra fees it costs you. After that it's the new usual, so a change in Stripe's pricing is told once, not every month.

What to do: for unpaid customers, open Revenue at risk and start with the customers who are 8 to 30 days late: still likely to pay, but not for much longer, and the moment a personal email does the most. A failed-payment spike for one week is often a card network problem or a batch of expired cards, and dunning recovers most of it. If it continues, look at your payment provider's retry and card-updater settings. For a rise in Stripe fees, open Payouts › Fees in percentages, grouped by fee type, to see what changed and since when.

Coming up

Examples: "Discounts end for 179 customers in December, adding about $13.2k in MRR", "$9.1k of MRR is already scheduled to cancel in November", "14 annual renewals in the next 30 days, 2.1× a usual month: about $48k to collect".

These look ahead at what's already scheduled, with the numbers from CMRR and Scheduled churn:

• Discounts ending: the month in the next three with the biggest scheduled increase from discounts ending, when it's big enough to matter and clearly more than discounts ending usually add in a month. As in CMRR, the increase lands on the invoice that first charges full price, so an annual plan's discount counts in the month it renews. This is good news, so it gets Noted rather than Working on it.
• Scheduled cancellations: next month already has at least a normal month's churned MRR scheduled to cancel, before any cancellations that aren't scheduled yet.
• A busy renewal month: clearly more annual renewals in the next 30 days than in a usual 30 days. The insight says how much there is to collect, and how many of the renewals are at risk because of a failed payment in the last year or a card that expires first.

Coming up insights are refreshed at every close (renewals during the day) and are always about what's ahead, so they never pile up in your history.

What to do: for a discount wave, tell those customers before the first full-price invoice. A heads-up gives them time to decide, instead of cancelling in surprise when the bill arrives. For scheduled cancellations, reach out to the biggest accounts while they're still customers. For a busy renewal month, open the renewals at risk on the Revenue at risk tab and ask those customers to check their card before the invoice goes out.

Changes in a group

Examples: "New customers on Team 4.1× normal", "Churned customers in DACH 2.6× normal", "Expansion MRR on Team below normal for 2 months".

A change in a group is told when one region or country, plan, or billing interval moves clearly differently from the rest of your business, even if your totals barely moved. The chart shows the group and the rest, each against its own usual level.

What to do: ask what's different about that group this period: a campaign, a partner, a price change, a competitor, a local event, a payment method. Ask AI is good at the first round of digging. See How insights are found for the rules.

Long-term differences

Examples: "Customers who started on Team now churn 16% more than everyone else (was 36% more)", "Customers on annual billing now lose 2.0× as many of their failed payments as everyone else", "Customers in Southern Europe now expand 43% more often than everyone else".

These are the moments a long-standing difference between a group of customers and everyone else appears, changes or ends. They're found at month-closes by comparing each group over the last year with everyone else of the same customer age. A difference that holds without changing isn't news, so it's only told again when it changes or ends.

What to do: a difference usually tells you who your customers are more than what to change. Customers on annual billing churn less partly because customers who pick annual billing were more committed to begin with, so moving everyone to annual billing wouldn't give you their churn rate. Use differences to decide where to look, to set expectations for each group, and to spot a group that's getting worse. Read Long-term differences before acting on one.

Other notes

• "Fewer new customers start with a trial" is told once when trials stop being how most of your customers arrive. From then on, insights follow leads and new customers instead of trials.
• An insight about a week covers the account's week, starting on your first day of the week, in your time zone. An insight about a month covers the calendar month in your time zone.