Connect your customer data. OpsPulse identifies the renewals most likely to churn, explains the reasons, and recommends the next best action. It reads the signals your team already generates — support tickets, QA reviews, survey responses, product usage, and CRM and billing data — plus the external events your own systems cannot see. Telephony is the one source still to land. Then it tells you which renewals are in trouble while there is still time to act.
Designed with the people who own the renewal number
Built from working interviews with Customer Success, RevOps and Support leaders carrying real mid-market renewal books. Every detector on this site exists because one of them said the signal mattered. Design partner names and results appear here as each programme finishes, and not a day before.
The warning signs sat in your systems for months: usage sliding, tickets escalating, the champion gone quiet. They were just in six different tools, and nobody was paid to join them up. By the time the renewal reaches someone's calendar, the decision has already been made on the customer's side.
Renewal signals sit in ticketing, telephony, QA, product usage and billing. Each system sees a fragment; none of them sees an account.
A red-amber-green dot with no arithmetic behind it cannot be argued with, so it gets ignored, and then it turns out to have been right.
One CSM covering sixty accounts cannot read every ticket. Attention goes to whoever shouted last, not to the renewal with the most revenue on it.
Nothing captures what was decided about an at-risk account, on what evidence, or whether the intervention actually moved the renewal.
Five stages, each constrained by the one before it. Detection is arithmetic; the language layer only describes what the arithmetic found. Nothing in the chain is a number a model invented.
Nine detectors compare a recent window against that account's own baseline. Nothing is raised until it clears a declared threshold, so a healthy quarter produces an empty feed instead of invented risk.
Attribution is by decomposition: contributions are shares of a measured decline and sum to 100%. The system never claims confirmed cause; the CSM who knows the account closes that gap.
Exposure is published as a range you can re-multiply from the printed inputs. Some exposures, compliance among them, return no dollar figure at all, because a fabricated one would be the least defensible number in the product.
Committing to a play snapshots the evidence as it stood. The loop closes when the same signals are re-read at renewal and the outcome is reported, including when the account left anyway.
Tickets · Calls · QA · Usage · Billing
Baselines · Anomalies · Renewal risk
Why this account, with the evidence
A named play, a named owner
ARR retained · measured at renewal
Each layer constrains the next. Aggregation decides what is knowable, detection decides what is real, prioritization decides which renewal is worth today's hour, and every recommended play is bound to a named owner, never free text.
What it reads today: support tickets and escalations, QA reviews, NPS responses, weekly active seats against contracted seats, account contract facts, and external events on the account. What it does not: telephony — there is no call data in this build, so the contact-centre panel renders locked on every plan rather than filled with numbers nobody measured. That is the rule throughout: a detector with no source reports itself unrun, and confidence is capped by how much of the book the engine could actually read.
A recommendation is not a decision. A decision is a human committing to a play on an account, on stated evidence, at a stated confidence. The ledger records that commitment (the risk, the confidence and the account's metrics as they stood at that moment), then re-reads the same signals on every pass and reports what happened by the renewal date.
Because the evidence is snapshotted and never back-dated, the platform can be shown to have been wrong. That is what turns "our health score said red" into a renewal forecast a board will accept.
Every renewal tool on the market reads the customer's behaviour inside your product. None of them reads what is happening to the customer. An account can be healthy on tickets, usage and NPS on the Friday and be a different company by the Monday, and no amount of ticket history contains that.
An acquired customer inherits the acquirer's vendor stack. The renewal is decided in a procurement review your champion is not invited to, usually months before the date on your contract.
Surfaced · riskRestructuring, credit events, a freeze on discretionary software spend. By the time this reaches your billing system as a late invoice, the outcome is fixed and the only open question is how much you collect.
Surfaced · riskThe person who chose you leaves, or the programme sponsor is reorganised away. Their login going quiet is the symptom, several weeks late. This one carries the lowest confidence of the four, and is scored that way.
Surfaced · riskThe one that runs the other way. A raise with a headcount plan behind it is a seat-expansion window with a deadline on it, and it is flagged as an opportunity rather than a risk — the same engine, pointed at the upside.
Surfaced · opportunityWhat is real today: the account surface, the four event types, the entitlement gate, the query path and the provenance rendering all run in the live product. Open the demo on a plan that carries External Signals and the panel is there, ranked most recently detected first, each event carrying its type, its confidence and its source.
What is not: the licensed feed behind it. Today's events are sample fixtures, and they say so — each one renders a sample badge and carries no source link at all, because a fabricated URL pointing at a real news domain is a citation that does not exist, and that is a worse failure than no link. Connecting Crunchbase, Tracxn or ZoomInfo is a procurement decision; the schema is built so the wiring is a swap.
And one thing we will not overstate: an external event is surfaced against the account, but it does not yet move the renewal risk score. Fusing it in is a roadmap item, and it is deliberately not a second score sitting next to the first — two numbers that can disagree about the same account is the health-score failure mode with an extra step.
Everyone below is looking at the same accounts and the same arithmetic. What changes is the question they arrive with.
"Which of my sixty accounts needs me this week?" A ranked list with the reason attached, so the hour goes to the renewal that is actually moving, not the customer who emailed most recently.
"Is the renewal forecast real?" Exposure ranges you can re-multiply from the printed inputs, and a ledger showing which risks were caught, which were acted on, and which churned anyway.
"Which escalation is a renewal problem?" Ticket volume alone cannot tell you. Tying queues to contract value shows which backlog is quietly costing ARR and which is merely noisy.
"What is net revenue retention going to be, and why?" One brief, generated from the same objects the CSM sees, so the board number and the account list cannot disagree.
The fastest way to know whether this works on your renewals is to point it at your renewals. No signup, no integration work, no call first. Below on the left is the ninety-second story. On the right is the experiment.
The story the film tells, in the order it tells it.
Northwind renews in six weeks. The health score is green. The CSM has spoken to them twice this quarter and both calls went fine.
Seat usage fell 34% in product analytics. Two integration tickets aged out in support. The champion's last login was in May. Three systems, three fragments, no account.
Against Northwind's own baseline the decline is 3.4σ. It clears the threshold, so it surfaces, with $180k–$240k of ARR attached and 43 days on the clock.
62% of the seat loss sits in the two teams whose connectors broke in the March SSO migration. Confidence 71%. Labelled a hypothesis, because the CSM knows things the data does not.
The play is committed to the ledger with the evidence frozen at that moment. At renewal the same signals are re-read and the outcome is reported: saved, or lost anyway.
Export tickets, QA reviews or NPS responses from the tools you already run, then drop the CSV into the Ops Floor. Your rows are parsed, mapped onto the account timeline and the same nine detectors re-run against them. The feed then changes because of your data, not ours.
One file is a test. Several is the picture. A support CSV on its own carries support signals, so what comes back is escalation load, resolution decay, QA drift and detractor drivers — real findings on your accounts, but not a renewal view. Add a CRM export and the same findings acquire ARR, a renewal date and an owner; add billing and usage and the detectors that need them start running. Every detector that still has no source says so rather than scoring on nothing, and the engine caps its own confidence by how much of the book it could read. You will always be told which half you are looking at.
How much comes back depends entirely on what is in your file, and which detectors can run depends on which signals it carries. That is the point of running it rather than watching a demo, and it is why we will not print an example count here.
Runs entirely in your browser. Nothing is uploaded to a server, nothing is stored, and the sample operation behind it is simulated.
Not a waitlist and not a discount scheme. A design partner shapes what the product detects and how it explains itself, in exchange for putting a real renewal book in front of it early.
The detectors tuned to your renewal motion, the findings on your own book from the first week, and direct access to the founders themselves rather than a support queue.
An export or a read-only connection, an hour a fortnight, and honest reactions, especially when a finding is wrong. A detector nobody argues with is a detector nobody uses.
$3M–$30M ARR, a renewal book someone owns by name, and enough history in your tools for a baseline to mean something, roughly twelve months.
A thirty-minute call and a look at the live product. If the fit is not obvious to both sides in that half hour, we will say so. Scope and commercial terms are agreed on that call, per partner — we are not publishing a standard price for the programme while pricing discovery is still running.
Ten places, taken one at a time so each one actually gets attention. Tell us what your renewal book looks like and where you currently find out too late, and we will tell you on the call what a partnership would involve on both sides.
The shape of the model is decided. The numbers are not, and we would rather say so than publish a figure we picked in a room with no customers in it.
One source is one connected system: a ticketing platform, a telephony feed, a QA tool, a product-analytics stream, a billing system. More sources means more for the engine to reason over, and that is a real cost to run.
How long the decision record stays alive and queryable. A renewal argument you can still reconstruct three years later is a different product from one you can reconstruct for twelve months, and it costs differently.
Unlimited read-only viewers on every plan. A renewal risk only gets acted on if the CSM, the RevOps lead and the CRO can all open the same evidence without someone buying them a licence first. Charging per seat would price the product against its own mechanism of action.
List prices are on hold until we have finished pricing discovery with the first design partners. We are asking them what a caught renewal is worth against what they pay for the tools that currently miss it, and we would rather arrive at a defensible figure late than anchor on a confident one now.
If you want a number for a budget conversation, ask us directly and you will get our current thinking, the reasoning under it, and a straight answer about how firm it is.
This is the same rule the engine runs on itself. Exposure is published as a range you can re-multiply; compliance risk is returned deliberately uncosted rather than given an invented dollar figure. A list price we have not tested would be exactly that invented figure, printed on the one page a buyer trusts most.
Engage renewal owners at mid-market SaaS companies to sharpen which signals actually predict a non-renewal, and which are noise.
A working engine that detects statistically, explains by decomposition and prices exposure as a range. ← you're looking at it.
Ten mid-market teams run it against their real renewal book, and the detectors get tuned to what each motion actually looks like.
Native connectors, scalable architecture and a go-to-market motion aimed squarely at the renewal number.
Every SaaS company has a system of record for what it sold and what broke. None of them has one for why an account renewed: what was noticed, what was decided about it, and whether that decision worked. That is the category, and renewals are the door into it.
Fifteen years across customer operations, customer experience and enterprise SaaS, and more recently executive search. The through-line: the people who own renewals are almost never short of dashboards, and almost always short of a defensible reason for the call they just made. OpsPulse exists to close that gap.
Enterprise AI, architecture and engineering. Owns the decision engine: the statistical detectors, the confidence floor, and the insight contract that stops anything reaching a screen before it can be traced back to a record.
Go-to-market, revenue and partnerships. Owns how the product reaches the operations leaders it is built for — pricing in practice, the partner and investor network, and the expansion into markets beyond the first one. The commercial counterpart to a product that refuses to overstate itself.
Thirty minutes, screen shared, your questions rather than our script. Tell us the size of your renewal book and where you currently find out too late. We will show you what the engine does with signals like yours.
You will be talking to Sudharshan, Founder & CEO, Sachin, Co-founder & CTO, or Venkatesh, Co-founder & CBO. No account managers, no SDR round-trip. They answer everything themselves.
If nothing happened, no mail client is configured on this machine. Copy the message below and send it to infohireloop@gmail.com, or use the direct link.