How Marketing Moves NRR

Nine of the ten top ranking SaaS marketing guides stop at the moment the contract is signed. That is where roughly half the revenue actually comes from, which makes post-sale marketing the largest unclaimed territory in the discipline.

This page is the operating model: which signals to watch, which five plays to run off them, how to decide when a save costs more than it keeps, and how to report the result without a quarterly argument with customer success.

The signal layer: what has to exist before any play runs

Nothing here works without product event data reaching your campaign tool. That plumbing is the project, and it typically takes four to eight weeks.

Two architectures dominate. The CDP route sends events from the product to a tool like Segment, which forwards them to the marketing platform in near real time. The warehouse route lands events in Snowflake or BigQuery, computes signals on a schedule, and pushes audiences back out by reverse ETL. Real time matters for a handful of triggers and not for most, so the warehouse route is the better default for teams who already have a data function.

Define these once, in writing, and put them in your marketing tracking plan alongside acquisition events. The most common failure I see is three teams each computing active users differently and then disagreeing about a dashboard for a quarter.

Play one: seat threshold expansion

Trigger: licensed seat utilisation crosses 85 percent, sustained for seven days. Audience: the billing admin, with the primary champion copied. Message: a plain statement of current utilisation, what happens at 100 percent, and a one click path to add seats at the existing rate.

Success metric is expansion ARR from accounts that received the trigger, within 30 days. Expect a 12 to 25 percent action rate on this play in products where seats are genuinely constrained, because the customer already knows they have a problem.

The failure mode is sending it to the champion instead of whoever holds the budget. Enrichment is worth doing properly here: a message about money going to somebody who cannot spend money produces a forwarded email and a two week delay.

Play two: usage tier upgrade

Trigger: metered usage reaches 80 percent of the plan allowance with more than ten days left in the period. Audience: admin plus the heaviest individual user, because the heavy user is the one who will argue internally for the upgrade.

Message the overage math, not the feature list. Show what they will pay in overage at current run rate versus what the next tier costs, and make the crossover explicit. Customers respond to arithmetic they can check.

Play three: feature adoption

Trigger: an account is below the cohort median on feature breadth after 90 days, specifically missing a feature that correlates with retention. Audience: all active users on the account, not just the admin, because adoption is a behaviour change and behaviours belong to users.

This is the play that looks like marketing and is really education. A two minute recorded walkthrough of the specific feature, sent to people whose usage shows they need it, outperforms a general product newsletter by a wide margin. Measure it on feature activation rate within 30 days and on the retention delta of the cohort at six months.

Honest caveat: feature adoption plays produce the least visible revenue of the five. They show up as retained ARR a year later, which makes them the first thing cut when a quarter is tight, and cutting them is usually wrong.

Play four: the dormant admin save

Trigger: the admin has not logged in for 21 days and weekly active users are declining. This is the highest severity signal in the list and the response should not be an email newsletter.

Route accounts above your named account threshold straight to customer success with the signal attached. For accounts below it, send a short human-sounding message from a real person asking one question: has something changed. Then escalate on reply.

Play five: downgrade instead of cancel

Put a downgrade option in the cancellation flow. Not a retention wall, not five screens of friction, one clearly presented cheaper plan that keeps their data and their account alive.

Downgrade paths typically retain 30 to 50 percent of the revenue a hard cancellation would have lost, and more importantly they keep the account addressable. A downgraded customer can upgrade again when their situation changes. A cancelled one has to be re-acquired.

The tradeoff is real: some customers who would have paid full price downgrade instead. Watch the rate of downgrades from accounts with healthy usage, and if it climbs above about 5 percent of the base, the option is too prominent or your tier gap is too wide.

The save offer decision table

Discounting to prevent churn is the most misused tool in this playbook. Here is the frame I use, and the NRR and churn calculator will run the numbers against your own base.

Situation	Offer	Why	When it backfires
Budget cut, healthy usage	Downgrade or annual prepay discount	Problem is cash timing, not value	If the discount becomes the new list price
Low usage, no champion	No discount, offer onboarding help	Price is not the problem	Discounting hides the adoption failure
Champion left the company	Free re-onboarding for the successor	Relationship loss, not product loss	Waiting until renewal week
Competitor switch in progress	Executive conversation, no blanket discount	Discount signals your price was wrong	Panic discounting spreads through the base
Genuine product gap	Roadmap commitment with a date, or let them go	Keeping them creates a bad reference	Promising a date engineering has not agreed

The arithmetic test is simple. A 40 percent discount on a 2,000 dollar monthly account that survives another eleven months retains 13,200 dollars, minus the cost of serving them, minus the risk the discount leaks into other renewals. If that number is below the cost of acquiring a replacement customer, take the save. If it is above, let them go and put the money into acquisition. The nuance is in the leak, and the leak is usually underpriced.

Win-back sequencing at 30, 90 and 180 days

Thirty day win-backs mostly fail. Nothing has changed for the customer, the decision is fresh, and the email reads as not listening.

Ninety days is the first real window, and the message should lead with what shipped since they left. One hundred and eighty days is the second, and by then a genuine product change or a pricing change gives you something honest to say. At twelve months, treat them as a warm prospect rather than a churned customer and route them into normal acquisition.

Segment churned logos by reason before sequencing. Customers who left over price, over a missing feature, and over a failed implementation need three different messages, and sending one message to all three is why win-back programmes report such poor numbers.

The measurement model that survives a review

Report expansion ARR attributed to the play, with the trigger date, the action date and the upgrade date. Keep the window at 30 days. Report it separately from customer success sourced expansion and resist the urge to build a shared credit model, because shared credit models consume more leadership attention than the revenue they describe.

Netting NRR into one number hides the story. An account base at 108 percent with 20 percent expansion and 12 percent churn is a different company from one at 108 percent with 9 percent expansion and 1 percent churn, and only the second one is healthy. The definitions sit in net revenue retention, and the segmented figures are in NRR and churn benchmarks.

Ownership, and the SLA that makes this real

Marketing should carry an expansion ARR target with its own budget line. Not a contribution percentage, not influenced pipeline, a number the CMO commits to and is measured against.

The boundary with customer success needs to be written down. The version that works: marketing owns one-to-many plays for all accounts and all plays for accounts below the named threshold, CS owns named accounts, and every triggered signal on a named account is delivered to the CSM within 24 hours with context attached. Put response time obligations on both sides.

We had the data and the tools for two years. What we did not have was one line in one document saying whose number it was.

Without that document, post-sale marketing stays a slide in a strategy deck. With it, the plays ship in a quarter.

What to do next

Start with the signal layer, not the campaigns. Pick the three signals above that you can compute today from existing product data, get them flowing into your campaign tool, and confirm freshness for two weeks before you send anything.

Then ship one play. Seat threshold expansion is the usual first choice because the trigger is unambiguous and the revenue is immediate. The message level detail is in expansion revenue email campaigns and expansion marketing plays, the segmentation mechanics are in lifecycle email segmentation with product data, and the broader revenue model sits in B2B SaaS expansion revenue marketing. For the metric definitions underneath all of it, start at SaaS metrics and analytics.