Onboarding to First Value in 30, 60 and 90 Days: A GUARD Activate Plan, an Early-Churn Check and a Day 90 Health Score

In my view, many onboarding plans are implementation plans: they track the vendor's tasks and stop at go-live. The customer bought something else: a faster close, a report on time, revenue out of a spreadsheet. Migrated data that ties out is none of these. It shows the old numbers survived the move, so call it readiness. First value is the first result the customer bought and did not have before.

A stalled account can look like a training gap. After a live test of the GUARD skills on a sample 40-account portfolio, I wrote: "A $97K account at 0.5% adoption was not treated as a training issue. It became a failed deployment and re-onboarding case." (My comment on my 10 May 2026 article.) A sample, not a client result.

In finance software the calendar decides when value can show: a close each month, a year-end and an audit each year. A plan that ignores it puts the first live close on the busiest week of the year. Outside finance, swap the close for your product's first real cycle, and the calendar for your customer's heavy dates.

The GUARD customer success guide picks up on established accounts; the Day 90 exit test is the hand-off.

The method is a checked handoff, a readiness gate, a first value the customer bought, a weekly milestone list and a written exit test. Two free Customer Success plugin skills and one optional CRAFT GTM tool save drafting. Treat each output as a first draft and check it.

The order follows the skills. The onboarding skill says "Activate ALWAYS takes precedence for accounts under 90 days old", and the health check says "default to onboarding-plan instead: baselines are not established". So the score comes last, and the milestone list watches until then.

Step 1. Build the 90-day plan from the sales handoff

About this example. Rillet is a real company. Rillet has not reviewed this guide. Facts about Rillet come from public sources, linked alongside with the date I checked them. Every number marked illustrative, and every customer, AE note, customer remark, churn reason, delay and product module in the example, is an input I chose to show the method. None of it is Rillet's data or a Rillet customer. This is a worked example, not a client result, and the tool outputs are exercise results, not Rillet's real strategy.

What is public: Rillet sells "an AI-native ERP platform built specifically for finance teams". IT Brief reported a $100 million Series C at a $1 billion valuation on 20 August 2026, with more than 600 customers (company-reported). Its implementation page says "We do the work. You do the validation." and "Six weeks later, we're sitting alongside your team for your first close", without saying what the six weeks count from. Its About page says "We measure ourselves by how many days we remove from your close." This example uses that measure.

Input, one paste, all illustrative, not a real customer: "Just signed, build the onboarding plan", vendor Rillet, then a 320-person US B2B software company, US parent and UK subsidiary, subscription plus usage billing, leaving a legacy ERP; $54,000 ARR; 12-month contract from Monday 2 November 2026. Sold: general ledger, revenue recognition, consolidation; AP automation after the first close. Bought because the close takes 12 working days, the board pack runs late and usage revenue sits in spreadsheets. CFO buyer, Controller champion, six finance users. AE notes: "first close in Rillet by mid-December", "the UK entity consolidates from the first close", 6 working days by the March 2027 close. Audit fieldwork starts late January. Unknown: UK currency and chart of accounts, the billing integration owner, CFO at kickoff.

What came back: a handoff check, five success criteria, four phases with owners and dates, and a risk register. It named the risk "Unverified promises, unknown UK data" and put one action first: trace all three AE notes to source, "proposal, contract or conversation". How to read it: the 6-day target falls on Day 158, outside the plan, so set a Day 90 checkpoint at kickoff.

The plan says the Day 30 milestone "cannot be a close", proposes balances that tie plus users working in the product, and says "Document this as the first win." Keep the milestone and rename it a readiness gate: by the vendor's own words, the vendor does that work. First value here is the November close, measured against the old one, with the result in the CFO's hands. One usage line recognised in the ERP would count too. The plan's next phase agrees, calling the first close "the first major win".

One decision is not the CSM's. If the fiscal year is the calendar year, a November cutover splits the audited year across two systems. The plan tests that with the auditor by 13 November. Baker Tilly, writing on one ERP product, says "Fiscal year cutover represents the most opportune time to go live with an accounting system." So a January go-live can be sound, even if it breaks the AE's date. The Controller, CFO and auditor settle that.

Step 2. Check your early-churn reasons by hand, and keep a weekly milestone list

Start from your own records: why did new accounts stall or leave early? This example uses three illustrative reasons and an example routine, not Rillet's: kickoff call, CPA-led implementation, weekly status email. For each reason, write what answers it, with an owner and a date. A reason with no owner goes into the risk register.

Then keep one list from Day 1: each milestone, due date, actual date, days late. Review it weekly, because the score waits for Day 90. A slip or a departure opens a review that week.

Write the proof test now. My suggested test: a before and after on one flow, product line or entity, from the buyer's own data, with the period named, who did the work, what ran outside the product, and the owner who signed it off.

The CRAFT GTM retention playbook is optional. It drafts a signal and message per reason, illustrative health weights and a touchpoint calendar, for you to edit. The weights and calendar start from the business model you choose. It sees only what you type, not your product or your customer's calendar, so the call on each reason is yours.

Step 3. After the first live close, turn the notes into a status update

Input, as a CSM would type it on Monday 14 December, Day 43, all illustrative: November closed in the new ERP with the implementation team on working day 9 (Friday 11 December), against 12 before. UK entity mapped, GBP, own chart of accounts, but intercompany eliminations ran in a spreadsheet; the Controller accepts that if December has them. RevOps lead, named at kickoff as integration owner, has the usage feed on 1 of 3 product lines, about 40% of usage revenue. CFO skipped kickoff, came to the Day 14 checkpoint, has not seen the November result. Controller now replies in about two days. Legacy ERP still in parallel; a senior accountant asked to run December there "to be safe for the audit". Auditor wants the opening balance tie-out by 15 January. Third bank feed needs the CFO's signature. AP specialist asks when AP starts. Year-end plan meeting, due by 18 December, not booked.

What came back: progress, twelve dated actions, risks and timeline impact, rated "Tight". It read the gain with care: "one data point at 9, with help". On the fall-back question it said "The request has a real audit reason, so do not dismiss it" and "Let the auditor's answer decide." Leave aside the partial health read it printed; the score waits for Day 90.

How to read it: in these illustrative notes, the fall-back question is reason 1, and a CFO with no result at Day 43 is reason 3. Write the November result to the proof test, with the limits beside the gain: working day 9 against 12, with the implementation team alongside, eliminations in a spreadsheet, the legacy ERP in parallel and about 40% of usage revenue in the ERP, signed off by the Controller. The 12 comes from sales notes, so compare it step by step with the kickoff baseline.

One decision outranks the rest: which system is the system of record for December and the audited year, and when the legacy ERP stops. Put a switch-off date in the plan. It is the hard evidence that fall-back did not happen.

Step 4. Score the cohort at Day 90, then apply the exit test

Input, one paste, all illustrative: a Day 90 check on an onboarding cohort, then one block per account, condensed in the table alongside. All four are on 12-month contracts and pay on time.

What came back: a ranked table, with an illustrative $24,000 at risk, 11.4% of the $210,000 cohort, and no score at Day 70: "Do not read this as healthy." That figure counts only the Critical account. A further $150,000 is Needs Attention, including the half-live $96,000 healthcare company, the larger exposure.

How to read it: the marketplace startup scored 34, Critical. Its fractional controller left at Day 60, no success criteria were written, and December took 11 working days against 10. The skill's first action: call the founder within 48 hours, "Do not defend the product on this call". Its alternative reading: if the slip and the doubt come from losing that controller, help the founder hand over to the next finance owner.

The software company scored 73, Needs Attention: "It needs finishing work, not rescue." December is also year-end, so neither close is a normal month yet.

The healthcare company scored 68, with 9 of 10 on expansion: it asked about a fifth entity while two of four are not live. The skill named the catch: "Without it the account scores 59 of 90, so the 68 should not hide the rollout gap".

The fintech company is unscored at Day 70. But its first close is still next month, weeks behind the software company's Day 40. That looks like reason 1 in plain view, so confirm the date this week.

The 60 line needs a ruling. The onboarding template says a score below 60 at Day 90 means "onboarding is not complete, extend with intervention plan", and passing 60 proves little. My suggested exit test, every part required:

1. Score: 60 or above. Within 10 points of 60, read the signals by hand.
2. Core job: the job the customer bought it for runs start to finish in the product, and the old system no longer runs alongside it.
3. The rest: everything else sold is live, phased in the contract, or on a new date the buyer accepted in writing.
4. Proof and path: one line meets the proof test, and the path to the sold target is written, with an owner.

A date alone never clears part 2. Healthcare fails it even with a dated plan: two entities are dark and consolidation sits in a spreadsheet. At $96,000 it gets the team's time first. The software company is close: ask whether the legacy ERP is off, get the buyer's date for product line 3, and write the path from 10 working days to 6. The marketplace startup gets an extended plan sized to $24,000. Tell sales: the doubt may be a sale to the wrong size.

In this illustrative cohort, none of the four accounts meets every part of the exit test. For the software account signed on 2 November, Day 90 is 30 January, before January has closed. Treat Day 90 as a review, and set each exit on the customer's calendar. Exit should change something: the ongoing plan takes the target and whatever remains.

This is the point where I would want a second pair of eyes: the Day 90 score and independent first-value evidence disagree. Bring me your sales handoff, milestone record and exit test, and we will work it through in 45 minutes.

1. Run the Onboarding plan skill on every account signed in the last two weeks, and settle each flagged promise with the AE before kickoff.
2. Put the customer's close, audit and board dates into the plan.
3. Rename your Day 30 milestone a readiness gate, and write down first value.
4. Go through each early-churn reason by hand, and start the weekly milestone list.
5. Book each account's Day 90 review, and write the exit test into your plan.

1. Ending the plan at go-live, or calling migrated data that ties out first value.
2. Taking the AE's promises into kickoff unchecked.
3. Leaving success criteria to "later", so the Day 90 review has nothing to show.
4. Scheduling the first live close on year-end or an audit.
5. Reading a Day 90 score above 60 as a pass, or letting a date stand in for a result.

• When High Adoption Hides Churn Risk: A GUARD Health Check, Save Plan, Expansion Case and QBR on One Book of Accounts
• How I Check Product-Market Fit, Churn and ICP Gaps Before Scaling Paid Demand Generation
• B2B Crisis Communication That Protects Renewals: Seven Real Incidents and a Worked Save Plan
• All 30 guides