rulebase-co Support-led growth
Someone has noticed support talks to more customers than any other team and asked what it could be doing commercially. The question is reasonable and the usual answers are bad.
The framing that works: support drives growth by removing the reasons customers do not grow, not by promoting. Those are different mechanisms with different costs, and only one of them survives contact with a support queue.
The mechanisms, best first
1. Removing adoption blockers. A customer who cannot get a capability working will not expand into it. Fixing that is unambiguously support's job, it requires no change of posture, and it is the largest of these effects in most businesses. This is where to start, and it is mostly an onboarding-friction and knowledge problem rather than a new programme.
2. Resolving the failure that is blocking a decision. Accounts frequently pause an expansion until an open problem is fixed. Identifying those and prioritising them is high-value and needs no selling — just knowing which open issues sit on accounts with a pending decision.
3. Closing the loop on requests. Telling a customer the thing they asked for now exists is the highest-converting, lowest-risk growth motion available to support, and almost nobody does it systematically. It requires only a record of who asked for what. Start here if you want a quick, safe win.
4. Contextual adoption nudges. Pointing out an existing, already-included capability that solves the problem at hand. Safe when it is relevant to the conversation and already paid for. Becomes promotion the moment it involves an upgrade.
5. Advocacy and referral. Customers who have had a genuinely good recovery experience are unusually willing to refer or give a reference. Timing and consent decide whether this is graceful or grubby.
6. Flagging commercial signal to the account owner. Detection only, and it is a handoff design problem rather than a growth programme.
Everything past 4 needs an explicit design; everything past 6 is sales.
The line, stated plainly
Support may help the customer do more of what they already bought. Anything that requires the customer to spend more is a commercial conversation and belongs to whoever owns the account.
Concretely, do not:
- Mention an upgrade, a paid tier, or a price in an unresolved conversation.
- Give agents revenue targets, quotas, or commission. This is the single change most likely to damage a support operation, and its effects show up in quality and repeat contact within a quarter.
- Ask for a referral or review from a customer whose issue is open, or whose issue was a failure you caused, until it is genuinely resolved and they say they are happy.
- Route a complaint, a vulnerable customer, or an account in financial difficulty into any growth motion.
- Add a promotional footer to support replies. It reads as advertising in a service interaction and it measurably reduces trust in the reply above it.
Design the safe motions properly
Request loop-closing. Needs a record of who asked for what, a trigger when it ships, and a message that is a genuine notification rather than a pitch. Send it from support, reference their original conversation, and do not attach an upsell. Measure adoption of the shipped capability among those told versus comparable customers not told.
Adoption nudges. Rules, not agent discretion: relevant to the resolved issue, included in their current plan, at most one, after resolution. Anything else and it becomes a habit that spreads to the wrong conversations.
Advocacy asks. Only after a resolved interaction the customer has confirmed they are happy with, only once in a defined period, and never for a case where you caused the problem. Make the ask separable from the support conversation so declining costs the customer nothing.
Measure it as an experiment, with a guardrail
Every one of these is testable and almost none are tested.
- Withhold at random. Tell half the customers who asked for a feature that it shipped; compare adoption and expansion. This is cheap, it is the only way to know the effect is real, and it converts an anecdote into a number that survives review.
- Report the guardrail every time: satisfaction, complaint rate, and repeat contact for customers who received the motion versus comparable customers who did not. If the guardrail moves the wrong way, the programme is net negative regardless of the revenue line — and support-side damage is slow, so you will only see it if you look.
- Watch agent-side effects — handle time, quality scores, and whether participation correlates with anything worse.
- Attribute honestly. Without a holdout, report influence rather than causation.
Traps
- The team gets a revenue target. Almost always the beginning of the end: quality drops, then repeat contact rises, then the revenue gain reverses. If a target is imposed, argue for a support-quality guardrail with a stopping condition attached, and get the stopping condition agreed in writing before launch.
- Measuring activity instead of outcome. Nudges delivered, features mentioned — these rise on command and mean nothing.
- Ignoring the trust cost. It is real, it is slow, and it does not appear in a monthly pipeline review. The guardrail metric is the only instrument for it.
- Assuming the requester is the buyer. The person who asked for the feature may have no authority and no interest in the commercial consequence.
- Consent. Outbound notification and advocacy asks are governed by marketing-consent rules regardless of how helpful the message is. Flag it; do not assume the support relationship covers it.
Present results to the user
- The mechanisms in scope, ordered, with the line between helping customers use what they bought and asking them to buy more stated explicitly.
- What is excluded — complaints, vulnerability, unresolved conversations, financial difficulty, consent gaps.
- The design for each motion in scope, as rules rather than agent discretion.
- The experiment — what is withheld from whom, and what will be compared.
- The guardrail metric and its stopping condition, agreed before launch.
- Results as influence unless a holdout ran, and the holdout design if a causal claim is wanted.
- Agent-side effects, so a revenue gain paid for with support quality is visible rather than discovered later.