Service and success are two doors into the same machine

I reckon the proactive-help part of customer success can use the same machinery as inbound service. The first thing I would count is what the two teams maintain twice.

They may keep separate answers and separate views of the same account. If they do, someone is paying to reconcile them, or the customer is finding the differences.

Start with a narrow slice of customer success: initiating help for someone who appears to need it. A prompt to set up a useful feature is an example.

Once the customer accepts, the work may resemble an inbound request for the same help. Both routes need the relevant answer and permission to perform any required action.

Who starts the conversation still matters. It changes whether the contact is welcome and what a useful result looks like.

So my claim is about sharing the work behind that exchange. It does not cover the whole customer success job.

In the five-cause piece, I propose labels for failed attempts to get help. They cover missing answers, retrieval, actions, avoidable questions and poor handling.

I would test whether those labels work for outbound help too. A missing prompt is one additional failure to record: the customer never received the offer.

Once the exchange starts, the answer may still be missing or hard to retrieve. An action may be unavailable, or the assistance may be handled badly.

The avoidable-question label may also apply. Perhaps the customer should not need help setting the feature up because the default could meet their needs.

I would not assume the causes match until reviewers have tried them on both kinds of interaction. The useful result is a shared fix, not a tidy diagram.

Count these first:

1. Requests that both teams answer. Check whether they use the same approved information or give different answers.
2. Content maintained separately for the same purpose. Distinguish duplicate work from material that serves a different audience.
3. Setup questions reaching support from accounts receiving proactive help. Check whether the outreach could reasonably have addressed that question.

That last count needs restraint. A managed account contacting support does not prove anyone failed; support may be the intended route.

It gives us a place to inspect the handoff. Did the teams know what the other had already offered or tried?

Those checks can establish a duplication cost without settling an argument about organisation design. They can also show that the apparent duplication has a useful purpose.

The measurement of outreach needs its own test. A comparison between contacted and uncontacted accounts is hard to interpret if someone chose the accounts based on likely need.

I cannot tell from outside a team how well it handles that problem. I would ask for its evaluation design before claiming that its outreach works or does not.

Where practical, I would randomly withhold optional outreach from part of an eligible group. Keep ordinary support available and compare a result chosen before the trial starts.

This is a proposal to test an extra prompt. It is not permission to withhold assistance that a customer needs or has been promised.

The team would need to choose a suitable group, trial duration and measure. It would also need enough observations to distinguish a useful effect from noise.

A holdout can help with that comparison, but it cannot settle every question. Accounts may share information, other teams may contact them and the result may take time to appear.

Record those limits before using the result in a budget decision. A trial that cannot detect the effect you care about should not be reported as proof of no value.

Even if shared help works, merging the teams does not follow. Renewals, commercial ownership and advice about a customer's business remain outside the narrow slice I started with.

My call is to test shared content and actions first. Keep access controls appropriate to the task and preserve the context of how each conversation began.

The content base piece describes the maintenance I would expect behind those shared answers. The service experience piece asks what evidence shows that the help actually worked.

Keep the reporting separate. Useful outreach may generate conversations while useful inbound help may prevent repeat contacts.

A single contact-volume target could reward one team for undoing the other's work. Decide what each measure means before combining the reports.

Remember the difference in consent too. Inbound service responds to someone asking; outbound help interrupts someone who may not want it.

Give people a clear way to decline and use that response in future contact decisions. A low complaint count is weak evidence that the prompts are welcome.

Where I might be wrong is in how much of the customer success job fits this proposal. Commercial and relational work may account for most of its value.

If so, sharing answers could save some maintenance without changing much else. I would size that saving before asking either team to reorganise around it.

A holdout may also be impractical for the accounts that matter to the decision. Contractual promises or the need to provide help can rule it out.

In that case, narrow the trial to optional outreach where it is appropriate, or use another evaluation method and state its limits. If the result does not apply to the wider accounts, do not claim it does.

The question to ask

Ask whoever owns proactive help what evidence would make them change the programme. Then choose a test that could produce that evidence.

What decision will you make differently when the result arrives?