Our social budget was almost entirely production and media. Responding to people was something the person who posted did when they had a moment, which in practice meant sporadically and slowly.
We changed that, mostly because complaints were going unanswered, and the effects went further than we expected.
What we found in the backlog
The audit was the thing that made the case.
Several hundred unanswered questions, many of them straightforward pre-purchase queries from people evidently considering buying.
A number of complaints that had escalated in public because nobody responded privately.
Several instances of other customers answering questions on our behalf, mostly correctly and occasionally not.
And a consistent pattern of people asking the same handful of things, which told us something about our own website that we had not learned from any analytics.
The response time finding
The clearest measurable effect once we started tracking it.
Questions answered within an hour converted at a noticeably higher rate than the same questions answered the next day. Not marginally — the difference was large enough to be obvious without any statistical work.
The likely mechanism is unremarkable. Somebody asking a pre-purchase question is in a buying frame at that moment. A day later they have decided something else or forgotten.
This is well established in customer service research generally and we had somehow not connected it to social.
What the role actually involves
More than replying, which is what we had assumed.
Triage, deciding what needs a response, what needs escalation, what is better handled privately and what should be left alone.
Judgement about tone, which is genuinely difficult and is the part that cannot be scripted.
Knowing when not to respond, particularly to bad faith engagement, which is a skill and which inexperienced people get wrong in both directions.
And feeding back internally. The person reading everything customers say is the best-informed person in the building about what is confusing, broken or missing, and that information rarely reaches anybody who can act on it.
The internal reporting was the biggest gain
The unexpected result.
We introduced a weekly summary of recurring themes, sent to product and to the website team.
Within three months that had produced several concrete changes — a rewritten delivery information page, a fix to a checkout step that generated repeated confusion, and a product description change.
Each of those addressed a problem that was generating support volume and losing sales, and none had surfaced through any other channel because nobody was aggregating the complaints.
What we got wrong first
Two mistakes worth describing.
We tried to handle it with a rota across the marketing team. It failed because nobody owned it, response times were inconsistent, and the tone varied depending on who was on duty.
We then tried scripted responses to improve consistency, which produced the recognisable corporate non-answer that makes people angrier than no response at all. Anybody who has received one knows immediately what it is.
What worked was one person owning it, with genuine authority to resolve things, and guidelines about principles rather than scripts about wording.
The authority point specifically
The single most important structural decision.
A community manager who has to escalate everything is a routing function, and the delay destroys the value.
Giving the role authority to resolve issues up to a threshold — refunds, replacements, corrections — without approval transformed both the speed and the tone of the interactions.
The cost of that authority being occasionally misused was substantially lower than the cost of the delays it removed, which we were able to demonstrate after a quarter.
What I would tell somebody making the case
Do the backlog audit first. Count the unanswered pre-purchase questions and estimate what they were worth. That number is usually enough on its own.
Track response time against conversion, which is straightforward and produces a clear result.
And frame the internal reporting as a research function rather than as customer service, because it genuinely is, and because it reaches budget holders who do not care about social media.
Out of hours
The practical question that always comes up and deserves an honest answer.
Complaints and questions arrive at all hours, and response time matters most at exactly the times nobody is working.
What we settled on was published response hours, stated clearly on the profile, plus an automated acknowledgement outside them that tells people when they will hear back.
That is considerably better than silence and considerably cheaper than staffing overnight. The thing people object to is being ignored rather than being asked to wait, and setting an expectation addresses most of it.
Measuring it
The thing that keeps the budget in place after the first year.
We track median first response time, the proportion of pre-purchase questions answered within an hour, and the conversion rate of people who asked a question against those who did not.
That last comparison is not a clean experiment and it is directionally clear, and it is what the finance conversation actually turns on.
We also count the internal changes prompted by the weekly summary, which is a crude measure and demonstrates the research value that the service metrics miss entirely.