Customer feedback automation
The short answer
Customer feedback automation means software does the repeat work: collecting from every source, grouping matching requests, ranking them and drafting the thank-you note. A person still decides what to build and sends the note. Annsa automates those four steps and leaves the decision and the send to you.
Customer feedback automation is software doing the repeat steps of feedback work, so a person only makes the calls.
TL;DR
- Automate the steps that repeat: collecting, grouping, ranking and drafting.
- Keep the calls with a person: what to build, what to say no to, and sending.
- Check the output weekly. A ranked list you never look at is a guess.
When to use it: when feedback arrives from several places and sorting it by hand is the job nobody gets to. When not: with a few customers a week, read it yourself. You will learn more.
What should I automate?
Four steps, and each one is dull enough to hand off.
- Collecting. Feedback should come in from where it is said: Slack, support chat, call transcripts, surveys, GitHub, a spreadsheet. If a person has to paste it, it will stop happening.
- Grouping. Ten people ask for the same thing in ten ways. Software can put them in one row, with all ten quotes underneath.
- Ranking. A fixed order you can read: what is broken, how many accounts, how much revenue, how urgent. The rule is written down, so you can see why row 1 is row 1.
- Drafting the note back. When a job ships, a draft to each person who asked, using their own words.
What should stay a person's call?
Anything that commits you.
- Deciding. What to build, and what to leave. A ranking tells you what customers asked for. It does not know your bets.
- Saying no. A no needs a reason in your voice.
- Severity. Whether something is broken is a judgment. A person's mark should win over any score.
- Sending. A reply goes out in your name. Read it first.
There is a name for the risk. Automation bias is the tendency to trust what an automated tool suggests and to dismiss information from other sources, even when that information is right. A ranked list is an input, not an answer. Look at the top five by hand each week, and read two raw quotes.
What does a good setup look like?
- Every source connected, so nothing waits for a person to paste it.
- The quote and the account kept on every request, so you can check the grouping.
- A visible rule for ranking, so you can disagree with it.
- A held state, not a drop. If you hit a limit, new feedback should wait, not vanish.
- A person sends. Drafts are free; sends are yours.
Use the list below to sort your own steps.
## Feedback work: automate or keep
Automate
- Pull feedback in from every source (no pasting)
- Group requests that mean the same thing
- Rank by a written rule (broken first, then accounts, revenue, urgency)
- Draft the note back to each person who asked
Keep as a person's call
- Decide what we build and what we leave
- Say no, with a reason in our voice
- Mark what is broken
- Read and send every note
Check weekly
- Read the top 5 by hand
- Read 2 raw quotes from the top row
- Look at what was held or dropped
How does Annsa automate feedback?
Annsa is the source of direction for your agents. It automates the steps above and leaves the calls to you.
- Collects. Feedback comes in from connected sources: Slack, Intercom, call transcripts, surveys (in-product, six templates), CSV and Google Sheets imports, GitHub and an API. Each request keeps the customer behind it, their exact words, the account and its revenue.
- Groups and ranks. Annsa groups matching requests. Within your bets, what a person marked broken comes first, then what customers asked for, then ideas. Then severity, how many accounts, revenue and urgency. A person's severity always wins.
- Drafts. When a job ships, Share Back drafts a note to each person who asked.
- You decide and send. You choose what to build. You read and send the notes. Annsa never decides for you.
See Feedback for the connected sources.
Source: Automation bias, Wikipedia.