Build the Listening Program: Query Set, Routing Matrix, Review Cadence

Marketing Academy · Field Work●Social Media Marketing

CoreBuild the Asset· 50 minutes

PolicyBazaar (PB Fintech)

Objective: Build the three foundational documents a real social listening program needs before any tool is purchased: a Boolean query set by use case, a routing matrix with named owners, and a review cadence.

You're setting up PolicyBazaar's first structured social listening program. Leadership approved budget for a mid-market tool, but nobody has defined what to track or who acts on what, and a tool without that groundwork is just a cost center.

Design query sets by use case, a routing matrix with named owners and SLAs, and a review cadence, before recommending which tool to buy.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets

FreeDraft the query sets, routing matrix, and review cadence before any tool purchase

Free, sharable with stakeholders for sign-off before committing budget to a paid listening tool

Paid upgrades (optional, faster/deeper)

Build all three documents free first. Buy the tool only after the routing matrix has named, accepted owners, a tool cannot fix an undefined routing plan.

Hootsuite(optional)

PaidRun the finalized query sets at scale with built-in routing and alerting once the matrix is signed off

Purpose-built for exactly the query-set and alert-routing structure this project designs manually

The process

3 steps

Step 01 of 03

Boolean query design by use case

The lesson's Layer 1 says build separate Boolean query sets per use case, brand health, product feedback, category, and crisis triggers, because combining them into one query creates noise.

PolicyBazaar sells insurance and loan comparison. What does a crisis-trigger query set look like versus a category query set for the same brand?

Google Sheets— One tab per use case, listing include terms, exclude terms, and example matched posts.

Procedure

Brand health tab: 'PolicyBazaar', common misspellings, 'PolicyBazaar vs [competitor]'
Category tab: 'best term insurance India', 'cheapest car insurance comparison', 'claim rejected' without the brand name
Crisis tab: 'PolicyBazaar' AND ('claim denied' OR 'fraud' OR 'lawsuit' OR 'data leak')
Exclusions tab across all sets: official PolicyBazaar handles, employee accounts, press-release syndication domains

Sample output

CRISIS QUERY SET
  Include: 'PolicyBazaar' AND ('claim denied' OR 'fraud' OR 'data leak' OR 'lawsuit')
  Exclude: site:policybazaar.com, @PolicyBazaarSupport, @PolicyBazaarIndia
  Sample match: 'PolicyBazaar denied my claim after 6 months, this is fraud' (Twitter/X, 340 likes)

Healthy

Four separate query tabs exist, each tuned to catch a different signal type without drowning it in the others.

Unhealthy

One combined query for 'PolicyBazaar' returns 4,000 mentions a week, 90% of them praise and support replies, burying the 2 crisis-shaped posts inside the noise.

What this means

A single combined query optimizes for volume, not for signal; separate use-case queries optimize for the decision each one is supposed to trigger.

So what do I do about it?

Symptom	Action	Effort
The listening dashboard shows thousands of mentions a week and nobody can find the ones that matter	Split the single combined query into 4 use-case-specific query sets with their own exclusion lists	30 min

YouYou can do this yourself, no engineering access required.

Step 02 of 03

Routing matrix with named owners

The lesson's Layer 4 requires every categorized mention type to have a predetermined owner and channel before the tool is even configured.

You have 4 query sets from Step 1. Who receives output from each one, and on what SLA?

Google Sheets— A single matrix tab: Signal Type | Owner | Channel | SLA.

Procedure

Product feedback (from category queries) -> PM team, tagged into Jira, weekly triage
Sentiment trend shift (from brand health queries) -> Brand team Slack channel, weekly threshold alert
Volume anomaly, 3x baseline in 60 minutes -> Comms on-call, paged immediately
Competitor mentions (from brand health comparisons) -> Sales intelligence, weekly digest

Sample output

ROUTING MATRIX
  Product feedback -> PM team -> Jira board -> Weekly
  Sentiment shift -> Brand team -> Slack #brand-health -> Weekly threshold alert
  Volume anomaly (3x/60min) -> Comms on-call -> Phone page -> Immediate
  Competitor mention -> Sales intelligence -> Weekly digest email -> Weekly

Healthy

Every signal type has exactly one owner and a stated SLA before the tool is purchased.

Unhealthy

The tool goes live, mentions accumulate in a dashboard, and 6 weeks later leadership asks why nobody acted on the sentiment dip flagged in week 2.

What this means

The routing matrix, not the tool's feature list, is what determines whether the program produces decisions or just reports.

So what do I do about it?

Symptom	Action	Effort
A listening tool is live but nobody can say who owns a given alert type	Build the routing matrix before finalizing the tool purchase, and confirm each named owner has accepted the SLA	30 min

YouYou can do this yourself, no engineering access required.

Step 03 of 03

Review cadence with a required decision

The lesson's Layer 5 says every review should produce a decision, a feature change, a message adjustment, a pivot, not just a slide deck summarizing what people said.

You have daily, weekly, and monthly review touchpoints defined. What decision does each one have to produce to count as complete?

Google Sheets— A cadence tab: Frequency | Attendees | Required Output.

Procedure

Daily (automated): anomaly alerts only, no meeting, comms on-call reviews and pages if needed
Weekly: PM + brand team review product feedback and sentiment trends, must produce one prioritization decision or message adjustment
Monthly: leadership review of brand sentiment and share-of-voice, must produce one resourcing or strategy decision

Sample output

REVIEW CADENCE
  Daily: Comms on-call, anomaly alerts, decision = page or dismiss
  Weekly: PM + Brand, product feedback synthesis, decision = 1 backlog item added or message adjusted
  Monthly: Leadership, sentiment + share-of-voice, decision = 1 resourcing or strategy call

Healthy

Every weekly review ends with one named decision written into the matrix, even if the decision is 'no action needed, monitor.'

Unhealthy

The weekly review is a 30-minute readout of what people said last week, with no decision attached, that repeats every week with nothing changing.

What this means

A review without a required decision output degrades into a status meeting; naming the required output in advance forces the discipline.

So what do I do about it?

Symptom	Action	Effort
Weekly listening reviews run 30 minutes and produce a summary slide, not a decision	Add a 'Required Output' column to the cadence tab and refuse to close a review without filling it in	5 min

YouYou can do this yourself, no engineering access required.

Final deliverable

Three linked documents: a 4-set Boolean query design, a routing matrix with named owners and SLAs, and a review cadence with required decision outputs.

See a reference example

Sample output

Robinhood listening program setup (excerpt)

QUERY SETS: Brand health, Category ('best trading app for beginners'), Crisis ('Robinhood' + 'lawsuit'/'outage'/'frozen account'), Exclusions

ROUTING MATRIX
  Volume anomaly -> Comms on-call -> Immediate page
  Product feedback -> PM -> Jira -> Weekly

CADENCE: Weekly PM review must close with 1 backlog decision; monthly leadership review must close with 1 resourcing call

Success criteria

You're done when you can:

4 distinct query sets exist with include, exclude, and example-match rows
Every signal type in the routing matrix has exactly one named owner and a stated SLA
Each cadence tier states a required decision output, not just an attendee list