Flagship research program · Independent work
Frontline Signal
What the store knows that the system doesn't.
Sixty friction moments logged on shift inside a large-format home improvement store, coded into nine operational themes, traced to structural causes, and sequenced into a pilot a store leader could actually run.
Conducted independently at a national home improvement retailer. All identifying details — brand, store and people — have been removed.
60
observations logged on shift
3 weeks
single store, PM-weighted
45
recurring — every or most shifts
9
themes after clustering
Executive summary
Frontline associates observe recurring friction that is invisible in aggregate data. When digital guidance and physical-store reality disagree, technology increases customer frustration rather than removing it.
The finding
Nine themes explain the 60 logged moments; five carry P1 priority and every one of them is structural rather than behavioral.
The proposal
A structured system for capturing, validating, prioritizing and routing frontline observations to the function that already owns the fix.
The ask
A cross-functional store walk and one small, measurable pilot — instrument the signal before changing anything expensive.
Framework
Capture, validate, prioritize, route, learn.
The discipline is in the consistency: identical capture structure across shifts is what lets separate observations be compared rather than merely accumulated.
- 01
Capture
One customer story or store-improvement note, tagged to a friction theme, in under two minutes.
- 02
Validate
Compare against repeat stories, operational data and work already funded. Anything already owned drops out.
- 03
Prioritize
Score frequency, severity, safety, sales impact, scalability and effort.
- 04
Route
Send each theme to the function that already owns the fix, with a named owner.
- 05
Test and learn
Run a narrow intervention, measure against baseline, then refine, scale or stop.
Methodology
What this sample is — and what it isn't.
The sample is small on purpose. One associate logging one store across three weeks surfaced 60 friction moments, which is the argument for capturing this at scale, not the argument against the data.
What it is
- 60 observations recorded on shift by one associate in one large-format store.
- A three-week window, weighted to evening shifts, written down in the moment.
- Grouped by root cause rather than by department.
- Directional evidence: enough to say what to look at, and in what order.
What it is not
- Not a statistically representative sample of a chain or a region.
- Not a customer survey or a replacement for existing operational data.
- Not a claim that every store carries these problems at this rate.
- Not a finished business case — that is what the pilot is for.
Field evidence
Verbatim first, interpretation later.
Three of the sixty entries. Each was logged in the moment with a customer objective, a business impact and a proposed smallest fix.
“They were ready to buy and had to wait to be allowed to.”
“Customers feel stuck needing to be the Hulk because no one is around.”
“She knew what she needed the second she walked in. She just had no idea where to go.”
Associate stories
The shift, as experienced.
Hover or select a beat to move through the arc of a single shift and the opportunity hiding inside each moment.
What they're thinking
“I need to finish my tasks and still support customers.”
Friction
Task productivity and customer service compete for the same minutes.
Opportunity
Make customer support easier to accept, route and close.
Affinity mapping
Sixty entries, grouped by root cause rather than by department.
An entry can carry more than one theme, so cluster counts intentionally sum to more than sixty.
Findability & adjacency
Heavy items & carts
Locked merchandise
Inventory truth
Journey mapping
Themes placed back onto the customer timeline.
Synthesis only becomes useful when it is re-attached to a moment in time that someone can own.
Moment
Identifies a home repair, maintenance or installation problem.
Motivated, slightly uncertain
Pain points
- Knows the goal, not the product language.
Opportunities
- Let people shop by project instead of requiring product expertise.
Moment of delightSomeone asks what they're building before asking what they need.
Root cause analysis
Trace it back until you reach something structural.
Observed symptom
A ready-to-buy customer waits eight minutes for a locked case to be opened — and sometimes leaves without the item.
The unlock request reached an associate verbally and then an overhead page, and neither was owned by anyone.
There is no response clock or acknowledgement on unlock requests, so nothing is late.
The people with keys are also the people mid-task, and the two demands are never reconciled.
Nobody records the request that failed, so the lost sale never appears anywhere it could be counted.
Root cause
There is no cheap capture mechanism for frontline observations, so the organization only ever sees the outcome of the problem — a lost sale — and never its cause.
Operational themes
Nine themes, prioritized by frequency and impact.
Findability and wrong adjacency
Customers arrive knowing the exact item and still can't locate it: tape split across three sections, dollies stored outdoors far from moving supplies, torches shelved with automotive. Fix: project-based adjacency tests plus location-aware help.
Heavy items, carts and lifting
Thirty bags of concrete loaded alone in the lot; drywall lifted with a child; large lumber in a grocery cart. Seniors reported buying less rather than walking back for a flat. Fix: a located “need help lifting” request with staged flats and a confirmed responder.
Locked merchandise
An eight-minute wait for a drill unlock and an escorted checkout. A $99 drill request went unanswered through both an associate and an overhead page. Fix: response-clock routing with named ownership.
Inventory truth
Items show in stock while still on the truck, or are stocked too high to reach. Fix: a one-tap “system says here, floor says no” flag feeding an inventory-confidence score — the same correction layer any AI assistant would need.
“No home” items
Merchandise with no assigned location gets stashed, ages into shrink and burns pick time. Nobody owns the fix because nobody is authorized to give the item a home. Fix: authorized, audit-logged local location records.
Wrong product for the job
Catches that only happened because an experienced associate was standing there — brake-contaminating lubricant, outdoor insecticide for an indoor rug, indoor caulk on a leaking roof. Fix: a “what are you using this for?” check at the shelf.
Language and accessibility
A single request pulled a second bilingual associate off task during a rush; another was solved with a translation app and luck. Fix: bilingual describe-and-visual search so a basic request doesn't consume two associates.
Coverage and response time
Service in some departments swings entirely on whether the veteran associate is scheduled; overhead announcements routinely produce no one. Fix: cross-training plus dispatch with acknowledgement.
Signage and small-hardware bins
Fifteen minutes digging through a mixed fastener bin; double-sided section numbering makes everyone overshoot by one. Fix: price-sorted labeled bins, scan-to-match, single progressive numbering.
Pilot
Instrument first. Change expensive things last.
Phase 01
Store walk
Walk the documented friction points with cross-functional partners and validate the current condition. Map each theme against work already funded and drop what's solved.
Phase 02
Signal capture
Test one customer story and one proactive-improvement entry per participating associate per shift. Measure capture burden from day one.
Phase 03
Prioritization
Identify the highest-frequency, highest-impact patterns and assign a named owner per theme.
Phase 04
Intervention
Two or three small changes: one adjacency test, one locked-item response process, one entrance help-and-app discovery point.
Phase 05
Measurement
Compare before and after against the walk-day baseline.
Phase 06
Recommendation
Decide what scales, what gets refined, and what stops.
Business value
Three measures that would decide it.
Each is observable inside a single quarter and comparable against control stores.
Time to find a product
Median assisted-find time in pilot departments, timed during the store walk as a baseline.
Leading · weekly
Locked-item completion
Share of unlock requests acknowledged with a named owner. Currently unmeasured.
Leading · per request
Location corrections
Volume and accuracy of associate-flagged mismatches. Baseline is zero — no channel exists today.
Lagging · quarterly
Objections
The arguments worth naming before someone else does.
Associates are measured on productivity. Does this add work?
Capture is under two minutes, voice-first, and framed as one story per shift — not a quota. If burden runs high in the pilot, the honest answer is to cut the form down or stop.
What happens when nobody owns the incoming signals?
The pilot routes to named owners per theme and measures follow-up completion. A signal system with no closing loop becomes a complaint box — that's the primary failure mode to test for.
How do you keep this from becoming a complaint channel?
Every entry requires a customer objective, a business impact and a proposed smallest fix. A second track counts problems associates prevented, not only problems they found.
Why should the frontline set the roadmap?
It shouldn't. The frontline supplies observations; the prioritization model, the operational owners and the pilot data decide what's worth doing.
Quick Help is the customer-facing concept this program produced.
Read Quick Help