How Smart Lock Factories Control Product Quality explains the gates inside a plant — incoming component inspection, in-process control, end-of-line testing.

Diperbarui pada: Mei 14, 2026

How to Check Smart Lock Supplier Product Quality

How to Check Smart Lock Supplier Product Quality

Checking a smart lock supplier's product quality is not an event you perform once during vendor selection. It is a surveillance program you run continuously on the goods you have already bought: incoming inspection on every shipment that lands at your dock, structured logging of every failure that surfaces in the field, and a batch-consistency comparison of each new lot against a sealed golden sample from the first accepted shipment. The factory's own quality control ends at the shipping mark. Your obligation starts there, because by the time a defective lock reaches a hotel guest or a rental tenant, it is your name on the invoice, not the factory's.

This manual is the buyer-side counterpart to the factory-side documentation on this blog. How Smart Lock Factories Control Product Quality explains the gates inside a plant — incoming component inspection, in-process control, end-of-line testing. Smart Lock Factory Quality Control: What Buyers Should Check tells you what to demand evidence for before production finishes. Smart Lock Factory Testing: What Should Be Tested? covers which tests a competent factory runs. None of those three articles help you on the morning a container arrives with 800 locks inside and three weeks of ocean transit between you and the factory floor. This article covers that morning, and every morning after it.

The short answer, up front:

How do you check a smart lock supplier's product quality? You run three data streams on delivered goods — per-shipment incoming inspection against a written acceptance standard, field failure tracking from installed units, and batch-consistency review against a sealed golden sample — then feed all three into a weighted supplier scorecard that decides reorder, corrective action, or disqualification. Quality you do not measure after delivery is quality you are borrowing against.

Where the Factory's QC Ends and Your Surveillance Begins

A supplier can run every gate described in the factory-side articles and still ship you a bad lot. Seals loosen in transit. A factory quietly changes a plastic supplier between orders. A firmware revision meant for a different customer lands in your SKU. Battery chemistry that passed validation ages six months in a port warehouse before it reaches your shelf. Factory QC protects the process; it cannot protect you from every outcome, and it certainly cannot protect you from drift you never measure.

The distinction that matters is timing and ownership. Factory quality control answers "did we catch defects before shipping?" Your surveillance answers "what actually arrived, what is happening after six months of use, and is lot 7 the product I agreed to in lot 1?" Those are questions only the buyer can ask, from the buyer's side of the shipping mark, with the buyer's own records.

Run the program as three streams, each catching what the others cannot see:

Surveillance streamQuestion it answersDefect classes it catchesRuns when
Incoming inspection (per shipment)"Does what landed match what we accepted last time and the spec sheet?"Transit damage, missing parts, wrong finish, dead-on-arrival units, packaging defects, gross assembly faultsWithin 48 hours of arrival
Field failure tracking (continuous)"What breaks after installation, and how often?"Battery drain, reader drift, motor/latch failures, firmware faults, finish corrosion, false unlocksEvery service ticket, logged weekly
Batch consistency review (per new lot)"Is this lot the same product as the lot that built our reputation?"Silent component swaps, firmware revisions, tolerance creep, material changes, spec-sheet driftAt receipt, before the lot is sellable

A distributor with all three streams reports failures to a supplier with numbers, dated logs, and photographs. A distributor without them says "some customers are complaining" and gets a shrug. The difference in negotiating position is enormous, and it costs only procedure.

Roles, Ownership and the Standing Rule

An SOP without named owners is a suggestion. Assign these roles before your next shipment, even if two of them are the same person wearing a different hat.

RoleOwnsNon-negotiable duties
Receiving InspectorStream 1Opens cartons, pulls samples, runs the receiving checklist, signs the lot record accept/reject the same day
Quality CoordinatorAll streams + recordsMaintains the golden sample shelf, the failure log, the batch comparison sheet; the single person who can say "this is the third reader failure this quarter"
Field Service / Support LeadStream 2Captures every ticket into the log within 24 hours — model, lot code, failure mode, resolution; never lets a fix happen off-record
Product Manager / OwnerDecisionsApproves dispositions (sort, return, rework), signs corrective-action requests, runs the quarterly scorecard and the reorder decision
Supplier Quality Contact (external)Their sideNamed person at the supplier who receives your lot results and owns the response clock

Standing rule: no conforming or nonconforming status on any lot until the inspector signs the record. Unsigned inventory sits in a holding zone and does not ship to customers. This rule exists because the first thing a warehouse under deadline pressure will do is "borrow" quality decisions from optimism.

Small teams (a two-person distribution shop) collapse these into one person plus a spreadsheet, but keep the sequence: inspect before sellable, log every field failure, compare every new lot. The roles scale down; the discipline does not.

The Golden Sample: Your Reference Object Before Anything Else

Batch consistency requires a "before." Without it, every new shipment is judged against memory, and memory is the weakest quality instrument in any warehouse.

Procedure G-1 — Establishing golden samples

  1. At your first fully accepted order for each SKU, pull 3–5 units before they enter sellable stock. One unit stays sealed in original packaging; two are opened for destructive comparison; one is a live demonstration unit (rotate it out yearly, as its battery and finish age).
  2. Photograph each unit: faceplate, edge, latch bolt extended, battery compartment, PCB side if opened. Save images to the SKU's record with the order number and lot code.
  3. Record the configuration string: model, firmware/app version if checkable, finish name, door-thickness range as printed on the box, credential types the unit accepts (fingerprint, RFID card at 13.56 MHz, passcode, mechanical key, app), and what the box contents list (screw pack, templates, cards, manuals).
  4. Store sealed units in the same conditions your customers' stock sees — not in a climate-controlled office. A golden sample that outlives its finish in a drawer is a false reference.
  5. Golden samples are re-sealed or refreshed only with written sign-off from the Product Manager. When you intentionally change specification (new handle, new colorway), both the old and new references stay on the shelf, labeled with effective order numbers.

If a supplier pushes back when you cite batch drift — "this is the same product" — the sealed unit from order one settles the argument in seconds. Bluff survives memory. It does not survive a taped box with a dated label opened on a conference table.

Receiving Inspection: The First 48 Hours After a Shipment Lands

This is the heaviest procedure in the manual, and the one with the highest return per hour invested. It applies to every inbound lot of every SKU.

Procedure R-1 — Receiving sequence

  1. Pre-open paperwork check (30 min). Match the packing list to your purchase order line items: SKU codes, quantities, finish codes, version notes. Discrepancy here is a stop signal — do not open product cartons until the mismatch is documented and photographed.
  2. External condition. Photograph any crushed, re-taped, or water-stained cartons before opening. Transit damage is a shipping-claim issue, not a quality issue, and the photographs are what separate the two categories later.
  3. Draw the sample. For lots under 200 cartons, inspect 10% by count, minimum 5 units. For 200–1,000, inspect 5%, minimum 20 units. Above 1,000, inspect 3%, minimum 50 units, spread across pallets — not 50 units from the top of stack one. These are buyer-side defaults; formal sampling plans classify findings as critical, major, or minor defects against an Acceptable Quality Level you and the supplier agree in the purchase contract. Whichever scheme you pick, write it down once and apply it identically to every lot, or your trend data is noise.
  4. Run the checklist below on each sampled unit. Record pass/fail per unit, not just per lot — per-unit rows are what let you compute defect density later.
  5. Disposition the lot the same day it finishes inspection (see the IF-THEN rules further down). Inspected-and-held is better than uninspected-and-shipped, but neither is a steady state.

Procedure R-2 — Receiving checklist per unit

#Check itemMethodFail class
1Carton contents match the printed list (locks, strike plates, screw packs, cards, templates, manuals)Open and count against listMajor
2Physical condition: no scratches, dents, gap inconsistencies, loose or mismatched finish (matte vs. satin visible side-by-side)Visual + hand feel against golden sampleMinor–Major
3Mechanical function: latch extends/retracts smoothly via knob/handle; deadbolt (if equipped) throws fullyManual operation, 5 cyclesCritical
4Power-up: insert batteries (typically 4 × AA on many models); unit boots, keypad backlight lights, no dead-on-arrivalDirect testCritical
5Fingerprint enrollment + 10 successful reads on one finger, 5 on a different fingerDirect testCritical
6RFID/IC card read: card accepted, no repeated double-promptDirect testCritical
7Passcode accepted, including random-digit padding if the model supports itDirect testMajor
8Mechanical emergency key turns without bindingDirect testCritical
9App pairing (TTLock / Tuya / Doodle as applicable): one unit paired per platform version you sell; check firmware string shown in appDirect test, save screenshotMajor
10Battery voltage on installed cells (new-cell reading expected; low voltage at arrival suggests storage age or shelf drain)Multimeter on 2 sampled unitsMajor
11Lot code / date code present and recordedVisual, photographMajor (breaks your traceability)
12Door-thickness range printed on box matches your spec expectation for the market you sell into (many hotel models run 36–100 mm; hidden-reader and glass models differ)PaperworkMajor

Severity definitions, buyer-side: Critical = unit or lot cannot sell or install as-is. Major = sellable after sort or with documented concession. Minor = cosmetic, record it anyway; a lot with 40 minor finish complaints is a trend you will want in six months.

The checklist earns its keep on items 3–9. A smart lock is a motor, a sensor, a battery budget and a radio inside a door furniture shell; a functional sample check of five cycles per unit catches the overwhelming majority of assembly defects that shipped through a rushed end-of-line station.

Verifying the Spec Sheet Against the Units You Received

Here is the part almost no distributor does, and the reason this SOP exists as a written procedure: turning the datasheet from marketing into measurement. Take a real hotel-lock spec block — the SLK-9936S published characteristics — and build the receiving table from it. Every line becomes something you can check, demand, or flag.

Spec line (as published)What it means at your dockHow to verifyIf it fails
Door thickness 36–100 mmBackset/fitting range the lock ships to handleMeasure bolt case and mounting plate; confirm against your market's standard door scheduleLot mismatch to your market = sort or concession claim before sale
RFID Mifare 1 @ 13.56 MHzCredential type locked into firmware/hardwareTest with the card stock you issue; check a foreign card is rejected, not just your ownWrong reader config = critical, quarantine lot
Read distance 0–5 cmUsable hotel card feel at the readerCards at 1, 3, 5 cm on 5 sampled units; note failures at ≤3 cmReader drift vs. golden sample = batch-consistency claim
Read time <0.25 sGuest experience at the doorStopwatch over 20 reads, sampled units; flag mean >spec or outliersRecord with timestamps; CAR trigger if clustered
4 × AA powerBattery logistics you will sell foreverConfirm compartment count/type; note any lot using different cell brands at arrivalCell-brand change without notice = documentation claim
Battery cycles >15,000Design-life claim you cannot test in a dayNot verifiable at receiving — this is a field-tracking number: your 90/180-day battery ticket rate per lot code is the real testField data builds or breaks the claim over time
Audit records up to 1,000Event log depth hotels will demandTrigger 10 events, read log via management software; confirm entries and timestampsLog depth short of spec = critical for hotel SKUs
5-pin cylinder, 3-point latchMechanical gradeThrow the latch; count contact points; check bolt alignment in strikeMisaligned latch = critical, assembly defect
Multilingual software, standalone or centralized modesYour support burdenInstall management software once per lot; test both modesAny new-lot software change = demand release notes

The last row of the battery-cycles entry is the bridge to Stream 2: some spec lines can only be surveilled over months of field data, which is exactly why receiving inspection alone is not a program.

Field Failure Tracking: The Slow Stream That Decides Renewals

Receiving inspection catches what you can see in an afternoon. The field catches what only time reveals: battery drain in a specific climate, fingerprint sensor failures after 5,000 cycles, a finish that pits near coastal properties, a firmware revision that broke app pairing two months after the lot shipped.

Procedure F-1 — Capture rules (IF-THEN):

  • IF a support ticket involves any lock unit, THEN the log entry is mandatory before the ticket is closed — no "replaced, moved on."
  • IF the unit is retrievable, THEN it returns to the Quality Coordinator labeled with lot code, failure mode and install date; do not discard a failed unit without photographing it and recording the code.
  • IF a customer reports a failure verbally, THEN support converts it to a log row the same day, even if resolution is pending.
  • IF the log tool is a spreadsheet, THEN it has one row per failure event, never one row per customer, with these columns:
Log fieldContentWhy it earns its column
Date receivedTicket date, not discovery dateEnables rate-per-period math
SKU + firmware/app versionAs paired or as labeledSplits hardware vs. firmware problems
Lot / date codeFrom unit or carton recordThe only field that links a failure back to a shipment
Failure modeFrom a fixed taxonomy: battery drain, reader fail, motor/latch fail, fingerprint fail, app/pairing fail, mechanical key bind, finish/corrosion, false unlock, otherFree text kills trend analysis
Units sold in lot (running)Denominator estimateTurns counts into rates
EnvironmentHotel / rental / residential; coastal or not; climate if knownLocalizes systematic causes
ResolutionReplaced unit / field fix / returned / unresolvedSeparates hardware from installer error
Photo refFilename of failure photoEvidence for corrective-action claims

Fix the taxonomy once and hold it. Nine months later, "battery drain, 14 units, lot 2024-06, three hotel projects" is a claim. "Some customers complain about batteries" is gossip with an invoice attached.

Procedure F-2 — Weekly review (15 minutes): the Coordinator scans new rows, checks whether any failure mode crossed three occurrences within one lot code, and raises any hit to the Product Manager as a probable cluster. Three same-mode failures from one lot is the standing trigger for a corrective-action request — a number you set per contract, but set deliberately, not ad hoc.

Batch Consistency Review: New Lot Against the Sealed Reference

Silent changes are the specific failure mode of long supplier relationships. Nobody lies to you; the line supervisor accepts the substitute keypad housing because the price and lead time were better, and your next 2,000 units feel slightly different under a guest's thumb. Single lots pass. Reputation erodes across lots.

Procedure B-1 — On every new lot, before sellable status, compare against the golden sample:

AttributeComparison methodJudgment call
Faceplate and finishSide-by-side under same light; note gloss/color shiftsPhotograph + record if visibly different
Keypad feel and backlightPress every key on 3 units; backlight uniformity in a dark roomNotch/click change = flag housing or membrane change
Handle torque and rattleSame hand on both unitsNew looseness = tolerance creep, flag
Battery compartment + contactsOpen both; compare cell count, contact style, brand of shipped cellsAny change unannounced = documentation claim
PCB / board layout (one destructive unit per major lot)Compare component placement, solder quality, conformal coating presenceBoard change without notice = demand explanation
Firmware version stringPair one unit, read app-reported versionVersion jump = request release notes before selling
Box contents and printContents list, manual language set, lot code formatMissing codes break your traceability — treat as Major

Destructive comparison on one unit per major lot feels expensive; it is cheaper than one unannounced board revision that reveals itself as a field failure spike in month four, when the affected lot is already installed across thirty properties.

IF the batch comparison shows drift, THEN the Quality Coordinator photographs both units, lists affected attributes, and issues the standard change query to the supplier before disposition: "Lot XXXX differs from accepted reference in these ways. Confirm intentional change or investigate. Response requested within 5 working days." Suppliers who made the change deliberately answer easily. Suppliers who do not know they made the change answer worse — and that answer is the more important one, because it tells you their internal process control is loose, which is exactly what the factory-side gate verification article was supposed to prove before production started.

Disposition Rules: What You Do With Nonconforming Goods

Inspection without pre-agreed dispositions produces panic decisions at deadline. Agree this ladder with your supplier in writing before the first order — ideally in the same conversation where you compare final candidates, because how a supplier reacts to these rules is itself a qualification signal (see How to Compare Smart Lock Suppliers).

Numbered disposition rules (IF-THEN):

  • D-1. IF critical failures in sample ≥ the lot's agreed critical limit → THEN quarantine whole lot; notify supplier within 48 h with photos and unit counts; supplier proposes containment (sort, rework, return) within 5 working days.
  • D-2. IF criticals under limit but majors over → THEN sort on your side at supplier's cost per written agreement, or return the nonconforming subset with a documented concession note.
  • D-3. IF only minor findings within rate → THEN accept; record defect density per 100 units in the scorecard anyway. Accepted-but-recorded is the whole game: individual minor rates mean nothing, the trend line means everything.
  • D-4. IF field failures from one lot code exceed your agreed threshold within 180 days of install → THEN corrective-action request: supplier explains root cause, scope (which lots share the component/batch), and preventive action; your containment decision (recall/sort/monitor) follows the scope answer.
  • D-5. IF two consecutive lots from the same SKU fail D-1 or confirm unannounced change → THEN that SKU goes to new-sample status: full inspection percentage, no concession accepts, until three consecutive clean lots restore normal rules.
  • D-6. IF findings attributable to transit (external damage photos, consistent single-pallet pattern) → THEN route to shipping claim, not quality claim; misrouting these destroys the credibility of every other claim you file.

Corrective-action request format (one page, five fields): what was found (numbers, dates, lot codes) · where found (receiving/field/batch review) · the evidence (photos, logs, golden-sample comparison) · what you need (root cause, scope of affected lots, preventive action, deadline) · consequence dates (what happens if no response by day 15: stop-shipment, reorder from alternate, claim per contract).

A CAR phrased as an accusation gets defensiveness. A CAR phrased as data with a deadline gets a root-cause report. Suppliers are more likely to answer the second one honestly, and the quality of that answer — specific cause and scope versus boilerplate reassurance — is scorecard material.

Run through the numbered rules a second time and they stop looking like six separate clauses and start looking like one lifecycle every lot travels, with the holding zone, the golden-sample comparison and the field log as its checkpoints.

stateDiagram-v2
    [*] --> Landed_at_dock
    Landed_at_dock --> Paperwork_mismatch_stop : packing list ≠ PO · document before opening
    Landed_at_dock --> Holding_zone : R-1 sample drawn · lot unsigned
    Landed_at_dock --> Shipping_claim : external damage pattern · D-6
    Holding_zone --> Accepted : minor findings only within rate · D-3
    Holding_zone --> Sorted_or_returned : criticals under limit · majors over · D-2
    Holding_zone --> Quarantined : criticals at agreed limit · D-1
    Quarantined --> Contained : supplier sort · rework · return within window
    Accepted --> Field_tracked : F-1 log per ticket · lot code recorded
    Field_tracked --> Corrective_action_request : three same-mode failures one lot · D-4
    Corrective_action_request --> Restored : root cause · scope · preventive action
    Sorted_or_returned --> New_sample_status : two consecutive failing lots · D-5
    Quarantined --> New_sample_status : confirmed unannounced change · D-5
    New_sample_status --> Accepted : three consecutive clean lots
    Accepted --> [*] : sellable · scorecard fed
    Restored --> [*] : scorecard fed

The Quarterly Scorecard and What It Controls

Every stream feeds one table per supplier per quarter. Weights are yours to set; the structure below reflects that field failures matter more than receiving findings, because the receiving numbers protect this quarter and the field numbers protect your brand permanently.

Metric (source stream)InputWeightExample scoring rule
Receiving clean-lot rateR-1/R-2 records20%% of lots accepted with zero criticals
Defect density at receivingChecklist per-unit rows15%Critical+major findings per 100 inspected units, trending
Field failure rate @ 90/180 daysF-1 log ÷ units sold30%Tickets per 100 installed, by SKU; vs. prior quarter
Batch consistencyB-1 sheets15%Unannounced-change incidents per quarter (target: zero)
Corrective-action responsivenessCAR responses10%On-time + specific-root-cause = pass; boilerplate = fail
Claims and goodwill behaviorDisposition records10%Sorts/concessions resolved per agreement vs. contested

Scorecard outputs, decided in writing at the quarterly review: ≥ threshold → normal purchasing continues, explore expanding SKUs against the 79-model catalog you already qualify with this partner. Drift down one quarter → raise it in the review call, request their side's data on the failing metric. Two consecutive quarters down, or any field-failure breach ≥ D-4 threshold → new-sample status (D-5) plus re-qualification of the factory process, using the sample evaluation workbook as the re-entry gate. Decline plus refusal to answer CARs → you are a quality-managed buyer with a non-participating supplier; open an alternate qualification track the same month, and keep the relationship commercially alive while you do.

The scorecard doubles as your renewal and pricing argument. "Your defect density on hotel SKUs doubled across three quarters and two of your CAR responses missed the deadline" is a sentence that moves payment terms, sort-cost responsibility, and stock allocation. Your opinion does not. Your data does.

Illustrative Scenario: 120 Hotel Locks for a Renovation Project

Consider an illustrative scenario — not a customer account, a composite worked example. A regional distributor quotes a 120-room boutique hotel renovation on the SLK-9936S-class hotel lock: 120 units plus 10 spares, RFID card operation, four storeys, June delivery, contractor installs in July.

Under this SOP: at receiving (R-1), the team pulls 26 units (the 200–1,000 band at 5% + minimums would give 20; the distributor adds a self-imposed floor of 25 for project-critical SKUs). The checklist finds one unit with a binding mechanical key (Critical), two cartons missing strike-plate screws (Major), and a backlight dimmer on three units than the golden sample — not a fail on the sheet, but the Coordinator photographs it onto the B-1 comparison anyway. Lot disposition under D-2: sort at distributor cost, supplier notified with counts. Project continues; all 130 units install with the same lot code recorded against the hotel's asset list.

September, month three of operation: the hotel reports three units whose guest cards are read slowly. F-1 capture rules mean each becomes a dated row with the lot code. Three same-mode hits from one code = probable cluster under F-2. The CAR goes out with the September data, the golden-sample read-distance note, and the dimmer-backlight photograph. The supplier's root cause comes back as a reader-board supplier change made in May and announced in a revision note nobody on the distributor's side had connected to the SKU. Two things happen: the affected lots get contained before winter occupancy, and the distributor's change-notification clause gets rewritten at renewal — with the scorecard line "batch consistency: unannounced change incidents 1" on the table when terms are discussed.

The scenario is invented; the mechanic is not. Notice that the distributor was not "wronged" dramatically — a supplier changed a component and most buyers would never have connected it to three slow doors. Surveillance converts a diffuse customer irritation into a specific, dated, resolvable claim with a cause and a scope.

Failure Patterns This SOP Is Designed to Catch

Five patterns account for most of the relationship-ending surprises in smart lock distribution:

  1. The slow component swap. No single lot fails; every lot since March feels slightly cheaper in the hand. Only B-1 golden-sample comparison against a reference from two years ago catches this, because no single lot breaks a rule.
  2. The post-warranty battery cliff. A field-failure rate that sits at 0.8% for ten months and hits 4% in month eleven. Without lot-code denominators in the F-1 log, you discover this as a pile of returns instead of a trendline you saw at 2%.
  3. The firmware mix-up. One pallet pairing to the wrong app version because a revision landed without release notes. Receiving item 9 catches it in week one instead of at 40 installers at 40 doors.
  4. The transit-damage quality claim. Misfiled under quality; the supplier correctly rejects everything; your credibility with the next five legitimate claims is gone. D-6 exists to keep claim categories clean.
  5. The concession that never ended. A one-time accept-with-sort quietly becomes the standard for four lots. D-5's new-sample status rule puts a hard ceiling on how long a known problem ships sellable.

Running the Program: The Cadence Calendar

  • Per shipment: R-1 sequence (paperwork → condition → sample → checklist → same-day disposition), B-1 comparison before sellable status.
  • Per week (15 min): F-2 log scan; count-mode clusters against the D-4 threshold.
  • Per month (30 min): Defect-density and failure-rate trendlines drawn; golden-sample shelf checked (seals intact, demonstration unit rotated).
  • Per quarter (2 hours): Scorecard compiled; supplier review call; renewal decisions taken against the thresholds; standing triggers and thresholds re-read — a threshold nobody reviewed in two years is a formality, not a control.
  • Per year: Re-seal a refreshed golden-sample set per active SKU; re-confirm spec-sheet numbers against current PDPs (published specs evolve; your acceptance criteria should evolve in lockstep and be dated when they do).

Total overhead for a distributor running a two-person warehouse: roughly one inspector-day per shipment plus half an hour weekly. The alternative cost is one hotel project where three hundred doors work "mostly fine" and your phone starts ringing at peak occupancy.

Where This Fits Alongside the Rest of the Quality Library

This manual presumes you already selected the supplier well. If that is not yet true, the upstream documents are the supplier due-diligence checklist, factory evaluation criteria, and the factory-side quality trilogy linked in the introduction. Surveillance does not substitute for qualification — it assumes it and then keeps it honest, lot after lot, on the side of the shipping mark where your customers actually live.

Every procedure above is a template, and every threshold in it is a contract negotiation waiting to happen. If you want the acceptance criteria, sampling bands and CAR response windows written into your next purchase conversation rather than improvised after the container lands, start from the buyer resources library, and where a spec line is not something you can verify at your dock, request the exact figures and documentation against your quote before you order — checking supplier quality is cheapest one shipment before you need it.

Keranjang Anda
Pengalaman Anda di situs ini akan meningkatkan dengan mengizinkan cookie Kebijakan Cookie
Chat on WhatsApp