The feed said the promotion was live
The brand's retail data showed four products listed at three stores. The promotion agreement called for a temporary price reduction and visible in-store signage.
The account team could see sales movement and reported inventory. It could not answer several basic questions:
- Was the product physically on the shelf?
- Was the promotional price displayed?
- Did the sign appear beside the correct SKU?
- How many facings were visible?
- Which competing brands sat next to the product?
- Was the display consistent across stores?
A retailer feed is useful evidence about the system. It is not evidence of the shelf.
The work becomes a bounded observation task
“Visit these stores and tell us what you see” would produce inconsistent notes. The Taskin brief must define the unit of observation before the participant accepts.
The requester supplies:
- three store addresses;
- four SKU names and reference images;
- a Tuesday collection window from 10:00 to 15:00;
- the fields required for each store and SKU;
- image requirements;
- store-access and privacy constraints;
- fallback rules;
- compensation and direct-payment terms.
For each SKU, the participant records:
- present, absent, or out of stock;
- number of visible facings;
- shelf position;
- regular and promotional price;
- promotion sign present or absent;
- adjacent competing brands;
- wide shelf image;
- close product and price image;
- collection timestamp;
- deviation or access restriction.
The participant does not move products, correct the display, photograph customers, misrepresent themselves as a retailer employee, or bypass a store's photography policy.
How Taskin executes the handoff
The merchandising agent identifies the evidence gap and searches Taskin for field-research capability in Moab. Taskin matches the request with Loren's documented location and in-store research capability.
Loren reviews the route, store list, time window, data schema, photography rules, payment terms, and fallback. She can clarify, decline, or accept.
At each store, Loren collects the same fields in the same order. If a product is missing, she records “absent” or “out of stock” rather than substituting a competitor. If photography is prohibited, she records the restriction instead of taking a hidden image.
The result returns as one record per store and SKU. The agent can compare the physical evidence with the retailer feed and promotion agreement without extracting facts from a free-form message.
What the structured result contains
A usable record includes:
- store and SKU identifiers;
- visit time;
- product state;
- facings;
- shelf position;
- displayed prices and currency;
- signage state;
- named adjacent brands;
- links to evidence;
- access or collection deviations.
The result shape:
store_id: moab_store_02
sku_id: sparkling_lime_12oz
visited_at: <actual timestamp>
product_state: present
facings: 3
shelf_position: eye_level
regular_price: <observed price>
promotional_price: <observed price or null>
promotion_sign_present: false
adjacent_brands: <observed list>
evidence: <artifact references>
deviations: <actual exceptions>Each field was filled from what the participant observed in the aisle, not from the retailer feed.
Acceptance test
The requester accepted the result only because:
- 1.every store was visited or has a documented access failure;
- 2.every SKU has a named state;
- 3.claimed prices have legible evidence;
- 4.timestamps fall inside the collection window;
- 5.images can be matched to the correct store and SKU;
- 6.missing fields include a reason;
- 7.the participant did not alter the shelf;
- 8.the result follows the agreed schema.
The acceptance test evaluates the observation process. It does not guarantee that the display remains unchanged after the visit.
What the brand did with the result
The agent compared:
- listed versus physically present products;
- agreed versus displayed promotional prices;
- planned versus observed signage;
- store-to-store execution;
- product position and competitive adjacency.
A human account owner then decides whether to contact the retailer, request a correction, change local media, or collect another observation.
Taskin supplies current physical evidence. It does not decide whether the retailer breached an agreement.
How the task was measured
| Measure | Evidence |
|---|---|
| Coverage | Stores and SKUs completed |
| Data quality | First-pass acceptance and missing-field rate |
| Speed | Request, acceptance, visit, and submission timestamps |
| Promotion execution | Observed price and signage versus agreed terms |
| Operational effort | Internal coordination time |
| Decision value | Actions taken from the findings |
No revenue impact is claimed from a set of store visits. What this case reports is what was observed, how quickly it returned, and which business decision followed.
When this pattern fits
This workflow fits store audits, price checks, shelf placement, promotion verification, visible inventory checks, public-location observation, and other tasks where the evidence must come from a specific place and time.
It is a poor fit when covert activity is required, photography is prohibited, the participant would need to access private systems, or the conclusion requires statistically representative consumer research.
What the brand kept on record
The brand retained the accepted brief, task reference, store list, timestamps, original images, returned schema, deviations, cost, and its acceptance decision. Store names and commercial terms are withheld here at the customer's request.
Label outcomes accurately:
- Observed: supported by the returned evidence.
- Customer-reported: supplied by the brand after receiving the result.
- Estimated: calculated from stated assumptions.