IoT Development for Data Consultancy
-
IoT Development
- $10,000 to $49,999
- Nov. 2025 - July 2026
- Quality
- 4.5
- Schedule
- 4.5
- Cost
- 5.0
- Willing to Refer
- 5.0
"They told us about the accuracy problem before we found it."
- Information technology
- Barcelona, Spain
- 51-200 Employees
- Online Review
- Verified
Sensor Fusion Lab developed a camera-based system for a data consultancy. The system detects the client's SKUs and flags when a product-facing shelf is empty. It also measures how long the gap lasts.
The system helped the client recover an estimated $16,000 in sales and achieved an overall detection accuracy of 94%. Sensor Fusion Lab delivered the project on time and within budget. The team was proactive in identifying and solving problems, and they communicated well.
The client submitted this review online.
BACKGROUND
Please describe your company and position.
I am the Associate & Senior Consultant of Best Data Solutions
Describe what your company does in a single sentence.
Best Data Solutions is a data consultancy based in Barcelona. We work across the full end-to-end data lifecycle — data architecture and governance, ETL and data transformation, visualization and reporting, advanced analytics and machine learning, plus custom development where an off-the-shelf tool doesn't fit. Our team works mainly with Power BI, MicroStrategy, Azure and other cloud platforms, and we handle SAP and big data integrations regularly.
Beyond project delivery, we offer ongoing support and maintenance and run training programmes, including our Business Data Master executive course. Every engagement is built to the client's specific situation rather than dropped in as a template.
OPPORTUNITY / CHALLENGE
What specific goals or objectives did you hire Sensor Fusion Lab to accomplish?
- Detect gaps on shelf, not gaps in the warehouse. We needed a camera-based system that could look at a defined shelf section, recognise the client's own SKUs, and flag when a facing that should hold product was empty — independent of what the retailer's own inventory system reported, which the client's sales team no longer trusted.
- Measure how long a gap sits, not just that it exists. A five-minute gap during restocking is normal; a six-hour gap is a lost sales day. SFL's job was to time each gap from first detection to resolution, per SKU, per store.
- Keep it about the shelf, not the shopper. The cameras face the shelf, not the aisle, and nothing about who is standing in front of it was ever meant to be captured. This kept the deployment out of employee- and customer-monitoring territory entirely and made retailer sign-off far faster than it would otherwise have been.
SOLUTION
How did you find Sensor Fusion Lab?
- Online Search
- Referral
Why did you select Sensor Fusion Lab over others?
- Close to my geographic location
- Pricing fit our budget
- Good value for cost
How many teammates from Sensor Fusion Lab were assigned to this project?
2-5 Employees
Describe the scope of work in detail. Please include a summary of key deliverables.
The project ran as a single, contained pilot in one supermarket rather than a wider rollout: in Phase 1, SFL installed shelf-facing cameras over the client's snack category — 8 SKUs — (stock-keeping unit ) in that store, trained detection against the client's specific packaging, and validated results against a manual shelf-walk done by the client's own merchandising team.
Phase 2 stayed within that same store and focused on getting detection reliable across all 8 SKUs and the store's specific layout — shelf height, lighting and planogram position — before any conversation about scaling further; and Phase 3 built the gap-duration logic, connected it to the client's merchandising team through automated alerts, and handed over a dashboard for the client's category manager.
RESULTS & FEEDBACK
What were the measurable outcomes from the project that demonstrate progress or success?
Average gap duration fell from 3.8 hours to 50 minutes once alerts reached the store's merchandiser directly instead of waiting for the next scheduled shelf-walk.
Estimated recovered sales of roughly €16,000 over the eight-month pilot, calculated by the client's own commercial team from the reduction in stockout hours against historical sales velocity per SKU — modest in absolute terms, but enough on an 8-SKU, one-store scale to justify the next conversation about wider rollout.
Detection accuracy of 94% overall, though this varied by packaging — the one clear-plastic multipack SKU sat closer to 85%, which is exactly the kind of gap the accuracy report was built to surface rather than hide. One recurring problem SKU identified, where gaps kept reappearing on the same product regardless of alerts — traced to a restocking habit specific to that store rather than a detection issue, which the data made undeniable rather than anecdotal.
Describe their project management. Did they deliver items on time? How did they respond to your needs?
On budget and on schedule, with one real hurdle along the way. The hurdle: the store's one clear-plastic multipack SKU was badly under-detected in the first weeks of the pilot — the model kept missing it against the shelf's cooler LED lighting. This was the client's second-highest-volume product in the category, so it mattered.
The response: SFL flagged the accuracy problem themselves before the client noticed it in the data, rather than letting a quiet false negative sit in the dashboard looking like a clean result. They retrained specifically on that packaging under that lighting, brought accuracy from around 68% to 85% within two weeks, and were honest that 85% was the ceiling for that packaging type rather than promising a number they couldn't hit.
Day-to-day communication ran through a shared Slack channel and a weekly call with the client's category manager, with SFL joining directly when packaging or detection questions came up rather than routing everything through us.
What was your primary form of communication with Sensor Fusion Lab?
- Virtual Meeting
- Email or Messaging App
What did you find most impressive or unique about this company?
They told us about the accuracy problem before we found it. It would have been easy to let the dashboard show a plausible number and let the client assume the multipack SKU simply wasn't going out of stock much. Instead they flagged it, quantified the gap themselves, and fixed it. For a system whose entire value proposition is "trust this number instead of your gut," that kind of self-reported honesty is worth more than the fix itself.
The second thing was how narrowly they scoped the cameras. Several of the client's retail partners were nervous about any camera near a customer-facing aisle, and SFL's answer — the field of view is the shelf, full stop, nothing else is in frame — closed most of those conversations in a single meeting rather than a month of back-and-forth with each retailer's legal team.
Are there any areas for improvement or something Sensor Fusion Lab could have done differently?
Packaging variety should have been stress-tested from day one, not discovered mid-pilot. SFL's initial model training leaned on the category's best-selling SKUs, which happened to have straightforward matte packaging. The clear-plastic multipack issue only surfaced once all 8 SKUs were live in Phase 2. With only one store and eight products to test against, there was no excuse not to include the awkward packaging in the very first training pass.
Lighting variation within the single store was underestimated at quote stage. The original proposal assumed one calibration pass would cover the whole shelf section; in practice the aisle has two different lighting rigs a few metres apart, and calibration had to be redone more granularly than budgeted. It didn't blow the project cost, but on a pilot this small it shouldn't have been missed in the initial site survey.
RATINGS
-
Quality
4.5Service & Deliverables
-
Schedule
4.5On time / deadlines
-
Cost
5.0Value / within estimates
-
Willing to Refer
5.0NPS
