FLM

Outsourcing to the customer

Hans de Zwart and I write a monthly series titled: Parallax. We both agree on a title for the post and on some other arbitrary restrictions to induce our creative process. For this post we agreed to write an essay of no more than 500 words discussing the title in relation to Knowledge, Innovation and Quality. You can read Hans’ post with the same title here.

Recently, I got a new job with a completely different application domain than my previous company. Instead of being an tech eLearning consultant, I am now ‘translating’ obscure imagery to easily interpretable maps; add value to raw measurements and sell that as a service (think buienrader.nl) or a product (think map). My company focuses on water quality so I will use that as an example.

Data sources in Remote Sensing traditionally come from a spectrometer or radar antenna (SAR) which is either on a space, plane or land/sea based platform. All of these instruments have 1 thing in common: scarcity. Space based instruments usually suffer from low spatial resolution (50m to 1km) and/or low temporal resolution (e.g. due to cloud cover). Plane and land based measurements are very expensive and thus not feasible for setting up real-time monitoring services.

But what if we could outsource the measurements to the customer? Then you, theoretically, have an army of measuring individuals at zero marginal cost. The problem is that you’d have to equip them with measuring instruments. But everybody already has such an instrument: each modern cell phone has a camera, GPS and internet connection.

People could just make a picture of the lake they want to swim. The software on their phone sends the image to a server where it’s added to a database with other measurements (a hybrid mix of phone shots, satellite images, scientific field campaign data etc.). Algorithms on the server then process all these data into high quality, near real-time water quality maps.

Now, what would that mean in terms of knowledge, innovation and quality for the company and for the customers?

Knowledge

Key to making customers do measurements is that there must absolutely be no need for a-priori knowledge. All the knowledge should be inside the software on their phone and on the server. In other words, the customers should be ‘stupid’ data collecting drones.

From the data dissemination perspective (maps), there shouldn’t be any ambiguous interpretations possible. It rains or it’s dry, the water quality is good or bad.

Innovation

Innovation is triggered by the fact that there will be a plethora of measurement types. These will differ in: spatial, temporal, spectral and radiometric resolution.

Smart algorithms have to re-sample and re-shape the data so they can be used as input for 1 map.

Quality

Besides the differences in scale of the factors above, there will also be differences in quality. Those will be more difficult to overcome. How can you ‘trust’ measurements from an unknown source? How do you ensure that a single ‘untrusted’ measurement doesn’t infect the overall quality of the output?

The quality of each individual ‘untrusted’ measurement must be considered bad. But by applying statistical techniques to a lot of ‘untrusted’ measurements the quality can be potentially better than a single ‘trusted’ measurement.

By outsourcing measurements to the customer, there is an enormous stream of free data becoming available. The challenge for companies like mine is to channel and filter these data.