For R&D teams with a technology looking for its application

Your technology works. Find the problem it pays to solve.

We start from problems someone is already paid to solve.

Fig. 01From technology to applicationsample data
The research ranks the uses. You choose.
tap a step

Step 1 · the capability. What your technology does, in measurable terms. Written as capabilities, with no market in mind yet: what it detects, how fast, and in which conditions.

Step 2 · problems with an owner. Problems someone is already paid to solve. Each one has a person in charge of it and a budget. That person is the one who will judge your technology.

Step 3 · scored on evidence. Each use gets a score, with the evidence behind it. How big the problem is, what people use today, and whether your technology does it better. Each score says if it is measured, estimated or not known.

Step 4 · prototype. A low-cost prototype of the strongest use, tried by the people who own the problem. What they do with it counts more than what they say about it.

Step 5 · your decision. License it, build a product line, test the next use, or stop. You decide on evidence, after spending a research budget and before starting a product program.

Fig. 02Proof

Research we have published, and the method we use on yours.

Published · co-authored. A neural network that finds tables and charts in digitized documents. About → A co-authored paper. The network is saliency-based. arXiv 1804.06236

Published · Walmart Labs. A machine-learning system that picked product images across a very large catalog. About → Add-to-cart rose for tablets, televisions and monitors, and fell slightly for laptops. The paper reports the drop too. arXiv 1811.07996; Causal-impact analysis over 3 weeks

Internal · design record. From the user's need, to the requirement, to the trade-off accepted. Written down for every design decision in our own product.

Argonne National Laboratory. Signal processing built into low-temperature scientific instruments. About → Algorithms that pull a faint signal out of noise, integrated with heavy mechanical engineering, together with several partners. One field brought into another. Meas. Sci. Technol. 19, 115802 (2008); Nanotechnology 20, 055705 (2009); Langmuir 25, 1885 (2009)

Walmart · computer vision. The technology already existed. The question was what to do with it. We found where it paid off: choosing the product photos shoppers see across a very large catalog. The paper above reports the result.

Patent → product · SnapCHART. A patented method became SnapCHART, a plugin that turns a chart into the table behind it. About → 600 beta users. Patent: US 10,726,252 B2

Start with secondary research on your technology, from $5,000 to $10,000.