Leveraging Employee Creativity by Upgrading the Suggestion Box
In a knowledge-driven world, employees are a technology company’s greatest assets. Effectively harnessing the native creativity of technologists can be the key to staying on the leading edge of innovation. That doesn’t just happen by itself, and one of the most important tools for leaders is also one of the oldest: the suggestion box.
The employee suggestion box dates back to the early 1880s, when the Scottish shipbuilders William Denny Brothers Ltd. became the first (known) organization to formally turn to employees for ideas. Toyota advanced the concept in the 1950s, after an internal study identified untapped employee talent as a form of corporate waste.
At Comcast, as in most companies today, the employee suggestion box is digital. While at first, that just meant that employees would e-mail their suggestions, rather than writing them on a slip of paper, we quickly realized that a digital version of an the old analog approach wasn’t going to work in an organization seeking to innovate at Internet speed.
Four years ago, I was asked to rethink how we gather and manage employee-generated ideas from our technology teams. At first, we used an Excel spreadsheet that tracked ideas, but when it got up to 500 ideas, it started to feel like a second full-time job. And we were constantly feeling like we’d already read something before.
Necessity being the mother of invention, we assembled a small team to build software that treats both idea creators and those who “consume” ideas as customers. We knew that the system would need to seamlessly connect ideas to the people in the best position to act on them, rather than depending on human intermediaries to pick and choose.
We call it a Machine Learning Correlation Engine. It consists of two parts: A front-end ideas portal, and a back-end engine that applies machine learning and natural language processing to sift through the ideas, and correlate them.
Here’s how it works: Our software developers, engineers and technologists go to the idea portal. There, they can submit an idea, vote on ideas, comment, check out challenge campaigns, and view idea status. It’s all hash-tagged, to see what ideas are trending. The back-end engine, using machine learning and natural language processing, compares it against ideas that are correlated. This helps us manage about 80% of the ideas that come in; the rest are individually analyzed and handled.
In year one, we received about 30 ideas. Now, it’s up to 8,000 a year. As a result, people often ask: How many of them saw the light of day? The answer is about 1,500 evenly divided between internal improvements and customer-facing advancements. Most of the ideas that get implemented are small – a tweak here or there to make things better or faster – but the combined impact is substantial.
Continuous improvement being a driving motivator, we routinely roadmap new features to both the ideas portal and the back-end. For instance, we recently built an API metrics portal that can slice and dice the idea data any which way, as it relates to innovation.