How Autometrics Built a Demand Sensing Model with Hundreds of Datasets

by   |   August 9, 2013 4:07 pm   |   0 Comments

With a product that costs thousands of dollars to make, market and sell, automotive companies need every possible insight into the whims of potential customers.

A company called Autometrics has built a business around sensing demand for automakers by tracking the Web browsing habits of potential car buyers and reporting them to the manufacturers on a daily basis.

The major car companies can use the demand data to adjust production schedules, tweak marketing campaigns and minimize inventory, greatly increasing profit margins in a very competitive industry.

Related Stories

SAP’s SmartOps acquisition signals move to add demand sensing to HANA for supply chain management.
Read the story »

In quest to build big data platform, Actian adds ParAccel to its list.
Read the story »

B2B firms inject data analytics into pricing, where instinct has ruled.
Read the story »

Inside L’Oreal’s plans to implement global supply chain analytics.
Read the story »

That’s the power of demand sensing: stitching together data from marketing, sales, production and supply chain systems to offer up-to–the-minute insights on what buyers want.

Autometrics pulls data from 150 different third-party car-buying websites, allowing the company to understand what cars are drawing interest from buyers, as well as when and where that interest lies. Using predictive analytics, Autometrics can use that information to predict sales figures and sell that information to the auto manufacturers as a service.

Autometrics has built a business around its ability to get the demand data car manufacturers can’t get. But many manufacturers in other sectors are starting to use demand sensing as part of their forecasting process by connecting sales, marketing and supply chain systems data from within their own organization. This gives them a better picture for their marketing and sales promotion effectiveness and can save money in more efficient inventory, warehouse and logistics management.

Data Sensing as a Service
Autometrics started in 2000, aggregating and integrating data for automakers and delivering weekly sales reports via email and the Web. In 2007, CEO and founder Stephen Shaw decided that there was a larger, more profitable prize and his company shifted gears.

“As the company started to evolve, we realized that the power was not so much in the integration of data itself, but more in the new capabilities that emerge once you have all this data in one place,” Shaw said.

The shift included an increased emphasis in building predictive models with data that wasn’t available to automakers. Shaw said the analysis of these datasets would “lead the automotive industry through to far greater efficiencies in production, marketing, and sales incentives based on the intelligence within data.”

In order to be successful, Shaw said his company had to get the data that would be predictive of future sales, something he calls “lower funnel prospects.” (The reference is to the purchase funnel model, describing a consumer’s activities of considering and then purchasing a product.)

Shaw said that getting the commercial agreements with the third party car-buying websites proved to be the most difficult part of the process; he started reaching out to the websites in 2004, and it wasn’t until 2007 that he felt his company had the critical mass of data to start selling insights to customers. Today, he has data from approximately 150 websites, he said.

Part of his commercial agreements with those sites prohibits him from citing them by name, but Shaw said that none of the sites belong to individual automakers. This is by design.

“If you go to, the only vehicle you can select from a list is a Ford vehicle,” Shaw said.  “If you go to a dealer website, similarly, you can only choose from the cars they sell there. When you go to a third-party site, generally you’re able to select any vehicle in the marketplace.

Autometrics is able to measure the effectiveness of marketing campaigns; this year, they noted in February, the Lincoln MKZ’s commercial “Phoenix” was the most successful car ad during the Super Bowl, increasing the number of prospects by 5,800 percent.

While marketers no doubt find that information interesting and useful, those types of insights have serious impacts on the factory floor, too.

“Our data is being used to make production decisions,” Shaw said. “Automakers [traditionally] have to use gut feel as to: Do they move a plant to third shift? Do they shift plants if they can between model A and model B? Making those kinds of decisions.

“Those decisions are worth tens of millions of dollars if they get them right,” Shaw said. “That decision can make or break a profit line for a given model in a year, and our data is one of the key data sets on the table at that moment in time being used for major decisions in that.”

Autometrics uses two pieces of what’s now the Actian Big Data Analytics Platform, ParAccel’s columnar database and Pervasive Software’s data integration software. Shaw said he was using both products before Actian’s acquisition spree, but he said the fact Actian has brought both products under a single umbrella is a signal he was on the right path.

Lance Speck, formerly of Pervasive Software and now Actian’s vice president of integration, said that Shaw is one of the first people to understand the value of data over applications.

“Autometrics [started] truly challenging the status quo,” Speck said. The company, he added, was “talking to people who typically don’t think in this way, and said, ‘You can gain competitive advantage, you can gain cost advantage, you can gain insight that you never could before. It’s all spilling out on the floor in the form of the data that’s available, and you’re not taking advantage of it.’”

Getting the Right Flavor to the Right Store
Autometrics’ business model is gathering data from other organizations and providing its demand sensing insights to its customers. But large companies can find similar successes by integrating their own data across silos and applying some predictive modeling.

In a case study published on analytics software company SAS’s website, Nestle’s Direct Store Delivery business increased its forecasting accuracy by 4 percent. Nestle Direct Store Delievery sells ice cream and frozen pizza, which is a competitive and complex market due to the wide variety of flavors and styles.

Nestle’s stated goal for this business is “right flavor, right time, right store.” Because of the complexity and competition, much of the business is promotion driven, so any insights into the potential success of the sales and marketing campaigns can help the business produce and ship the right product to the right store to better meet demand.

By integrating data across marketing, sales and the supply chain and running SAS’s analytical products, the company has been able to reduce inventory, storage costs and freight costs, according to Geoff Fisher, Nestle’s director of supply chain.

In a video Q&A posted on SAS’s website, Fisher said the increase of four percentage points in forecast accuracy has had a good financial impact.

“If you looked at the raw numbers, our finance people would tell us that we’re flat,” Fisher said. “I would tell you: great, because the business grew, the channels grew and we were able to maintain the service level at the current inventory. So it’s a net reduction, it’s a cost avoidance.

“We would have had to build and protect against forecast accuracy that we don’t have to build now,” Fisher added. “There is real capacity that we’re not using, and there is really variable labor dollars in our plants that we’re not having to spend against products that aren’t going to get sold—or worse—that we didn’t make products that could have been sold.”

Email Staff Writer Ian B. Murphy at

Home page photo of cars for sale by Flickr user Helgi Halldórsson. Used under Creative Commons license.

Tags: , , , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>