MS Logo 2

Scaling geospatial solutions for Microsoft’s global reach

Successfully scaled geospatial data acquisition and production pilot project into a steady state program producing high quality imagery and LiDAR data for >90% of US roads for Microsoft’s Streetside product, >75% of US population and dozens of international cities for their photorealistic 3D city product.
KPI

400%

Scaled Production Team by 400% within 4 weeks

KPI

15%

Reduction in cost of 3D city production year over year

KPI

>300

Drivers Staffed Across North America

Company:

Microsoft Bing Maps

Industry:

Technology

Services:

Finance & Operations, Program Management

Project Summary

Successfully scaled geospatial data acquisition and production pilot project into a steady state program producing high quality imagery and LiDAR data for >90% of US roads for Microsoft’s Streetside product, >75% of US population and dozens of international cities for their photorealistic 3D city product. Established processing and production workflows to train ML models on building footprint, road sign, license place, face, and other object detection for incorporation into proprietary base maps.

The Challenge

Microsoft Bing Maps sought to develop geospatial data products rivaling Google, specifically Streeside, global ortho, and 3D cities. After a successful pilot project, they needed to scale up the original vendor team of 50 operators to meet aggressive production goals while accounting for the seasonality of data collection. Initial snags in Microsoft’s software required a radical increase in manual production capacity to hit the project deadline on time, within budget, and to quality specs.

The Strategy

Core to our strategy was aligning our entire team around the awesome project we were all a part of by clearly defining how each function was vital to reaching our overarching objectives and celebrating our accomplishments publicly and regularly as a whole team.

Moreover, by honing the ideal operator profile, aggressively recruiting, establishing a well defined org structure, and developing a detailed training program we scaled the production team from 50 to over 180 within four weeks. Although space constraints and quality demands limited our ability to add more staff to accommodate the necessary production targets, by implementing creative incentives and adopting a leader-leader management philosophy, we effectively managed the rapidly established team while yielding an average of 63 hrs of production per operator per week for a sustained 12 week period.

At the same time, we established feedback loops between our vendor production team and Microsoft’s engineering team to not only fix the initial software development snags but drive radical improvements in tooling and processing pipelines, reducing the cost of 3D city production by 15% year over year.

Upon accomplishing our initial delivery of 70 Streetside and 40 3D cities, we transitioned the team to steady state operations by carefully managing collection backlogs and production staffing to retain our talent and meet quality requirements.

We developed a long term hiring and resource plan, creating detailed models of data collection capacities across North America based on staffing levels, seasonally variable sun angle, foliage, and daylight hours. These models enabled efficient staffing, minimal attrition, and the successful delivery of each milestone. This included scaling our Streetside collection fleet from a dozen sensors to >120 sensors staffed by over 300 drivers operating regionally.

Cross functional communication and coordination was paramount to ensuring smooth data transitions from acquisition to processing, production, QC, staging, and publication. By establishing clear organizational structures, cooperative leadership between functional roles, and providing opportunities for feedback and advancement we drove efficiencies and yielded <5% voluntary attrition over the life of the project.

Functional Areas Included:

Aerial Data Collection and Production

  • AOI Definition
  • Aerial Acquisition
  • Ingest
  • AT
  • DSM Production
  • 3D Edit

Streetside Data Collection and Production

  • AOI Definition
  • Streetside data acquisition
  • Dispatch
  • Ingest
  • Imagery QC

Machine Learning Training

  • Data center operations
  • Object ground truthing and classification
    • Building footprints from satellite and aerial data
    • Signs, license plates, faces, addresses, business data, etc from Streetside imagery

The Results

Despite initial software setbacks, the team not only successfully met the initial aggressive delivery target on time and within budget, but went on to deliver on time, to spec, and on budget for 24 continuous months. We produced high-definition 3D maps covering 75% of the US population, Streetside maps for over 90% of US roads within 2 years, and trained the models for Microsoft’s building footprint program. The project was subsequently sold to Uber.

SITE Consulting

We’ve worked in the field and in the office, so we understand what it takes to get your hands dirty and finish tough jobs. The data and cross functional coordination required to support field teams, attract new customers, and get paid can easily fall by the wayside or worse, get in your way – that’s where we come in.

Implementations