SiGNALS Analytics, LLCConfidential Information 1

Attachment C – Initial Market Analysis

This attachment describes a preliminary view of the market for first responder mobile broadband within

Colorado

March 23, 2017

Analytics, LLC

DISCLAIMER

This RFP was prepared using funds from the State of Colorado - Governor’s Office of Information Technology. The statements, findings, conclusions, and recommendations are those of the author(s) and do not reflect the views of the National Telecommunications and Information Administration (NTIA), U.S. Department of Commerce (DOC) or FirstNet.

Some analyses were prepared by the State of Colorado - Governor’s Office of Information Technology using funds under award 08-10-S13008 from the National Telecommunications and Information Administration (NTIA), U.S. Department of Commerce (DOC). Other analyses referenced were done with non-SLIGP funds, consistent with the requirements of SLIGP funding.

Introduction (1/3)

• Attachment C contains three separate but complementary studies:

– A preliminary Demand Forecast estimating the addressable market for wireless broadband among Public Safety Entities (PSE) within Colorado.

– A view of Radio Access Network Utilization (%) over time, on a county-by-county basis, using the Demand Forecast and the Sample Network Design (Model #8) plus a few additional assumptions.

– A Survey of mobile broadband-related end-user needs and behaviors of Public Safety Entities within Colorado.

• The preliminary Demand Model draws upon primary and secondary research to estimate the demand among Public Safety Entities for mobile broadband voice and data. It attempts to estimate:

– Number of first responders, by year (addressable market)

– Number of devices per first responder, by year

– Usage per device per month (voice and data), by year

Introduction (2/3)

• The preliminary Demand Forecast develops a viewpoint specific to Colorado, and different from the numbers published by FirstNet in its RFP.

• The preliminary Demand Forecast also attempts to estimate the number of first responders (addressable market) by county. This exercise is important because network resources are geographically located. In general there is a lot more radio resource per POP in rural areas, because they are coverage-constrained, than in urban areas.

• If a network similar to the Sample Network Model is deployed then the “excess capacity” is rural areas will last significantly longer than the excess capacity in urban areas.

• The preliminary Demand Forecast leverages the Survey results. The analysis was done in 2015 and the results have not been updated. An analysis done today would be different to the degree to which trends visible in 2015 shifted from the anticipated trajectory.

Introduction (3/3)

• Next, we attempt to show the relationship between the Sample Network Design (Model #8) and the results of the Demand Forecast.

– This analysis requires additional assumptions about adoption of mobile broadband services and the market share associated with FirstNet verses commercial mobile broadband offerings.

– We currently don’t know enough about FirstNet’s future offering to predict adoption or market share with any confidence. We therefore use simple “placeholder assumptions” that the bidder can replace with his or her own assumptions.

– The results are significant in that they show utilization growing at different rates in different geographic areas. These findings are driven more by demographics and network design than by penetration and market share.

– An view of utilization is possible only when one has attempted to forecast demand and when one has developed a sample network design, which few states, other than Colorado, have done.

• Colorado will undertake a number of analyses in 2017 that may produce additional significant insights.

Other Notes

• Colorado makes no commitments about user adoption with these demand forecasts. These are estimations only; no agency or user is required to purchase CO-RAN or FirstNet services or devices.

• Colorado’s two federally recognized tribes, the Southern Ute Indian Tribe and the Ute Mountain Ute Tribe, are reflected in data for Archuleta, La Plata and Montezuma counties when not identified specifically.

I. Demand Forecast

Introduction (1/2)

• In the following pages we will present a comprehensive view of possible demand for first responder mobile broadband service from 2015 through 2032. We will identify:

– The types of organizations / functions that the state considers public safety entities (PSEs)

– The number of potential users by entity / function

– The geographic location of potential users by county

– The number of types of devices

– The monthly usage per first responder, in GBs (for data) and in minutes of use

(for voice)

• In developing these forecasts we leveraged:

– Local data sources

– Federal data sources

– Local survey results

– Analyst forecasts of broad consumer trends in adoption, wireless and wired broadband usage, and categories of devices

Introduction (2/2)

• We will forecast the size of the addressable market. We will not address the rate of adoption of FirstNet services, since this will depend upon a set of value propositions (coverage, feature/functionality, and pricing) that have yet to be defined.

• In estimating market size we identify the general categories of entities to include. In some cases we include 100% of those employed in the space. In other cases we have prorated the numbers based on the portion of employees we believe would reasonably need access to a first responder network.

• We have assumed the widespread availability of consumer / Band 14 devices (since virtually every mobile device on the market supports multiple bands). We have further assumed that a consumer could purchase a consumer device and be given Band 14 access to use occasionally for a critical job function.

• We believe that there will be wide differences in the amount of “first responder” usage per person (as opposed to ordinary consumer usage) based on job function. For simplicity, we have not discounted usage at this point in time to reflect different types of users.

Addressable Market – Number of Potential Users

Number of Potential Users - Methodology

• We started with a broad categorization of government employment (federal, state, tribal, and local) and identified those job functions that would be 100% first responders (e.g. police, fire, EMS) and those job functions that would have a tangential first responder requirement (e.g. schools / educational institutions).

• We further analyzed the distribution of government functions (federal , state, tribal and local) by county within the state of Colorado. We discovered that the percentage of the population of Colorado employed in government functions (6.99%) was similar to the portion employed in government functions in the United States as a whole (7.83%).

• We identified relevant roles and assigned a percentage (up to 100%) of those functions to first responders. We arrived at 6,872,084 first responders nationally, of which 118,452 (in 2015) reside in Colorado.

• We then mapped those first responders to the 64 counties of Colorado (inclusive of the two tribes), based on known government employment by county.

• Finally, we “sanity-checked” the result against very local “bottom up” sources that are very accurate, but represent only a subset of the intended users.

Where Are the First Responders?

1%

2%

19%

6%

11%

7% 2%

0%1%

3%

11%

7%

8%

3%

1% 0% 3%

8%

1% 5%

Judicial and legal

National defense and international relations

Police protection

Fire protection

Corrections

Highways

Air transportation

Water transport and terminals

Public welfare

Health

Hospitals

Parks and recreation

Natural resources

Water supply

Electric power

Gas supply

Transit

Education

Postal service

The Department of Homeland Security

The numbers above represent people, not organizations

How Does this Forecast Compare with other Data Sources?

20,000

40,000

60,000

80,000

100,000

120,000

Police, Fire, EMS (circa 2014)

INCLUSIVE Current Estimate

N um be r o f R es po nd er s

Comparison of Data Sources

Other

Police, Fire, EMS

Numbers from very different sources tell a similar story

First Responders in Colorado, Forecast 2015 - 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032

Population of Colorado 5,439,290 5,532,049 5,628,099 5,726,144 5,825,318 5,924,692 6,025,325 6,126,642 6,627,956 7,095,836

TOTAL 118,452 120,472 122,563 124,698 126,858 129,022 131,214 133,420 144,337 154,526

We expect that the number of eligible first responders associated with public safety entities will increase over time in proportion to the population

First Responders by County, Forecast 2015 - 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

First Responders by County, Forecast 2015 - 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

TOTAL 118,452 120,472 122,563 124,698 126,858 129,022 131,214 133,420 144,337 154,526

Number and Types of Devices

What About Devices?

• FirstNet Colorado conducted a survey of public safety entities in Colorado to understand what kinds of mobile broadband devices are in use and who purchased them.

• The survey had 147 responses (each representing an organization). We have used information from the survey – which seems to align with our expectation from external data sources – as the starting point for our beliefs about the number of devices and the usage per person.

• We have scaled the number of devices going forward, based upon observable trends in the consumer marketplace that we believe will apply to first responders, as well.

• Here are a few data points:

– There are currently 1.2 devices per first responder, many provided by the agency, and many personally owned.

– Some devices support voice; others are data only

– Some devices are handheld; others are installed in vehicles

– The graph on the next page shows many of these permutations

Device Ratios

5%

7%

1%

0%

20%

40%

60%

80%

100%

120%

140%

DEVICES /

Person

ENTITY

Voice + Data

DEVICES /

Person

ENTITY Data Only

DEVICES /

Person

ENTITY

Other

DEVICES /

Person

M2M

DEVICES /

Person

Individual V&D

DEVICES /

Person

Person

Vehicle

DEVICES /

Vehicle

De vi ce R at io s (

ot al S ta ff, U nl es s O th er w is e In di ca te d)

PSE-Provided Devices: Voice + Data vs. Data

7,082, 64%

2,870, 26%

1,072, 10%

Other

Number and Distribution of PSE-Provide Devices

5,925, 54%

1,412, 13%

505, 5%

1,157, 10%

1,458, 13%

567, 5%

PSE-Provided Devices: Persons vs. Vehicles

7,842, 71%

3,182, 29%

Number and Type of Devices, Forecast 2015 - 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032

Devices in Use by County, Forecast 2015 – 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

Devices in Use by County, Forecast 2015 – 2032

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

TOTAL 145,695 153,366 161,284 169,411 177,718 186,172 195,015 204,243 256,147 317,907

Usage per Subscriber and per Device

What About Usage? (1/2)

• We were able to calculate from our survey a weighted average usage per device, across the Colorado.

• The level of usage, in GBs per month, matched almost perfectly, what one would expect in the consumer marketplace.

• We started with that figure, 2.64 GBs per month per device, and scaled it going forward, based on the industry’s view of mobile broadband data usage in the United States.

• We have a reasonable accurate forecast for the next five years. We then continue to grow usage based on the growth rate of the fixed internet through 2032. It is important to distinguish between growth rates in different geographies (e.g. the United States vs.

the world) and to distinguish transient effects (migrations from feature phones to smart phones or from one radio access technology to another) from long term effects (growth in demand due to increasing capable devices, more radio spectrum, incremental technology advances, etc.).

What About Usage? (2/2)

• Separately, we calculated voice usage per person based on the most recent available

CTIA figures on voice usage. We then distribute that voice usage across all applicable voice-capable devices (e.g. more devices does not necessarily mean more voice minutes).

• Data is calculated in GBs and voice is calculated in minutes (MOUs)

• Even though mathematically one can calculate the number of bits in vocoded voice, these are not the same as bits in – for instance – a file being transferred. Voice is extremely latency sensitive and some bits are much more important than others. Also, LTE has radio layer techniques that can be applied to maintain voice quality under adverse conditions on the premise that voice is usually more important to the user than a comparable amount of data.

• We estimate the capacity of the network for voice alone and for data alone then attribute an appropriate number of “equivalent GBs” for voice usage. The “cost” of voice is therefore greater than the number of bits it represents. Even so, data now dominates the demand for network resources, so how we calculate voice usage is becoming less important every year.

Data Usage per Device per Month (GBs)

0.

1.

2.

2.

3.

3.

2.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

M2M Module Smartphone 4G Smartphone Tablet 4G Tablet Laptop Colorado FIRST

RESPONDERS

U

SA

G E pe r D ev ic e pe r M on th (G

Bs , W id e Ar ea N et w or k)

Colorado First Responder Usage vs. Consumer Usage

GBs and Minutes per Month, Forecast 2015 - 2032

Voice is flat (increasing only in response to an increased number of users) but data will grow exponentially, reflecting Moore’s law, the growth in

Internet traffic in general, and the rapidly increasing capabilities of devices.

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 Usage Usage per Device

Data Usage by County, Forecast 2015 – 2032 (GBs / Month)

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

Data Usage by County, Forecast 2015 – 2032 (GBs / Month)

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

TOTAL 384,636 493,491 623,784 773,397 957,662 1,184,169 1,464,143 1,810,010 5,201,330 14,791,589

Voice Usage (Millions of Minutes / Month)

Voice Usage (Millions of Minutes / Month)

Increasing Data Usage per Device and Devices per User

2.6 3.2 3.9 4.6

5.4 6.4

7.5 8.9

20.3

46.5

0.5

1.0

1.5

2.0

2.5

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

45.0

50.0

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032

LT

E De vi ce s p er F irs t R es po nd er

Da ta U sa ge p er D ev ic e (G Bs p er M on th

Usage per Device

Devices per Subscriber

This chart shows yearly intervals through 2022 then 5-year intervals through 2032. The change in slope after 2022 reflects, to a large extent, this change in the horizontal axis.

II. Utilization of Sample Network

Estimating Network Utilization (1/4)

• In examining demand relative to the capacity of the network several critical questions arise:

– How much of the total available capacity will first responders consume?

– When will the network run out of capacity?

– Will first responders generate enough revenue to cover the cost of the network?

• In this section we will attempt to answer the first two questions.

• In each case the simple answer is “it depends”:

– It depends upon the year (we assume that data consumption per device is increasing at 18% per year in perpetuity. In addition, users increase and devices per user increase)

– It depends upon the location (the most populated counties have 1,500 times the demand of the least populated counties)

– It depends upon how the network is enhanced over time

Estimating Network Utilization (2/4)

• Our analysis makes several significant technology assumptions:

– We assume that no new sites are constructed

– We assume that no initiatives exist that would cause an increased portion of traffic to be offloaded to Wi-Fi or to LTE small cells

– We assume that no other spectrum is used (e.g. LTE-U, other mobile operator bands, millimeter wave, etc.)

– We assume that overall system capacity increases at 5% per year (much less than demand) after 2018 as a result of technology improvements in the network and in devices.

• We assume that only a portion of the subscribers in the addressable market purchase service:

– We assume 50% (penetration) of first responders in the addressable purchase service from someone. Many might not because they are office workers or otherwise don’t need prioritized wide area wireless broadband communications.

– We assume 50% purchase from existing commercial mobile operators (all of whom will be able to offer prioritized services, if they wish, similar to FirstNet) while the other 50% purchase from FirstNet or the CO-RAN (market share).

– These are “placeholder” assumptions, to make the numbers work, not assumptions we necessarily believe. It is simply too early to know what market share FirstNet or the CO-RAN will achieve and why.

Estimating Network Utilization (3/4)

• One would expect adoption to occur slowly over time as the network is built, as early adopters try the service and like it, as PSEs become convinced of its quality, etc. In this section we disregard these effects entirely. We assume – for simplicity – that 25% (e.g.

50% x 50%) of the addressable market adopts instantly. The bidder can easily adjust the number to reflect other more realistic assumptions.

• It is important to note that our definition of “100% capacity” contains a number of important margins such that “100% utilization” is a level that the network can confidently deliver without any visible degradation of service.

– At levels of utilization above 100% there is increased risk of degradation in hot spots and/or at busy hour

– We assume that 1 Mbps of system capacity in a particular location delivers 47.08

GBs of billable data usage per month

– We similarly assume that 1 million minutes of voice usage consumes the equivalent of 87.56 GBs of billable data usage

– These ratios reflect differences in directionality, differences in the shape of busy hour curves, differences in overheads, differences in channel loading, etc.

– Of course, it is important to compare local demand with local capacity. We analyze capacity utilization, therefore, on a county-by-county and year-by-year basis.

Estimating Network Utilization (4/4)

• We ran the radio system analysis at 100% of downlink capacity (resulting in near worst case interference levels) but assumed in the traffic calculations that at most (during momentary peaks during the busy hour) half of that capacity is used per sector.

• The analytical result is therefore slightly more conservative than the results produced by respondents to the FirstNet RFP, based on the assumption sets in the RFP. In the RFP the radio system design is simulated with a 50% load. In such a scenario interference levels are also substantially lower and therefore the calculated data speeds are higher.

• We assumed that first responder data traffic is 67% downstream, consistent with beliefs articulated in recent conference proceedings. Consumer data traffic, of course, is more asymmetrical. Voice traffic is symmetrical.

• We have color-coded the utilization tables as follows:

– Orange, if utilization is greater than 80%

– Red, if utilization is greater than 100%

Required Capacity (Voice + Data) in Equivalent GBs / Month

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

DATA and VOICE USAGE by County (Equivalent GBs per Month)

Required Capacity (Voice + Data) in Equivalent GBs / Month

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

TOTAL 390,882 499,843 631,565 781,314 965,715 1,192,359 1,472,473 1,818,480 5,210,493 14,801,398

DATA and VOICE USAGE by County (Equivalent GBs per Month)

Required Capacity (Voice + Data) in Mbps

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View

SYSTEM Capacity Required (Mbps Up + Mbps Down)

Required Capacity (Voice + Data) in Mbps

2015 2016 2017 2018 2019 2020 2021 2022 2027 2032 County View