A tenant’s perspective on Brazil’s towercos: the need to diversify services for 4G

How has 4G network planning and rollout program management evolved in the age of shared infrastructure? How much flexibility can towercos offer a 4G rollout in Brazil, and what does it really cost? TowerXchange secured rare insights from the front lines by speaking to Steve Roberts, who until recently, was heading up the rollout of 300 sites for 4G startup ON Telecom.

TowerXchange: Please introduce yourself Steve, what’s your background and what brought you to Brazil?

I’ve been in telecoms for 30 years, starting out in the UK when everything was being privatised – I was involved in the first installations of the Orange network. I came to Brazil in 1997 and found Brazil going through the same process the UK had done 10-15 years earlier – the deployment of large cellular networks had to be done by tomorrow. We started by rolling out TDMA – here known as Banda B – which grew into GSM, 3G and now 4G rollouts. Nortel Networks first brought me over to Brazil, then I ran turnkey projects for Alcatel, Huawei (activating 1,000 GSM sites in São Paulo in six months!) and Nokia Solutions Networks.

I took all that turnkey deployment project experience to become one of six founders of ON Telecom, a 4G fixed wireless network startup with a specific coverage objective; we had the rights to install sites in 133 cities in the state of São Paulo, a high income region with lots of class A and class C customers.

We started up in April 2012 in a small office in São Paulo with a concept to design and implement 4G networks with the minimum of manpower and cost, and the maximum coverage and QoS – QoS can be, and still is, a big differentiator in Brazil!

It took us eight months to make our first 4G connection from São Paulo to New York – a Skype call to the office of our sponsor George Soros – then we deployed four fibre rings and almost 300 sites over the next two years – the majority of which were through co-locations on shared towers.

TowerXchange: Tell us about network planning at ON Telecom – how did you prioritise sites within the rollout?

From my experience with the vendors I learned that when operators ask for full turnkey solutions, it never quite works out as easy as you hoped. So we decided to do all our network planning, design and implementation in-house. Engineering capability and products in 2012/13 for TDD-LTE weren’t common, it was FDD that was being driven. Much depended on China Mobile who were the benchmark at the time, so we decided that we would retain control of design to avoid vendors selling us more equipment than we needed – we felt we could do the same job as the vendors for 30% of the cost, while maintaining quality. This was a massive cost saving.

When it came to tower infrastructure and equipment deployment, we planned everything and did it by ourselves, and used the same subcontractors as the vendors, negotiating a big discount if we paid on time – paying on time can also be a big differentiator in Brazil!

This methodology is still be used to date.

When deploying sites, we moved through a hierarchy of five deployment scenarios, in declining order of preference:

1. We preferred to share rather than build towers wherever possible. It almost always made more sense to rent than build, even where we were exposed to high ground rent, or where the height our equipment was to be mounted on the tower was not ideal. So we established an LPU with each towerco with an agreed rent for two, three of four square meters.

2. If we couldn’t co-locate on a shared tower, our second alternative was using a rooftop site. But rooftop sharing is always problematic in Brazil. Many rooftop sites are not properly licensed, and it can be difficult to deal with other operators, so we often built our own rooftops to ensure flexible access and to secure a significant reduction in rent (opex).

3. If no existing shared tower nor convenient and properly licensed rooftop location was available, our third alternative was to commission a build to suit site. The opex was a lot higher than sharing an existing site. But the cost was only part of the problem; time to market was the other. Site hunting takes as long as it takes, but the real drawback is the pain of licensing – permitting new sites is notoriously slow in Brazil, but necessary (it’s too risky to build and sort out permitting afterwards – I’ve seen illegal sites having to be torn down at great expense and significantly impacting subscribers).

4. Option four was to build our own greenfield sites. We built two and actually tested the waters to see if we wanted to become a towerco. But with the enormous capex involved, we felt we should devote the company’s limited capex and sales resources into our core business; selling modems had to take priority over selling space on towers.

5. The option of last resort was to share with an operator. Even though infrastructure sharing is mandated in Brazil, the operators make it very difficult, and we often found that their towers were in a poor state of repair. It would take six months to a year to use their site down to internal “Red-Tape” and the cost to strengthen their tower meant our site would no longer be profitable.

TowerXchange: How flexible were the towercos able to be given the unique requirements of a 4G fixed wireless network?

We were able to secure special rates from some of the towercos based on our LPU and the relatively small amount of space our equipment needed on the tower. We often had to be able to be very flexible about heights because we did not want to lose the location – our equipment could be mounted almost anywhere on the tower. However ON Telecom is a fixed wireless network so we always had to work closely with our RF Engineering team so as to maintain maximum coverage. The concept is very different from a “Mobile Network”.

Some of the smaller build-to-suit towercos would offer six months “free” tower space rental so as to accelerate their portfolio. Others would invoice towers based on fixed quotas per month, if we passed the forecast the additional towers would only be invoiced at the end of the year. Delaying rental costs for six months certainly helped with cash flow for a startup! Note: all contracts are based on ten years rental.

At a few marginal sites representing cities and towns where our forecasts suggested we couldn’t be certain we would generate adequate revenue to justify the lease, a couple of the big Brazilian towercos granted us a “sale or return” six month cancellation clause, a gentleman’s agreement which gave us the flexibility to enter new markets.

At all our sites, all the infrastructure and installation was taken care of by the towerco. And in most cases any tower strengthening costs were borne by the towerco too.

TowerXchange: What proportion of your initial sites were leased from towercos?

In our first year of rollout, approximately 70% of our sites were on shared sites leased from towercos, the rest were all rooftops, so all our sites came from our preferred deployment scenarios one and two as mentioned previously. Using towerco sites and rooftops minimised up front capex and maximised speed to market.

In many cases we used desktop software to plan transmission, but it was mandatory that we  had to visit the site to conduct line of site surveys. We surveyed every site as we could never be sure there was actually structural capacity available for our equipment, even if the towerco stated there was space; sometimes we would find the site was more loaded than the towerco realized. We’d present such surveys to the towerco who could then have illegally installed equipment removed. This helped them to rent more space and saved on tower strengthening.

In year two, we moved into the interior of the state of São Paulo and had no option but to deploy more build to suit, and in some cases had to share some existing operator sites, which slowed us down.

TowerXchange: Give us a rough comparison of time to market, build to suit site compared to co-location on an existing site?

A greenfield build to suit site could take anything from five to twelve months to get on-air, after we had gotten through the complex and extremely bureaucratic licensing process. While a towerco existing site open to being shared might take two to three months – and most of that time was taken up by the internal towerco process once the site was qualified, and by transmission planning.

RF coverage is of course the focus of a project like this, but full respect must be given to Tx, in our case mainly microwave. No transmission = no site! There could be plenty of sites out there, but factors such as tower height, space (larger microwave antennas were used the further we progressed into the “interior”), unavailable frequencies or channels meant topology redesigns. However in general the existing “shared” site had a much quicker time to market compared to build to suit.

Operators’ sales forecasts are based on coverage, they don’t take into account the time it takes to get on air and how the competitive landscape might have shifted in the meantime. So time to market is critical.

We tended to use build to suit in cities where the big operators were not yet present, hence there were no existing towers! This was good for the towercos because where we went meant that the other operators could consider new areas, therefore giving the towerco the opportunity to lease up more tower space.

TowerXchange: How did the spectrum ON Telecom were using for 4G affect network planning?

Since ON Telecom were running 4G on 2.6 GHz, our sites were limited to coverage of 500m. With 10MB/s plans, we had to concentrate the signal, and distances greater than 500m would limit speed. This meant our sites had no room for maneuver. Limited distance and a fixed service meant planning had very little flexibility to reposition a site.

In the second phase of the project we lost a lot of existing towerco sites due to inadequate coverage that we could get out of them. Our calculations were based on “Households” as a rule, amongst other calculations thrown in, such as competition and infrastructure costs, so covering partially an estate of 1,000 houses for example meant “GO/NOGO” decisions had to made per site. Out of the 300 sites activated we probably had a database of over 1,000 sites that were worked upon or looked at.

ON Telecom is one of several 4G players in Brazil, the biggest of which, Sky Brasil, uses TDD-LTE, achieved 100,000 customers in 2014 and is reportedly targeting 500,000 clients.

TowerXchange: How does network planning change in an environment like Brazil where 71% of the towers are owned and marketed by towercos?

In order to plan fixed cell site topography, we made extensive use of the databases from ANATEL, and asked the towercos for fresh databases every month. Planning a rollout where there are shared towers available means there are time to market and opex cost considerations as well as RF planning considerations – you have to take a more pragmatic view.

RF Planners and Engineers might moan about having to use sub-optimum sites, but you have to look at households first and RF planning second. Every site had to be evaluated in terms of the total business case; opex, capex, number of households divided by number of competitors et cetera – all these factors were evaluated in our GO/NOGO.

TowerXchange: Were there any noticeable differences between co-locating on an American Tower (ATC), SBA or Grupo TorreSur (GTS) site compared to co-locating on one of Brazil’s independent developers’ sites?

The big three towercos were often cheaper than the smaller towercos’ sites, but the smaller towercos were more hungry, flexible and proactive such as Brazil Tower Company (BTC) and QMC.

However, one of the larger towercos in Brazil were generally our preferred partners both because they were fantastic to work with, and because they had bought a lot of sites from Telefónica in São Paulo State where the land was still owned by Telefónica, so there was no ground rent pass through, which would reduce the cost of such sites by as much as half. Their sites were always given priority in the engineering planning just because of the saving on the Ground Rent, critical for a new operator of our size.

TowerXchange: How do the costs compare; leasing a ground based tower versus a rooftop versus build to suit?

2sqm on a ground based tower would cost us an average of R$2,000-2,500 (~US$700). Rooftop rents were higher, so the site cost around R$3,000-4,000 (~$1,000). About half of that was ground rent. If we built the rooftop ourselves then we would halve our opex costs.

If we commissioned a build to suit, 100% of the rental cost was borne by us until additional tenants were added. So build to suit is comparatively expensive, hence our preference to install our own rooftops when geographically possible. If you are in an area with no high buildings then you have no choice.

We would encounter situations where the big operators might be offering landlords double what we’d pay the landlord to use a rooftop, but because we moved so much faster, the landlords would rather take R$2,000 from us right away than wait for the big operators to make up their minds.

Outside Campinas, in the smaller municipalities, the rents would be lower.

TowerXchange: How does the build to suit market function in Brazil?

Once a towerco is registered and has built a relationship with the operator, they ask for a forecast of tower builds. The operator releases search rings and the build to suit companies go out with a nominal plan and come back with three candidate sites which might include existing structures and/or new builds. That’s the same as in any market, but in Brazil this can take ages as the towerco’s engineering guys don’t always have access to the engineering guys at the operator.

Licensing remains the biggest challenge to building new sites in Brazil. Once our RF guys qualify that we need to get a tower built at a given location, we go into pre-licensing, seek clearance of the aviation authorities, present the environmental issues, and only then begin the main licensing process. Often we would visit the local Mayors and emphasise our provision of Internet (rather than telecoms) services, often offering to provide free Internet to local schools. We’d take one of the smaller build to suit towercos with us like Brazil Tower Company or QMC, and we’d often get permission granted on that basis.

The larger towercos often didn’t want to do build to suit for ON Telecom; they felt we were too small.

At ON Telecom I did nearly all the acquisitions myself, our hunters came in and qualified sites, and the smaller, hungrier towercos often took it on from there. We did this to gain time. Also because of our limitations on specific coverage, focus was needed!

In a build to suit contract, it is important to clarify whether the anchor tenant pays fees from the moment the tower is built and ready for installation, or whether fees are payable from the moment the foundations are dug.

TowerXchange: TowerXchange concluded a study which suggested there may be less than 50,000 towers in Brazil, accounting for the towers of the towercos and the big operators, a significant difference from the 70,000 towers most people suggest there are in the country. Are we missing any?

There are a number of towers run by SCMs (licensed small operators) providing service where the big operators do not yet provide internet coverage. Even in the interior of São Paulo State you come across the antennas of a lot of ISPs with unlicensed frequency, providing low quality Internet service to maybe 2,000 customers in a small town. There are hundreds of them across the country and they account for a lot of sites. I’m not sure you should include such sites in your analysis though – it’s unlikely that many are shareable, and a proportion of them are probably unlicensed.

Brazil also has a number of utilities companies creating telecom spin-offs, some of which have microwave networks. For example, CPFL (energy company based out of Campinas) is providing fibre over power lines, and there are similar projects in the Northeast.

The other distorting effect on the tower count in Brazil is that when a towerco has built a tower, it’s registered to ANATEL by the tenants not the towerco, so the same tower can be registered multiple times by co-locating tenants.

TowerXchange: Help us understand the demographics of the 5,570 “cities” (municipalities) ANATEL the Brazilian regulator has targeted for coverage.

ANATEL sets aggressive coverage targets, but the reality is that, with the cost of deployment, the operators will do well to close the gap by 50%, and they seem to get fined every year.

Let me give you an example. One of the cities in ON Telecom’s footprint was Campinas, a municipality in São Paulo State. In a tier two city like Campinas, the core of a metropolis with a population of 2-2.5mn, residents’ have a healthy disposable income to spend on mobile and internet services, whereas many of the uncovered cities on ANATEL’s targets are in the North and Northeast of Brazil where the demographics become less attractive for operators and by extension for towercos. It’s the same in Australia, where I also worked for a couple of years – the government tries to force rural coverage through obligations in the license, but the economics do not add up. Deploying fibre is not cheap, and the alternative of coverage plus microwave towers has inherent opex and capex challenges. The challenge to connect Brazil’s unserved and underserved cities is as much to do with transmission as passive infrastructure.

TowerXchange: How complete is the fibre rollout in Brazil in terms of the proportion of towers still relying on microwave backhaul?

Outside of ‘The Triangle’ (between Rio de Janeiro, Belo Horizonte and São Paulo) where there are good fibre networks, the majority of Brazil’s towers are still relying on microwave or wireless transmission.

At ON Telecom, if a site didn’t have microwave backhaul that we could install it was automatically rejected. Also we left all our towers infra fibre-ready, with a post at the entrance with ducting to get to the Node Bs. This was done as part of the scope by the towercos regardless if we were ever to get fibre to the site.

A few of the tier two operators use fibre or have their own fibre, but the larger operators focus on their core business or on supporting their backbone rather than backhaul. Sharing someone else’s fibre is expensive. Deploying Leased Lines is still costly, these days needing higher link capacities. Values are coming down, but they still have a way to go.

A lot of the fibre in Brazil is deployed “over air” – mounted on posts on the street, which is relatively cheap and fast to deploy. In most cities you see spaghetti junctions of cables on posts, some legal some illegal. However, the problem is that this means it is incredibly expensive from an O&M point of view, and it’s exposed to hazards like sugar cane fires, bush fires and lightning storms. Putting fibre underground may be safer and more reliable, but it’s too expensive and slow when it comes to planning permissions and securing concessions from all the different parties. In this context, microwave still has a lot going for it.

4G will put more strain on transmission networks. Fibre is still backbone oriented in Brazil, seldom extending the last mile using microwave, especially in the North and Northeast of the country.

I’m now leading a startup which will be called Arqueiro Telecom which will provide high capacity (1.2-1.4GB) last mile microwave links. We have intelligent data that helps us understand when operators are in trouble when it comes to capacity and reliability, and we can help them breathe a little bit.

TowerXchange: How would you characterise the maturity of heterogeneous networks in Brazil? Do you see a lot of demand for metro cells and DAS?

In general almost everything is deployed on macro towers in Brazil. But with demand for data rising, and with data relying on speed, all the operators are asking towercos for metro cells and even femtocells.

At ON Telecom we deployed a couple of small cells from Airspan to extend the RF signal in specific apartment blocks.

A heterogeneous layer will be added to Brazil’s macro network over the next two to ten years, and the towercos need to help the operators to think differently. However at the moment it seems like the big towercos are focused on integrating their acquisitions and maximising tower cash flow from their huge inventory of macro sites – the largest of them must be billing over US$100mn per calendar month now. As long as ATC, SBA and GTS remain focused on macro sites, the smaller towercos will continue to think that the best way to create capital value is through building big ground based towers.

So I think there’s a gap in the market for someone focused on Brazilian operators’ increasing need for small cells and DAS, non-line of site transmission, and point to multi-point systems, which have been used in Europe for years but have seen slower uptake in Brazil.

Using DAS as an example, there might be 30 multi-band distributed antenna systems in various Brazilian shopping malls, but that represents a fraction of the addressable market for DAS. For example, there are huge condominium and housing estates in the interior of São Paulo State, gated developments hiding mansions where cabling fibre or even copper will be too expensive to achieve return on investment. The only way to provide internet service is through 3G modems – but how do you get signal in there? You can’t build a big macro tower next to US$2mn mansions. This represents a gap in the market for someone who is prepared to think like an engineering company, building an RF plan and solutions, put small cells on lamp posts or in unseen places (mini stealth sites), “lite” camouflage sites and DAS on clock towers……this can be endless. You should be thinking of hiring an architect with creative civil works involved and not just good old concrete and steel methods.

Taking this to an operator instead of your customary three candidates for a macro site is where I see some of the future. You make your money on volumes of sites/solutions but also you will have a happier customer and guaranteed growth.

TowerXchange: Please sum up how you see the future of Brazil’s tower networks.

Everything is unstable in Brazil at the moment – the government, the economy – but the tower industry is thriving and growing.

Amid such growth, it’s difficult for the towercos to stop and think about what the operators might need in two years time. The operators’ engineering departments are saturated and over-worked. Within niche markets, a smaller company offering intelligent solutions for street furniture and IBS could overtake the big incumbent leaders with their huge base of macro sites. It’s already time to start thinking about deepening infrastructure to the next level of providing a tower plus equipment.

There is always a risk when big, established market leaders sit back on their laurels – Nortel Networks have gone, Nokia are all but gone – innovators have got to think about smaller, lower structures. With the spectrum frequencies and wireless technologies today, sometimes mounting a little antenna on the side of a shop is more attractive than wading through red tape to build another macro tower.

Brazilians want internet access, and if they can get it they’ll use it. It’s just a question of how best to get it to them.