Asparuh Koev is a successful serial entrepreneur. Currently, he is a CEO of Transmetrics, a solution for cargo transport companies that uses Big Data and predictive analytics. Asparuh holds a Bachelor’s Degree in Computer Science from the American University in Bulgaria and an MBA degree from the Vlerick Business School in Belgium.
“Big Data: Improving capacity utilization of transport companies” will explore the practical benefits, IT tools and challenges of implementing a big data solution in a traditional industry (cargo transport), using as a showcase a predictive analytics project done for Speedy.
2. Transmetrics – the company
This presentation
2
Convincing business to try big data
Legacy landscape challenges
Tool set challenges
3. Kilometers on road are
24% empty
“Full” trucks only carry
57% of capacity
EU CO2 reduction target:
124 MT / year
EUR cost reduction target:
11% cost reduction
Empty space in the cargo supply chain, according to the World
Economic Forum
3
4. Large transport companies have the most to gain
Large transport company P&L structure (typical, source: Roland Berger strategy
consultants)
Net
Revenue
85%
15% 7%
4%
Capacity
Cost
Gross
Profit
Direct
Costs
Indirect
Costs
4%
Net
Profit
-11% 2-3x
Modest decreases in capacity cost
leads to dramatic increases in profit
4
5. Typical situation in daily transport
Today, there is poor capacity utilization in the transport between terminals,
for terminal – based products (Groupage or LTL), both domestic and
international
5
Terminal
Customer
locations
Terminal
Customer
locations
6. Transmetrics provides the missing 80% of data by predicting future customer
transport bookings, based on the basis of historical data and demand variables.
This results in efficient capacity plans with less empty space in vehicles.
Shipping
History
3-5 years
Customer orders
Consolidations
Linehauls
Events
Carrier contracts
Customers
...
+
Shopping
days
Industrial
seasonality
Month
end
Fairs and
events
Customer
Forecasts
School
holidays
Public
Holidays
New
tenders
Competitor
events
Gained and
lost
customers
Network
plan
changes
New
Product
launches
Commo-
dity
prices
=
Efficient
Capacity Plan
Forecasted
Customer
bookings
7. Transmetrics enables the loading factor of each linehaul to be
forecasted a few days in advance
Forecast:
next
Wednesday
departure
Forecast:
next
Thursday
departure
Forecast:
next
Friday
departure
Unused
capacity
Likely to have too much
unused space:
action needed
Should be OK,
no need for action
Forecasted groupage
orders via data mining
Likely to be overloaded,
need to make
a contingency plan
8. VPN
Shipping history
Shipments, capacities,
contracts, events
Transmetrics
servers
Transmetrics
servers
Transmetrics
Cargo transport
predictive
optimization
product
: SaaS product with a daily usage scenario
Customer IT
systems
Customer IT
systems
Transport
Management
System
Transport
Capacity Planning
System
Reports
for users
Reports
for users
Forecasts
Optimized schedule
runs periodically
Cloud - SaaS
Subscription
€ 2,500 per country
per month
8
9. First commercial implementation: DPD network
First implementations
9
Implementations in discussion with
Live since October 2015
In progress (go-live Q2/2016)
Romania
10. Transmetrics – the company
This presentation
10
Convincing business to try big data
Legacy landscape challenges
Tool set challenges
11. What does it mean for technology
11
CEOs are willing
to pay for
solutions to
BUSINESS problems
We have to translate
The right business problem
To a BIG DATA solution
And convince them that it
will work
13. Started with a very well established idea
13
Idea of Transmetrics came out in 2012 …
We already had the:
… Know how of business problem
… Understanding of data
… Understanding of algorithms
… Contacts with potential customers
… Some funding
14. The problem in getting a big data project going
14
Organization is overloaded
with daily problems
Properties of big data ideas:
… The results are unsure
… No benefits in this quarter
… Not a “burning issue”
Normal answer is “… great
idea, but not this year”
15. Yet … a number of challenges that took 3 years
15
Get someone to
take the idea
seriously and give
feedback
Get someone to
give us data to
work with
(DHL Express
transferred data
from October 2013
to February 2014)
Get someone to
agree to
implement in
production and
pay for the
solution
Met with over 30
transport
companies
What mattered:
trusting us as
people
Out of the 30, 3
companies joined
What mattered:
visionary CEO
Out of the 30, 1
company joined
Key factor:
visionary line
manager
16. The winning hand
16
The key factor to convince
management: the BIG SIZE
of potential benefits
22. State of the technology at CARGO customers
22
Legacy – 1990s
Some BI / data warehouses for operational needs
No data mining capability
Hard to do large data extracts
23. Data sizes … 4 billion records, with 200 columns each
23
4 billion records with 100+ fields each
>is much more than>
4 billon key-value pairs
Filter by multiple columns
Joins
Etc.
24. Data flow from origin to big data system
24
Legacy
Legacy
Legacy
Legacy
offices
Legacy
Cenral
repository
Transmetrics
Staging
CSV
Transmetrics
DWH
VPN
access
Customer controlled Transmetrics controlled
$3@#!!!! Transmetrics doesn’t work!!!
Late data
Wrong data
System changes
25. The importance of
25
Data quality measurement system + sign off
Data tracing / audit logs
Automated monitoring
= Be ready to prove that “garbage in = garbage out”
= Push problem back to customer IT
26. Transmetrics – the company
This presentation
26
Convincing business to try big data
Legacy landscape challenges
Tool set challenges
27. Technology challenges for a big data startup
27
Open source /
community
Performance
Support
Security
Legacy/
big vendor
Cost
Privacy
Lock in
Other startups
Uncertain quality
Uncertain roadmap
Risk
28. Tool set
28
Big data repository: Mammoth DB
• Stores main transaction data
• Main analysis cube = 2.5 billion records
• Most queries take < 1 min
Integra-
tor
Server
(PC)
DB Server 1
DB Server 2
DB Server 3
DB Server 4
DB Server 5
...
One virtual database
Other tools used
29. Key mistakes to watch out for
29
Started with tools that don’t scale (Pentaho, web2py)
Didn’t invest in data quality framework until late
“It’s all about the data” … underestimated the UX
30. Thank you for your attention!
Transmetrics AD | Asparuh Koev, CEO | +359 888 400 348 | asparuh.koev@transmetrics.eu