TM Forum Community

 View Only
Expand all | Collapse all

CEM (Network Probe feed) vs CDR Analsyis - Why CEM

  • 1.  CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 08, 2019 09:25
    ​Good day all , I'm in the middle of an implementation of a CEM solution which the main data feed is probes on majority of interfaces.   Comparison of what unique insights does a CEM platform provide ,compared to a CDR Analysis is stalling the adoption of the insights for use-cases.

    I'm seeking for any experiences around this
    #CustomerExperience

    ------------------------------
    thembinkosi ndebele
    MTN Group Limited
    ------------------------------


  • 2.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 03:45
    Helle Thembinkosi,
    Sound like you problem network elements, is that right.
    If so I guess you ETL the various data formats and build data sets that will be correlated and fed to analytics solution.

    If that is right, and you are building use cases. 
    Base on my assumptions, My suspicion is that you are heavily NOC focused.

    Customer Experiences are drawn from instances way beyond the network. Network alone, in this world is a tunnelled vision of the customers' experiences.
    If you expand that to various BSS systems you potentially get a broader view, but still inside "your own world" view, and not an ecosystem view from which you are part of.

    ------------------------------
    Cato Rasmussen
    Independent
    ------------------------------



  • 3.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 04:22
    Your assumption is correct ,the early focus of the program is to look at experiences based on network usage, key data sets are ingested from BSS to enrich the data sets.  While the NOC is impacted by introducing 'customer awareness' capability , it is not the focus.

    Majority of use-cases are around cross and upsell based on predicted objective NPS and user behavior analysis.

    The friction is where the business is asking , what additional data we receiving from probing the network interfaces compared to analysing call data records.​

    ------------------------------
    thembinkosi ndebele
    MTN Group Limited
    ------------------------------



  • 4.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 10, 2019 09:01
    Thembinkosi,
    A heads up, if I may.
    When scope is widen beyond network probing, you will most probably find that your use cases no longer relevant. I my experience, and it has been the case time after time after time, use cases tend to be one directional and when going backwards major holes appears. This will just magnify when scope get broadened.

    Also, contrary to Customer Care and CRM that are internally success measured, CEM can only be measured by customer.

    I suggest you read intro to Behaviour Economics. Highly relevant to Customer Experience. 

    Finally, be aware that Customer experience goes way beyond utility Expereince, that part is only 10-20% of the whole. Telco tend to think of the utility part as 80-90%. That is also why use case approach tend to be of moderate value for Managing Customer Experiences.



    ------------------------------
    Cato Rasmussen
    Independent
    ------------------------------



  • 5.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 04:22
    Hi Thembinkosi,

    A few of the reasons why a probe-based solution allows for a more detailed CEM analysis and thus use cases I can think off is:
    • Probes will give you more information on the quality of the sessions, while CDRs are usually just descriptive that a session or call took place, its length and not much more.
    • Probes can get you closer to realtime use cases. With CDRs you are not getting such a direct feed from active sessions.
    Of course you need to get down to details of which probes you have to have a conclusive analysis. From my experience there are huge differences in what a probe provides depending on your license agreement and such. Devil is in the details as well in this topic.
    In any case, CDRs still provide a lot of good information it may be not cost-effective to get with probes. Voice calls info (not VoLTE), for instance, is already available for rating-charging, and it is typically not very effective to put probes for the voice call quality (maybe in an sampling fashion, but not for every call). I would keep CDRs for those.

    Hope it was of some help. Regards.

    ------------------------------
    David Espadas
    Ericsson Inc.
    ------------------------------



  • 6.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 04:50
    ​Hi Tembi,

    In many of the CDR focused  use cases ,Probes data can only enrich the insight .So it is better to avoid building duplicate use cases fed from separate data sources ,but the positive insight that Probes data can add to some of those use cases is vital for in time and business and operations  decision making .

    ------------------------------
    AYDA Honarmand
    MTN Group Limited
    ------------------------------



  • 7.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 09:21
    Edited by Alfonso Miranda Jan 09, 2019 12:34
    Thembinkosi

    It all comes down to the incremental value vs the incremental cost to be determined by use case been implemented, probes will give you a better more granular view of the network experience the customer was exposed to, a good example is a low resolution photo vs a high resolution one, if all you need is to identify if there is a car present probably the low resolution will do, if you need to be able to read the license plates then the high resolution becomes a requirement of the use case.

    We have used both CDRs and probe data for CEM use cases associated with holistic customer experience quantification, probes will give you a much more accurate mapping of locations and consumption volumes, even desegregated by the types of traffic, however, if you use long enough periods you can also obtain a fair approximation of routine mobility patters with CDRs which on aggregate are probably enough to drive capacity investments but not necessarily enough to customized retention campaigns.

    ------------------------------
    ------------------------------
    Alfonso Miranda
    Bell Labs Consulting
    Alfonso.Miranda@Bell-Labs-Consulting.com
    ------------------------------
    ------------------------------



  • 8.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    Posted Jan 09, 2019 10:11
    Hi,

    I believe that you need to establish the firt 3 or 4 use cases that will drive your business case. We have identified the following use cases as being most promising when using network-based customer experience KPIs:
    1- Reduce churn based on low-scoring KPIs customers
    2- Generate upsell by targeting spcecific user category usage profiles
    3- Reduce costs by identifying quickly low-performing devices
    Hope you succeed in your project!
    Happy New year!

    ------------------------------
    Charles Coutu
    Videotron G.P.
    ------------------------------



  • 9.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 09, 2019 15:02

    Probe data (xDR) contains the key information for subscriber experience while CDR contains charging and usage information. To build a good solution (use cases) for CEM, my suggestion based on my experience is to correlated data from at least xDR (from probe), CDR, and CRM (user profile). After the data are correlated, you will be able to:

    a. define and adjust measurements (index) formula for performance, quality, availability, reliability, usability, etc.

    b. define and adjust customer experience measurement formula

    c. define dimension for the CEI (customer experience index) so that you can do drill-down/roll-up/drill-cross/drill-through to find the BI in data

    d. define reports and dashboards for in-experienced person or person who does not have the necessary expertise to navigate the data to find the answers for questions such as:

        i. what customer experience is for overall, by region, by user category, by service bundle, by application category, by application, by APN, etc.

        ii. where the customer experience is not good

        iii. etc.

    If you also want to find out the answer for "why" customer experience is not good, you will then need to combine with OSS system such as iNMS and/or SMS (service management system).



    ------------------------------
    Fazhong Deng
    OSSEra, Inc.
    ------------------------------



  • 10.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 10, 2019 06:19
    ​100% agree with approach shared by Fazhong Deng .That's the optimal way .

    ------------------------------
    AYDA Honarmand
    MTN Group Limited
    ------------------------------



  • 11.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 14, 2019 10:46
    Fazhong

    Although I agree with the overall spirit of the proposed process I will like to add a couple of caveats:

    - The correlation of personally identifiable information with specific network consumption data might be subject to strict restrictions on some geographies (GDPR in Europe for example), there are anonymization techniques to ensure users privacy is protected but this alone will require its own subject of discussion.
    - There is not really a good approach to a "Customer Experience" formula, here is where Machine learning is a really great ally to build some understanding, with the consideration that network quality in general only explains between 5% to 35% of the expressed CSAT (by the way, Customer surveys are another vital data point to build understanding around customer experience not listed on the formula before), therefore, the more information that can be added from the customer journey, the better chance the ML models will have to extract meaningful and predictive relationships.
    - Finally, not all customer assess value the same way, therefore the model need to be build on top of an actionable segmentation around the customer base as the data volumes allow (most limiting factor tends to be the volume of surveys) to be conducive to significant and actionable results.

    Regards,

    ------------------------------
    ------------------------------
    Alfonso Miranda
    Bell Labs Consulting
    Alfonso.Miranda@Bell-Labs-Consulting.com
    ------------------------------
    ------------------------------



  • 12.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    Posted Jan 16, 2019 10:43
    Edited by Keith Milner Jan 17, 2019 05:30
    I have been involved in some implementations of probe-based CEM systems on TDM voice and VoIP networks in the past.

    The primary use case for these systems is in understanding the service quality being provided to the end-user. As others have stated, this is not entirely possible using CDRs and other non-probe data alone (note that most networks can emit additional statistical data beyond CDRs, including data on call completion rates, ASR, average PDD, etc.)

    There are two types of probe and, in many implementations, both types are used together:

    Signalling probes will capture and provide analytics for signalling data. This can be used for:
    • Operational route analysis - are there routes where (for instance) the ASR or PDD are higher
    • Fraud detection (CDRs can be used for this also, but signalling probes give more detail and can be real-time)
    • Data to support specific fault root-cause analysis (for instance, why is a particular class of calls failing, or calls on a particular route, etc.). This not only improves the customer experience, but can result in improved network efficiency and increased revenue as fewer call attempts fail or are abandoned
    Less commonly, voice quality probes can be used. These tend to be used with VoIP networks more than old TDM networks. These attempt to measure the voice quality by various methods, which can include capturing and analysing the media stream, and analysing the metadata in the stream protocols (typically RTP). These enable quality measurements to be made including measures of jitter and latency. In some cases it is also possible to obtain MOS and/or PESQ scores which give a very good indication of the voice quality experienced by the customer on a call-by-call basis. Combined with other metadata such as PDD, this can provide a very good measure of the total call quality for the customer, especially as it includes the impact of interconnected networks.
    This can be used for:
    • Analysis and tracking of call quality across the network and on specific routes
    • Fault-finding on specific calls, in support of customer tickets, etc.
    • Threshold based alarming for early detection of congestion, poor quality routes, etc.

    Note that most of the benefits and use-cases of these systems are directly operational: they support the ability of the Operations team to understand how the network (and the networks of Interconnect partners) is operating and to optimise the efficiency and quality of the service.

    However, there are some potential revenue benefits including increased revenue from having better call completion rate and fewer abandoned calls, as well as the reputational benefit of having a good quality network.

    I hope that is useful.

    Keith



    ------------------------------
    Keith Milner
    Superlative Solutions Ltd.
    ------------------------------



  • 13.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 17, 2019 08:41
    This is all well, but it is NOC not CEM. It may provide some information that can be used, BUT IT IS NOT CEM!

    ------------------------------
    Cato Rasmussen
    Independent
    ------------------------------



  • 14.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    TM Forum Member
    Posted Jan 17, 2019 09:45
    Thembinkosi,
    please consider this; If the aim of CEM is to change peoples perception of your company as a service provider -> do not start with data analytics. That will only give you a scientific collection of data which assumes and give you the illusion that the world of tomorrow's customers will be the same as it is today and was yesterday. I will not move the needle. You will operate in the paradigm of the past.

    You need a much more innovative approach, and that cannot be based on data analytics because it belongs in the part of the world that will change, and we have no view or data of it. 
    to what degree was Henry Ford reliant on past data (if I had asked people what they wanted, they would have replied: Faster horses.
    Steve jobs did not rely on data analytics when coming up with iPod, iPhone, MacBook and their ecosystem.

    Starting with probing and NOC oriented activities will insure large investment, minimal customer effect, most likely only have a bleep of positive NPS before realising the failed investment,

    ------------------------------
    Cato Rasmussen
    Independent
    ------------------------------



  • 15.  RE: CEM (Network Probe feed) vs CDR Analsyis - Why CEM

    Posted May 02, 2019 08:35

    Hi,
    Passive probes on all interfaces will give you deep insights in the QoS and network, in addition you will have data on subscriber level 24/7/365 in more or less real-time. The data is much richer in content than billing CDRs and TAP files

    One issue you are facing is that most vendors of CEM type of systems look upon CEM from a network perspective and not from a customer perspective and they are designed for persons with quite deep technical knowledge.

    A CEM typ of system with data from passive probes should give you information about QoS related issues tied to a single subscriber and all other data you can pick up from the network.

    A system based on CDRs are less rich in content and will limit your QoS view of a subscriber.

    Take a look at www.subtonomy.com we have use cases out of the box in our applications that may give you some inspiration. However, we try to build applications that are used in different departments like customer care, NOC/SOC etc. We don't provide a frame work were you build your own use cases, in our case we deliver them to you out of the box.

    We are also mixing different types of information in our solutions, e.g. if you don't have passive probes on the data interfaces we can then use CDRs and also complement with speed test information.

    If you are in Nice ping me and we can share our knowledge in this field.

    Regards Fredrik



    ------------------------------
    Fredrik Edwall
    TO BE VERIFIED
    ------------------------------