Failing Ahead – What We Discovered at Cisco from a “Failed” Digital Orchestration Pilot



The fashionable buyer expertise is fraught with friction:

You converse to a buyer consultant, and so they inform you one factor.

You log into your digital account and see one other.

You obtain an electronic mail from the identical firm that tells an fully totally different story.

At Cisco, we’ve got been working to establish these friction factors and evaluating how we are able to orchestrate a extra seamless expertise—reworking the shopper, companion, and vendor expertise to be prescriptive, useful – and, most significantly, easy. This isn’t a straightforward activity when working within the complexity of environments, applied sciences, and shopper areas that Cisco does enterprise in, however it isn’t insurmountable.

We simply closed out a year-long pilot of an industry-leading orchestration vendor, and by all measures – it failed. In The Lean Startup Eric Ries writes, “when you can not fail, you can’t be taught.” I totally subscribe to this angle. If you’re not keen to experiment, to strive, to fail, and to judge your learnings, you solely repeat what you realize. You don’t develop. You don’t innovate. It’s good to be keen to dare to fail, and when you do, to attempt to fail ahead.

So, whereas we didn’t renew the contract, we did proceed down our orchestration journey geared up with a 12 months’s price of learnings and newly refined course on the best way to deal with our initiatives.

Our Digital Orchestration Objectives

We began our pilot with 4 key orchestration use instances:

  1. Seamlessly join prescriptive actions throughout channels to our sellers, companions, and prospects.
  2. Pause and resume a digital electronic mail journey based mostly on triggers from different channels.
  3. Join analytics throughout the multichannel buyer journey.
  4. Simply combine knowledge science to department and personalize the shopper journey.

Let’s dive a bit deeper into every. We’ll have a look at the use case, the challenges we encountered, and the steps ahead we’re taking.

Use Case #1: Seamlessly join prescriptive actions throughout channels to our sellers, companions, and prospects.

Right now we course of and ship business-defined prescriptive actions to our buyer success representatives and companions when we’ve got digitally recognized adoption obstacles in our buyer’s deployment and utilization of our SaaS merchandise.

In our legacy state, we had been executing a collection of advanced SQL queries in Salesforce Advertising and marketing Cloud’s Automation Studio to affix a number of knowledge units and output the particular actions a buyer wants. Then, utilizing Advertising and marketing Cloud Join, we wrote the output to the activity object in Salesforce CRM to generate actions in a buyer success agent’s queue. After this motion is written to the duty object, we picked up the log in Snowflake, utilized further filtering logic and wrote actions to our Cisco companion portal – Lifecycle Benefit, which is hosted on AWS.

There are a number of key points with this workflow:

  • Salesforce Advertising and marketing Cloud just isn’t meant for use as an ETL platform; we had been already encountering outing points.
  • The companion actions had been depending on the vendor processing, so it launched complexity if we ever wished to pause one workflow whereas sustaining the opposite.
  • The event course of was advanced, and it was tough to introduce new really helpful actions or to layer on further channels.
  • There was no suggestions loop between channels, so it was not potential for a buyer success consultant to see if a companion had taken motion or not, and vice versa.

Thus, we introduced in an orchestration platform – a spot the place we are able to join a number of knowledge sources by way of APIs, centralize processing logic, and write the output to activation channels. Fairly shortly in our implementation, although, we encountered challenges with the orchestration platform.

The Challenges

  • The complexity of the joins in our queries couldn’t be supported by the orchestration platform, so we needed to preprocess the actions earlier than they entered the platform after which they may very well be routed to their respective activation channels. This was our first pivot. In our technical evaluation of the platform, the seller assured us that our queries may very well be supported within the platform, however in precise apply, that proved woefully inaccurate. So, we migrated probably the most advanced processing to Google Cloud Platform (GCP) and solely left easy logic within the orchestration platform to establish which motion a buyer required and write that to the right activation channel.
  • The consumer interface abstracted elements of the code creating dependencies on an exterior vendor. We spent appreciable time attempting to decipher what went flawed by way of trial and error with out entry to correct logs.
  • The connectors had been extremely particular and required vendor assist to setup, modify, and troubleshoot.

Our Subsequent Step Ahead

These three challenges compelled us to assume otherwise. Our purpose was to centralize processing logic and connect with knowledge sources in addition to activation channels. We had been already leveraging GCP for preprocessing, so we migrated the rest of the queries to GCP. As a way to clear up for our have to handle APIs to allow knowledge consumption and channel activation, we turned to Mulesoft. The mixture of GCP and Mulesoft helped us obtain our first orchestration purpose whereas giving us full visibility to the end-to-end course of for implementation and assist.

Orchestration Architecture
Orchestration Structure

Use Case #2:  Pause and resume a digital electronic mail journey based mostly on triggers from different channels.

We centered on trying to pause an electronic mail journey in a Advertising and marketing Automation Platform (Salesforce Advertising and marketing Cloud or Eloqua) if a buyer had a mid-to-high severity Technical Help Heart (TAC) Case open for that product.

Once more, we set out to do that utilizing the orchestration platform. On this situation, we would have liked to pause a number of digital journeys from a single set of processing logic within the platform.

The Problem

We did decide that we may ship the pause/resume set off from the orchestration platform, but it surely required organising a one-to-one match of journey canvases within the orchestration platform to journeys that we’d wish to pause within the advertising and marketing automation platform. Using the orchestration platform truly launched extra complexity to the workflow than managing ourselves.

Our Subsequent Step Ahead

Once more, we appeared on the identified problem and the instruments in our toolbox. We decided that if we arrange the processing logic in GCP, we may consider all journeys from a single question and ship the pause set off to all related canvases within the advertising and marketing automation platform – a way more scalable construction to assist.


One other strike in opposition to the platform, however one other victory in forcing a brand new mind-set about an issue and discovering an answer we may assist with our current tech stack. We additionally count on the methodology we established to be leveraged for different sorts of decisioning resembling journey prioritization, journey acceleration, or pausing a journey when an adoption barrier is recognized and a really helpful motion intervention is initiated.

Use Case #3: Join analytics throughout the multichannel buyer journey.

We execute journeys throughout a number of channels. For example, we could ship a renewal notification electronic mail collection, present a customized renewal banner on for customers of that firm with an upcoming renewal, and allow a self-service renewal course of on We acquire and analyze metrics for every channel, however it’s tough to indicate how a buyer or account interacted with every digital entity throughout their total expertise.

Orchestration platforms provide analytics views that show Sankey diagrams so journey strategists can visually overview how prospects interact throughout channels to judge drop off factors or significantly crucial engagements for optimization alternatives.

Sankey Diagram Sample
Pattern of a Sankey Diagram

The Problem

  • As we set out to do that, we discovered the biggest blocker to unifying this knowledge just isn’t actually a problem an orchestration platform innately solves simply by way of executing the campaigns by way of their platform. The biggest blocker is that every channel makes use of totally different identifiers for the shopper. E-mail journeys use electronic mail deal with, internet personalization makes use of cookies related at an account stage, and the e-commerce expertise makes use of consumer ID login. The foundation of this difficulty is the shortage of a singular identifier that may be threaded throughout channels.
  • Moreover, we found that our analytics and metrics staff had current gaps in attribution reporting for websites behind SSO login, resembling
  • Lastly, since many groups at Cisco are driving internet site visitors to, we noticed a big inconsistency with how totally different groups had been tagging (and never tagging) their respective internet campaigns. To have the ability to obtain a real view of the shopper journey finish to finish, we would want to undertake a standard language for tagging and monitoring our campaigns throughout enterprise models at Cisco.

Our Subsequent Step Ahead

Our staff started the method to undertake the identical tagging and monitoring hierarchy and system that our advertising and marketing group makes use of for his or her campaigns. It will enable our groups to bridge the hole between a buyer’s pre-purchase and post-purchase journeys at Cisco—enabling a extra cohesive buyer expertise.

Subsequent, we would have liked to deal with the information threading. Right here we recognized what mapping tables existed (and the place) to have the ability to map totally different marketing campaign knowledge to a single knowledge hierarchy. For this specific instance for renewals, we would have liked to deal with three totally different knowledge hierarchies:

  1. Get together ID related to a singular bodily location for a buyer who has bought from Cisco
  2. Net cookie ID
  3. Cisco login ID
Data Mapping Example
Information mapping train for Buyer Journey Analytics

With the introduction of constant, cross Cisco-BU monitoring IDs in our internet knowledge, we’ll map a Cisco login ID again to an online cookie ID to fill in a few of the internet attribution gaps we see on websites like after a consumer logs in with SSO.

As soon as we had established that stage of information threading, we may develop our personal Sankey diagrams utilizing our current Tableau platform for Buyer Journey Analytics. Moreover, leveraging our current tech stack helps restrict the variety of reporting platforms used to make sure higher metrics consistency and simpler upkeep.

Use Case #4: Simply combine knowledge science to department and personalize the shopper journey.

We wished to discover how we are able to take the output of a knowledge science mannequin and pivot a journey to supply a extra customized, guided expertise for that buyer. For example, let’s have a look at our buyer’s renewal journey. Right now, they obtain a four-touchpoint journey reminding them to resume. Clients can even open a chat or have a consultant name or electronic mail them for extra assist. Finally, the journey is similar for a buyer no matter their probability to resume. We have now, nonetheless, a churn danger mannequin that may very well be leveraged to switch the expertise based mostly on excessive, medium, or low danger of churn.

So, if a buyer with an upcoming renewal had a excessive danger of churn, we may set off a prescriptive motion to escalate to a human for engagement, and we may additionally personalize the e-mail with a extra pressing message for that consumer. Whereas a buyer with a low danger for churn may have an upsell alternative weaved into their notification or we may route the low-risk prospects into advocacy campaigns.

The objectives of this use case had been primarily:

  1. Leverage the output of a knowledge science mannequin to personalize the shopper’s expertise
  2. Pivot experiences from digital to human escalation based mostly on knowledge triggers.
  3. Present context to assist buyer brokers perceive the chance and higher interact the shopper to drive the renewal.

The Problem

This was truly a slightly pure match for an orchestration platform. The problem we entered right here was the information refresh timing. We would have liked to refresh the renewals knowledge to be processed by the churn danger mannequin and align that with the timing of the triggered electronic mail journeys. Our renewals knowledge was refreshed in the beginning of each month, however we maintain our sends till the tip of the month to permit our companions a while to overview and modify their prospects’ knowledge previous to sending. Our orchestration platform would solely course of new, incremental knowledge and overwrite based mostly on a pre-identified main key (this allowed for higher system processing to not simply overwrite all knowledge with each refresh).

To get round this difficulty, our vendor would create a model new view of the desk previous to our triggered ship so that each one knowledge was newly processed (not simply any new or up to date information). Not solely did this create a vendor dependency for our journeys, but it surely additionally launched potential high quality assurance points by requiring a pre-launch replace of our knowledge desk sources for our manufacturing journeys.

Our Subsequent Step Ahead

One query we saved asking ourselves as we struggled to make this use case work with the orchestration platform—had been we overcomplicating issues? The 2 orchestration platform outputs of our attrition mannequin use case had been to:

  1. Customise the journey content material for a consumer relying on their danger of attrition.
  2. Create a human touchpoint in our digital renewal journey for these with a excessive attrition danger.

For primary, we may truly obtain that utilizing dynamic content material modules inside SalesForce Advertising and marketing Cloud if we merely added a “danger of attrition” area to our renewals knowledge extension and created dynamic content material modules for low, medium, and excessive danger of attrition values. Performed!

For quantity two, doesn’t that sound form of acquainted? It ought to! It’s the identical downside we wished to unravel in our first use case for prescriptive calls to motion. As a result of we already labored to create a brand new structure for scaling our really helpful actions throughout a number of channels and audiences, we may work so as to add a department for an “attrition danger” alert to be despatched to our Cisco Renewals Managers and companions based mostly on our knowledge science mannequin. A suggestions loop may even be added to gather knowledge on why a buyer could not select to resume after this human connection is made.

Failing Forward

Discovering Success

On the finish of our one-year pilot, we had been compelled to consider the techniques to realize our objectives very otherwise. Sure, we had deemed the pilot a failure – however how can we fail ahead? As we encountered every problem, we took a step again and evaluated what we discovered and the way we may use that to realize our objectives.

Finally, we found out new methods to leverage our current programs to not solely obtain our core objectives but additionally allow us to have end-to -end visibility of our code so we are able to arrange the processing, refreshes, and connections precisely how our enterprise requires.

Now – we’re making use of every of those learnings.  We’re rolling out our core use instances as capabilities in our current structure, constructing an orchestration stock that may be leveraged throughout the corporate – a large step in direction of success for us and for our prospects’ expertise.  The end result was not what we anticipated, however every step of the method helped propel us towards the fitting options.