ASOS looking closer into its processes

10 top tips for implementing Process Mining

As ASOS continues to expand its Intelligent Automation capabilities, the second instalment in our Automation blog series focuses on Process Mining.

Process Mining provides a near real-time x-ray into how a business process is actually running based on data. Its visualisation dashboard can be interacted with to monitor KPIs, find inefficiencies, analyse root causes and even act on the insight.

Here are my top tips to consider:

#1 Data source health check

Perform a proactive health check against each of your most likely target systems and agree a data connectivity design upfront. Consider:

• Process mining requires a minimum set of event data from each data source (Unique ID, Activity description and Timestamp). Check that this data is available, and with sufficient history.
• If no event history exists, estimate the average cycle time of a process and multiply 10–20 times to determine how long to wait before enough history is collected to make process mining worthwhile!
• Your data architecture design may mandate a staging area (e.g. Data Lake) between source enterprise systems & the 3rd party process mining solution. Check whether this already contains the data that the use cases will require.

#2 Delivery Critical Success Factors

At its simplest, each process mining implementation will set KPI’s, connect to data and analyse the process.

The steps to an ASOS Process Mining delivery

The following Critical Success Factors are recommended to set up for success:

• Secure time from experts before starting: Process Owner, Business SME, Data Owners and local Tech teams
• Design your data pipeline with Data Architects & Cyber Security
• Use the product vendor to assure your implementation project
• Invest in change management to communicate regularly, ensure adoption and sustain the new way of working

A motivated user creates a win-win situation, you get great engagement and time investment helping project success while the stakeholder learns a new skill, obtains great insight to unseen business data and can help lead the line to improve company performance!

#3 Set and focus on Key Performance Indicators (KPI’s)

Work with business stakeholders up front to understand the primary KPI’s you want to move the dial on, then stay laser focussed on them.

• Once you have the KPI’s, you can then identify the data sources you will need to collect, or use to calculate, the required data dimensions. Progress with agility if not all data is immediately available.
• Add cost and sustainability data into the KPI construction, this will complement the value story.
• Mining exposes lots of variants, we’ve seen over 10,000 in one process! Whilst this makes for an impressive demo or shock tactic, remain focused on the primary value KPI’s. Don’t be swayed by outliers, and only pivot by exception.

#4 Make the tool easy to adopt

The quantity of insight can be overwhelming, so focus on the user experience and be aware of other day-day inputs & dashboards the stakeholder may already use.

Predominantly the mining tool is engaged with in two ways:

1. Exploratory investigation
2. To confirm a hypothesis

Build for the layman and make it easy to use. Tips to improve the experience:

• Check the activity descriptions makes business sense. Create a mapping translation if not
• Add onscreen guides to help each user role navigate the tool to achieve their goals
• Provide a pop-up glossary of definitions
• Iteratively develop the dashboards with the target users to consider their needs and encourage engagement

ASOS spotlight on a process

#5 Take action!

As you engage more with process mining, you will identify more execution gaps and potential value to chase after, but don’t constrain process mining to just be ‘the insights project before the improvement project’ — use its automation capabilities to actually improve execution management.

Initially invest time to streamline and eliminate waste from your processes first, so you don’t automate a bad process and achieve a bad result quickly!

Spark your users’ imagination with some real-world automation examples e.g. notifications triggered by SLA breaches through to automated corrective actions taken directly within your enterprise systems (as allowed by their API’s).

Just take care with any time lag (from source to target) in your dataset that you may be using to make decisions to execute against.

#6 Widen the process net

Disclaimer: I haven’t done this one yet but think it is a good idea :-)

Part of your end-to-end process may fall outside the boundary of your companies’ responsibility, so there may be a lack of transparency with your supplier performance other than end outcome.

However, why not check to see if they are also invested in Process Mining and explore the possibility of linking your two solutions together in order to share amazing insight and work in partnership to achieve process excellence and eliminate waste?

Company cultures and commercial arrangements may well restrict the viability of this, but if enabled offers a true end-end view.

#7 Be mindful of hidden costs

Process mining replicates data, usually in a 3rd party cloud solution, your data pipeline may dictate several hops from source to target destination as well. The data will also grow in time through delta increments.

There is an infrastructure and storage cost associated with each hop. Tips to mitigate cost include:

• Ensure only the minimum amount of data is moved between hops
• Truncate data where possible
• Review your history requirements (e.g. to measure KPI trends over time or assist analysis of variants with long cycle times)

The ROI may be so astonishing this is a small fry consideration, but if you look after the pennies…

#8 Don’t assume the data pipeline just works

The quality of the data is vital to the quality of the process mining insight, and the users need to be able to trust it because they may act on the insight.

Ensure there is an element of reconciliation between every data hop, so you can be confident the data that left ‘a’ arrived at ‘b’ in tact and as expected.

#9 Sustain the change, support the sponsor & expand!

Process mining is so much more than a technical project. It enables achieving process excellence and highlights the execution gaps stopping you from achieving the maximum performance of a process.

A project will tangibly deliver a dashboard and an initial list of framed value opportunities, useful, but alone will not sustain the change as you transition to BAU.

For your sponsor, hold regular sessions to:

• Demonstrate the progress & value achieved
• Share the data and user feedback to help them evangelise amongst their peer group
• Review expansion opportunities aligned to strategy
• Regularly assess that the operating model matches the pace and investment you want to work at

For each process, establish a cadence to stay focussed on value and encourage the proactive use of the insight by:

• Weekly team review, Monthly SteerCo & Quarterly value assessment

Or

• Slot a ‘value’ agenda item into established governance; this will also help alignment and empowerment with a broad set of stakeholders.

Finally, help and inspire the team by utilising independent expert advice against your dataset to provide more insight you may have overlooked. Product vendor value architects is a good place to start for this service as well as external communities of practice.

#10 Process Mapping

Using mining to check a processes conformance against a standard and finding violations to inform your data driven investigations is a very powerful feature, and one of the first areas you would start to look for clues for inefficiencies, non-compliance and bottlenecks.

To instigate conformance, you need a process map; options to source one range from:

1. Use an existing one (if lucky enough to have one!)
2. Manually create a process map — which may involve lots of people, workshops and time
3. Use process mining to auto generate a process map

Our initial preference has been to use a Subject Matter Expert or Business Analyst to manually map the happy path process, mastered in a Business Process Management tool. Then import it into the mining tool to start conformance checking.

We have then used the violations highlighted and also the auto generated process mining map to help provide insight into the completeness and accuracy of the process map, iterating accordingly.

Interested in reading more? Automation blog series first instalment: “Top tips for implementing RPA”

ASOS Digital Workers

ASOS are hiring across a range of roles in Tech. See all our open positions

Scott Mathers leads the Intelligent Automation team at ASOS. In his spare time he enjoys keeping fit, juggling a 2nd career as a Dad taxi and following SpaceX!