The Million Dollar Message

In enterprise, customer experience issues are seldom neat and tidy. Work is always backed up, and teams in a siloed environment aren’t always able to see how things connect, and prioritisation across teams can be hard to coordinate. This story shows how you can take vague organisational objectives like ‘increase sales’ and ‘improve customer retention’, and use basic techniques like journey mapping, surveys, and documentation to understand your existing processes better, which enables you to discover opportunities. By starting with the basics, you can resist the temptation to solution to your competitors, and discover enhancements you can make that don’t add to your burdens.

This case study will take you through a deep dive of customer behaviour around order fulfilment, and shows you how in large chaotic environments sometimes successful CX research outcomes are choosing *not* to build something, or discovering and busting assumptions that will pay dividends to the team in the future.

Problem Statement

A review of call center statistics showed a disproportionate amounts of customer queries asking where their order is. A review of the case notes for each call showed that some orders were late, but not to the extent we expected for the call volume.

Background

Orders were delivered to homes by a partner with a fleet of contractor-drivers who collect the order from the store and deliver it to the customer’s home. Customer enquiries about order status involve both companies, leading to long wait times while investigations happen.

We had a hunch based on competitive analysis that customers would defer calling the support center if they could find out the status of their order another way, but we wanted to explore the problem before honing in on a solution.

If the order is late past a certain threshold, it is practice to offer credits to the customer by way of apology, and orders cancelled before delivery can’t be returned to the store and end up as waste. The support agent’s potential time saving also adds up to tens of thousands of dollars per quarter.

Investigatory research

Documenting the journey

We were surprised to learn that there was no design resource that explained what the interactions and systems actually were – the constant pace of change, siloed teams, partner API design cycles had all built up to a bit of a mess. As Fulfilment as a ‘product’ didn’t have any screens at the time – it was the umbrella for activities that occurred after checkout ending in either a successful receipt of goods, or an unhappy path – our main touchpoint was transactional messaging in the form of SMS and email. 

We began the process of creating journey maps and service blueprints that showed screenshots of the app state, transactional messages sent via email and SMS, sending systems and triggers, and API events. We read API documents, trawled the ocean of information in Confluence, spoke to at least 10 tech leads, and placed own grocery orders to capture real life interactions. These would become lasting artefacts published across the company, and valuable references we refer to in future research expeditions. We also captured the transactional messages of our top 2 competitors, and gave them the same treatment.

Competitor analysis

We looked at Walmart, Target, Shipt, Amazon, Instacart, and Uber Eats. Walmart and Target use partner services to deliver orders similar to our Retailer. Shipt is owned by Target, and is their delivery partner – but also competes with Instacart as a grocery aggregator offering multiple brands in one experience providing interesting contrast of one company with two journeys. Amazon, Instacart, and Uber use their own labor and tech stack rather than having a partnership with an API.

Strategy 1: Create visibility with driver tracking

Looking at the top 5 grocers in the USA, only Amazon and Instacart offer GPS map views of their drivers, and only Instacart approaches the fidelity of the interfaces popularised by Uber Eats and other real-time delivery dashboards. While an appealing user interface (it was the idea on the tip of everyone’s tongue we spoke to), it is a complex system to implement and costly to run requiring periodic API calls, a data connection to make them with, and a solid GPS signal – which is not a reliable thing in dense cities. The last thing we want to do is go to the trouble of building a map screen, and have inaccurate or slow data generate more customer concern. Understanding this, maps were off the table for a first iteration until their value could be quantified against other options.

Strategy 2: Empowering the drivers

Shipt stuck out as having almost no fulfilment experience in the app – just transactional messages, and didn’t even appear to send standardised messages – drivers were typing updates and sending them as they saw fit. Uber Eats augments their push and SMS messaging capability with direct 2-way communication. This is an interesting strategy – experienced drivers are best positioned to know what’s up, and a human touch can provide surprise and delight. It can also be an inconsistent disaster, and creates a compliance and standards problem along with a privacy headache. You are effectively making these drivers customer service ambassadors for your brand and part of your organisational voice – be careful what you wish for.

Strategy 3: Sequencing transactional messages

Mapping competitor messages out along a timeline, we could see that everyone approached their sequencing differently.  Some had more trigger points and milestones than others, some had conditional logic. Instacart, being a digitally native company had invested in a rich dashboard that became the focal point of the app while an order was being fulfilled.

We surveyed some of our customers to get some baseline sentiment on expectations around messaging channel. We had earlier research that indicated people preferred SMS to email, and we saw the same result again. In a force rank exercise, 69% wanted SMS for the primary communications, 25% wanted email, and the remainder wanted Push. I was surprised at how low the demand for push notification was, but this reflects the technology adoption curve for the industry and the fact that SMS just works. It doesn’t make sense to over index on product innovation unless it’s an innovation play – otherwise, enhance the existing technology first.

We took the opportunity to gather quick data on which timeline events customers preferred to receive SMS for, and were pleased to see it lined up with our current implementation. Customers want to know when their order has left the store (‘I need to remember to be ready’), and when the order is imminent (‘I need to be ready soon’). The majority were less interested in milestones like ‘order staged’ that relate to the back of house process – only when it affects the choices they are able to make in the app, like substitutions.

Customers expressed a desire to know about late and early orders. Early orders took us somewhat by surprise, but the impact to customer’s lives is just as disruptive as late orders – the social contract is broken, and while early orders have plausible business reasons behind them, customers quite rightly become incensed at the idea of missing a delivery that occurs before the agreed upon window. 

Expectations are critical

Consider the following scenarios and how they make you feel. You have invited a friend over for dinner at 6pm:

1. Your friend arrives between 5:55-6:05, or something close to that.
2. Your friend arrives at 5:00 with no warning. You are in the middle of finishing your dish to get it in to the oven, and now that task is competing with welcoming your friend.
3. Your friend arrives at 7:00 after texting you at 6:30 to say they’re stuck in traffic. Dinner is going cold on the table. When they arrive, they apologise and say they were delayed leaving because work ran over.

Now consider how your opinion changes if you know the following:

• In the early scenario, your friend left a bit early to beat the traffic, coming straight from the office.
• In the late scenario, your friend knew there was a big meeting scheduled at the end of day, and traffic is usually really bad at rush hour.

Everybody would prefer things just ran on time, but your emotional response changes based on the expectations set through communicating, and when the knowledge became available. If you feel like your friend knew information you could have used to better manage your time, then you’re going to feel negative if they didn’t bother to convey it, or that they made arbitrary decisions about their timeline without considering yours. 

If your friend does this to you every time you have them over, eventually you’ll stop inviting your friend – but consider how empathy and communication can alter both the impact of delay as well as your emotional response to it.

Expectation-setting in fulfilment customer experiences is similar. Being on-time is best, but allowing the customer to have full situational awareness and reasonable expectations is a good baseline. Repeat customers will learn your fulfilment sequence and start to notice patterns of poor performance.

Our key discovery on this theme was that our message sequence didn’t warn customers an order was running behind. Customers would see a message saying it was with a driver, but wouldn’t hear from us again until we were 5 minutes away. This left a span of potentially hours of no communication.

Using maps to identify opportunities

Our journey maps also showed us patterns we weren’t explicitly looking for. When we started, the assumption was that most early & late order conditions arose after the order had left the store. Mapping the process starting all the way back at the ‘thinking about shopping’ allowed us to question different customer mindsets about when they schedule their order and break up their week, and understand where and how delays occur. 

From this, we noticed that there were inferrable conditions about orders that could be surfaced to the customer to allow them to make better use of their time, similar to the dinner scenario we covered above. 

• If you know at 9am that your delivery capacity is lower than orders booked, and there’s no spare, then you know you’re going to run behind. 
• If you know an order isn’t ready, but its pickup deadline is approaching, you know it will be delivered late.
• If an order has left the store, the delivery window is almost over, but the customer isn’t up next, then you know it’s going to be late

Sprint outcomes & summary

From this research we identified and published the following principles for effective transactional messaging:

1. Information about order timelines should be relayed as soon as you know it, providing it is actionable and accurate.
2. If you don’t have full situational awareness – don’t fudge it. Inaccurate information is worse than nothing at all.
3. Message events should be triggered off events that are relevant to the customer, not the business process.
4. Expectation-setting is critical in your customer communications. If you don’t give customers a clear idea of what process will happen, you set yourself up for disappointment. Customers will form their own expectations based on their past experience and your competititon.
5. Even though you’ve made the sale, satisfying fulfilment experiences can make or break a returning customer due to the peak-end rule.

In addition to our documentation, we made recommendations to create and prioritise work items:

• We could see from the customer research that there was room to improve simply by rewriting the SMS copy to be more direct and place the critical info at the start. Sprints began to test and deploy new language for SMS.
  • Removing just one unhelpful SMS from the sequence at that scale saved almost a million dollars a year
• We were able to deprioritize an expansion of push notifications and the need to build heavy dashboard UI, large efforts which wouldn’t have provided as much ROI

• Discovery sprints were undertaken to test approaches and build solutions for the following:
  • Offering customers a choice of rescheduling on orders that will be very late, or waiting for the late order
  • Using the event stream to send conditional messages to customers warning them their order is late
  • Surfacing order fulfilment status through the UI 

Business Outcomes

• After releasing early/late alerts, call center inquiries for ‘where is my order’ fell by 30%
• Corresponding customer goodwill credits fell by $80,000+/year
• Food waste due to cancelled orders fell by 20%
• Removing an unnecessary SMS saved approximately a projected $1million per year