Branded Content

Who Owns the Delay Between Data and a Decision?

A pipeline can meet its delivery target while a customer-facing system still acts on an outdated record. CDOs need a measure of success that follows the data all the way to its use.

Written by: Mike Anderson | CTO & Founder at Tealium

Updated 8:00 AM EST, November 18, 2026

post detail image
Mike Anderson | CTO & Founder at Tealium Mike Anderson founded Tealium in 2008 and has served as the chief architect behind the company's market-leading tag management and unified marketing platforms.

A warehouse can refresh on schedule, an API can meet its response target, and a service representative can still be looking at an outdated customer record. Every team may be meeting its commitments. The customer still gets the wrong answer.

I came into web analytics when companies were trying to understand activity on their sites by processing web server logs. As traffic grew, those files grew, and so did the time needed to analyze them. A business could be attracting more customers while waiting longer to understand what they were doing.

That history still influences how I judge a data architecture. I want to know when the information becomes usable in the system making the decision. A successful transfer is only one step toward that.

CDOs need to establish who is responsible for that gap. Someone has to own whether usable data reaches a business decision in time, across all the systems and teams involved. That responsibility should be explicit when an architecture is funded and designed.

Assign ownership across the handoffs

Suppose a customer changes a delivery address and then contacts service to confirm it. The order system has the change, but the service application reads an analytical copy that updates overnight. All the scheduled jobs can run successfully while the representative confirms the old address.

Faster ingestion into the warehouse would not necessarily fix this. The delay might sit in the service application’s refresh schedule or a cached view. Depending on the cause, the answer could be to query the order system directly, update the cache, or send order changes as events. You have to follow the change all the way to the person or system that needs it.

This is where technical ownership and business accountability need to meet. The service leader can define what the representative must know during the conversation. The application and data teams can work out how to provide it. An accountable owner needs authority to resolve failures across those boundaries.

The CDO’s role is to secure that agreement and make sure the work has funding. Assigning the problem to the team that runs the warehouse is insufficient if the delay sits elsewhere and that team cannot change it.

Measure what was available when the decision was made

Pipeline completion tells you that a job is finished. To understand whether it helped the business, you also need to know what the receiving application could actually use.

For an important customer interaction, I would track the share of decisions where the required information was available and current enough to use. That means agreeing on the required fields, how quickly relevant changes must appear, and which permissions apply. A recent update attached to the wrong customer record does not meet the requirement.

In the address example, measure how long it takes for a confirmed change to become available in the service application. Then check whether it was available during the customer’s conversation (Website, Mobile App, Contact Center, In Store). An average pipeline time can conceal the cases that cause repeat contacts or incorrect confirmations.

This also separates an architecture measure from a business result. Getting the correct record to the representative is something the data and application teams can verify. Whether that reduces repeat calls or improves resolution needs to be measured with the service team. Both matter when deciding whether the investment worked.

Spend on the delays that matter

Once the requirement is clear, the architecture choices become easier to evaluate. Some interactions need an update within seconds. A quarterly forecast may be well served by a scheduled batch. I would be reluctant to fund a real-time redesign without a clear account of what the current delay costs the business.

That cost might be repeated service work, an avoidable customer error, an offer sent after a purchase, or a new experience recommendation when the current experience hasn’t completed. The business owner should help establish the consequence and the acceptable delay. The engineering team can then estimate the cost of meeting it, including monitoring, failure recovery, and ongoing support.

Data gravity remains a constraint. As more applications depend on a large repository, moving data and changing those dependencies becomes harder. Sometimes it makes sense to bring processing to the data. Sometimes an operational system needs a small, current update that should not have to wait for a broader analytical workflow.

The choice could be a direct query, an event stream, or processing within an application or device. Historical analysis can remain in the warehouse. Each approach has costs and limitations, and each needs appropriate access and privacy controls. The required response time and the information needed should guide the choice.

This is also a reason to inspect the existing flow before buying another platform. If a receiving application refreshes only once a day, making the upstream pipeline faster may have little effect. Fixing that handoff could be the most useful investment.

Decide what AI should do when information is missing

The same delivery gap matters more when an AI agent can take action. In the address example, a service representative might ask the customer to verify the details. An agent could proceed with the record it has unless the application is designed to check.

A model’s training cannot tell it that an order changed five minutes ago. It needs access to the relevant operational information when it responds or acts. Current context helps, but the team still needs to test whether the system retrieves the correct record and uses it correctly.

Before giving an agent authority to change an order, I would want a clear answer about what happens when that record is unavailable or cannot be verified. The application might retry, pause, or refer the request to a person. The business and technical owners should agree on that behavior before deployment and test it with delayed updates and unavailable sources. It belongs in the acceptance criteria alongside response time and accuracy.

Start with one consequential decision

I would start with one customer interaction where late or incomplete information creates a recognizable cost. Trace it across the existing systems, agree on what must be available and when, assign an owner, and measure how often the application meets that requirement.

Then fund the change most likely to improve that result. It may involve a different delivery pattern, a better connection, or a change in how an application reads its data. Use what that work reveals to decide where the approach should extend.

Before approving the next data investment, I would ask the team to show one important decision it will improve, how the required information will reach that decision in time, and who is accountable when it does not. Those answers give a CDO a basis for judging whether a technically successful project will be useful to the business. Customers know data about them is being collected. The question is whether that collection results in value to them.

Related Stories

November 19, 2026  |  In Person

Cincinnati Leadership Summit

Westin Cincinnati - Downtown

Similar Topics
Artificial Intelligence
Data Management
Diversity
Testimonials
background imagebackground image
Community Network

Join Our Community

starElevate Your Personal Brand

starShape the Data Leadership Agenda

starBuild a Lasting Network

starExchange Knowledge & Experience

starStay Updated & Future-Ready

logo
Social media icon
Social media icon
Social media icon
Social media icon
About