Drawn from the work, not from any one client. The loop is the shape Continuum was built around.
Theory and Practice was a consultancy of model builders: PhDs in predictive analytics and
behavioural economics, doing work for some of the largest companies in North America. Like
most services firms, it wanted to be worth more than its contracts, which means becoming a
product company. It also did not want to become an analytics company, and it did not want
to build dashboards.
I joined as its fractional Chief Product Officer, a few working sessions at first, then two and a half days a week, from 2022 to 2024.
People ask for data. Very few know what they will do with it.
I talked about this with the CEO more than anything else. A client would ask for the
data. Ask what they would do with it and some had an answer and some did not. When they
did, the data they wanted often would not solve the problem they thought they had.
The other half was delivery. A client would sign believing its data was clean. Fortune 50
companies often believe that, and often it is not. Then the project waited on the client's
own infrastructure, or a privacy sign-off, and slipped by weeks.
A bootstrapped company protects its cash, and that is the CEO's job. So the next contract
still gets signed, sometimes at a discount, because payroll does not wait for a client's
data team. Then a clean project arrives on time, the two delayed ones come back at once
with their own urgency, and a small team has three projects landing in the same month.
Nobody did anything wrong. That is what ebbs and flows look like from inside
delivery.
Six engagements and a cleaning block, drawn to no scale. The lengths are illustrative; the mechanism is not.
The answer was not a price list. It was plumbing, and a product.
Nothing on this list worked on its own.
Capacity you could see coming. I worked closely with the operations leads
to predict capacity against projects, brought in Jira and a handful of small process
changes, and looked at the whole run of the business: the pipeline, the positioning, the
resourcing and the implementation.
A data engineering team. Instead of a custom build for every client,
pipelines built to a data standard most of the industry already followed, so the time from
a signed agreement to the first models running came down sharply.
Time protected for the future. The team was often booked past capacity.
We set part of it aside for real research and development, because nothing new gets built by
people who are always on the next deadline.
Continuum. The idea I brought: one product instead of four services, for
the customers whose data let you build once. Many models learning continuously, a layer
managing them, and next best actions coming out the other end. The vision was to let
customers venture fearlessly, on decisions they could trust.
The dashboard nobody wanted. The buyer was usually the chief marketing
officer, not the data team, and marketing teams were running experiments that pitted their
own sales teams against our predictions. Without a dashboard they could not watch the loop
work. So it became the feedback loop: progress the CMO could see, the trigger to retrain a
model when it drifted, and the place to suggest the next thing the existing models could
do.
A layer for how people decide. The team already worked in forecasting and
behavioural economics. I added the experience layer: how a prediction is presented, at the
moment someone has to act on it, changes whether they act.
Drawn from the work. The order is the argument; the lengths are not a measurement.
Who did what. The CEO protected the cash and the company through a
bootstrapped stretch, which is the job, and the reason there was anything to build on.
I worked most closely with the head of production. Much of what changed in delivery was
hers as much as mine.
A front-end developer with strong design instincts ran Continuum as its product manager.
His ideas and his execution made it real; with the wrong person in that seat it would not
have worked.
The data scientists and the engineers who built the pipelines did the work this page
describes, and many of them have since moved on.
My part: the idea and vision for Continuum, the experience layer, the capacity work with
operations, the case for a data engineering team and for protected research time, and the
call that a CMO needed a dashboard. I worked with the product manager until it was running,
then stepped back and advised him.
Continuum and its team were part of what Mphasis acquired.
If your team keeps asking for data, ask which decision it is for.