Learn from what worked. Plan what works next.

ubina is an independent marketing effectiveness consultancy. We use marketing mix modelling to show what really drove your sales, and to plan the budget, media and promotions that will drive them next.

Talk to us See our results
Illustration: past sales as a solid line up to today, then a forecast that widens into a shaded uncertainty range Today What happened What happens next 90% credible range Past Future

Selected client results

FMCG · Promotions

$10.3m

Extra annual profit from a re-optimised promotions plan, in one product range and one market.

Travel · Media mix

$10m

Yearly revenue gain identified by rebalancing the advertising channel mix.

Publishing · Media

+7%

More subscriptions, while cutting total media spend by 5%.

Financial services · Forecasting

−30%

Lower forecasting error, by modelling long-term and macroeconomic trends.

What we do

Answers to the questions behind your marketing budget

How much did each channel really deliver? What should we spend next year, and where? Which promotions pay back? We answer with evidence, and show how sure we are.

Marketing mix modelling

Our specialism. We measure how much each factor, from TV and digital to pricing, competitors and the weather, really drove your sales.

Media and budget optimisation

The right total budget and the best split across channels, timing, spot lengths and creative, using each channel's diminishing returns.

Promotion and pricing

The true profit of each discount, mechanic and display, net of cannibalisation and brought-forward sales.

Forecasting and scenarios

Sales and profit forecasts for setting targets, plus "what if" scenarios to test plans before you commit to them.

Test and learn

Store and region tests that isolate the effect of in-store activity, new creative or channels that models alone can't separate.

Simulators and dashboards

Interactive tools that put the models in your team's hands, so they can plan budgets and test scenarios themselves.

We also help get your data ready for advanced analysis, take on modelling projects for other consultancies and agencies, and train in-house teams to run the work themselves.

How we're different

Human-led, AI-powered data science

AI agents do the heavy lifting. An experienced modeller makes every judgement, and stands behind every number you see.

  • Human-led

    You talk to the person doing the work

    A senior modeller frames the questions, makes the modelling decisions and presents the results. Nothing is handed off to a junior team you never meet.

  • AI-powered

    Faster projects, lower costs

    We use agentic AI tools for data preparation, coding, diagnostics and documentation. Every step is reviewed by a person, so you get the speed without the guesswork.

  • Bayesian

    Results that say how sure they are

    Our models, built mainly in Python, show the likely range of every result, not just a single number. That makes it clear where to act with confidence and where to test first.

  • Independent

    Free to recommend what's right

    We aren't tied to a media agency or a software platform, so our advice follows the evidence.

  • Lean and flexible

    Senior expertise without big-consultancy costs

    A small core team, based in Colombia, that scales up when a project needs it. Low overheads mean more time on your project for less.

How a project runs

From your questions to decisions you can act on

  1. Agree the questions

    We work with you to decide which decisions most need evidence, and what success looks like.

  2. Gather and prepare data

    Sales, prices, media, competitors and anything else that moves your business, collected and cleaned.

  3. Build and test the models

    We build the models, then check them against data they haven't seen before we trust them.

  4. Deliver recommendations

    Clear findings through presentations, workshops and interactive tools your team can keep using.

  5. Support implementation

    We stay on hand as you put the plan into action, and answer the follow-up questions.

Case studies

What our clients changed, and what it was worth

FMCG · Promotion optimisation

$10.3madditional annual profit

Promotions looked as if they performed randomly. Once we accounted for everything else affecting sales, a clear and profitable plan emerged.

Read the case study

Challenge

The client judged promotions on simple volume uplifts per product. Cannibalisation, brought-forward sales, competitor activity, its own advertising and the weather weren't accounted for, so the same discount in the same position seemed to work differently every time, making planning unreliable.

What we did

Using promotional calendars, EPOS data and data on every other factor affecting sales, we built marketing mix models that measured the true impact and ROI of each promotion, by retailer and by product.

Result

A clearer view of the real revenue and profit of each promotion let the client cut cannibalisation and allow for brought-forward sales. The re-optimised plan added $10.3m in annual profit across one product range in one large European market.

Travel · Drivers of demand

$10myearly revenue gain identified

The client believed its market was too complex to model. A pilot project proved otherwise, and led to a wider roll-out.

Read the case study

Challenge

An international travel company couldn't see what was driving bookings, or how to balance spending on media, discounts and incentives for indirect sales channels.

What we did

We set up a pilot covering part of the business, modelling bookings against every factor, from advertising, pricing and promotions to things outside the client's control. For levers that rarely changed, such as sponsorship, we designed a test-and-learn plan.

Result

We identified $10m a year in revenue gains from a better advertising channel mix, and improved demand forecasts, which are central to the client's capacity planning. The pilot led to a planned roll-out across further areas of the business.

Publishing · Media optimisation

+7%subscriptions on 5% less media spend

After moving its titles behind a paywall, a publisher needed to know what its advertising was really doing for subscriptions.

Read the case study

Challenge

The group advertised across many channels, including its own print titles. It wanted to know each campaign's impact on subscriptions, how big the budget should be, and the best media mix.

What we did

We modelled subscriptions for each title against all marketing activity and pricing, plus competitor activity, the weather and the economy. This gave a cost per subscription for every campaign and channel, and showed where extra spend stopped paying back.

Result

The total budget was close to optimal, but some channels were clearly stronger than others. A modest change to the mix, with 5% ($2m) less spend overall, increased subscriptions by 7%.

Financial services · Forecasting

−30%forecasting error

Simple forecasts kept missing because longer-term trends were being ignored, which made it hard to plan resources.

Read the case study

Challenge

A financial services provider used simple demand forecasts for business planning. They were persistently inaccurate, causing problems in allocating resources.

What we did

Working with the client's insights team, we stitched together overlapping data sets into ten years of history, including the 2007–08 financial crisis. We then modelled the macroeconomic and other external factors behind long-term demand.

Result

We identified new factors for the client's own forecasting models and gave long-term trends the right weight. Forecasting error fell by roughly 30%, allowing resources to be allocated far more efficiently.

FMCG · Shopper marketing

+21%shopper marketing ROI, $0.5m more annual profit

In-store activity showed little visible uplift. Before running it again, the client wanted its true return.

Read the case study

Challenge

After several waves of in-store activity, including for a new flavour, the client couldn't see much sales uplift and wanted to compare its ROI with other marketing.

What we did

We built matched test and control groups of stores, with and without the activity, and used statistical methods to isolate the uplift, calculating ROI for each activity by store type.

Result

On average the activity paid back on revenue, but some activities lost money in some store types. Our changes to the plan raised overall shopper marketing ROI by 21% and annual profit by $0.5m across the brand.

Retail · Store estate

+19%revenue in remaining stores, with 13% more footfall

A tech retailer closing stores wanted to know which closures would send the most customers to the stores that stayed open.

Read the case study

Challenge

After a first wave of closures, the retailer didn't know which stores were best to close next, in terms of the impact on the rest of its estate.

What we did

Using three years of footfall and revenue data by store, before and after the first closures, we measured how distance to closed stores, and the features of the remaining stores, affected their footfall and revenue.

Result

We designed a selection strategy for the second wave that would deliver 13% more footfall and 19% more revenue in the remaining stores than the previous approach of simply closing the weakest stores.

Sectors we've worked in

  • FMCG / CPG
  • Retail
  • E-commerce
  • Telecoms
  • Banking and financial services
  • Insurance
  • Travel and hotels
  • Publishing
  • Pharmaceutical
  • Education

About us

Robert MacKenzie, founder and director

Robert is a marketing mix modelling specialist from the UK, with many years of work for household-name brands across consumer goods and services, in the UK and around the world.

Before founding ubina in Colombia, he worked with leading marketing effectiveness consultancies, leading projects from end to end: running the client relationship, managing the analyst team and building much of the modelling himself.

He set up ubina to offer that senior expertise without big-consultancy costs, combining Bayesian modelling with agentic AI tools. ubina works as a lean, flexible team, bringing in trusted specialists as projects need them.

  • Python
  • PyMC
  • R
  • Stan
  • Agentic AI
  • Power BI
  • Shiny

Contact

Tell us what you're trying to decide

Whether it's next year's budget, a channel you're unsure about or a promotion plan, we'll tell you honestly whether modelling can help, and what it would take.

Based in Colombia, working with clients worldwide.