– enabling the data driven decision process in retail

Build on ADI™ (Automated Decision Intelligence) answering:

Where, what, which, why, when, how

Gilling launches ””

With comes a new type of data driven marketing automation systems for analysis, planning and execution of online campaigns: URM-systems – unknown relations management systems!

Facebook and Google´s target group and segmentation data are combined with detailed geographical, demographical and traffic data in a resolution of 100m. This makes it possible to create many microsegments and many campaigns in one work process, and to launch campaigns in highly optimized advertising zones, for example “around all shops in a chain in shopping distance”, or around “all highschools”, or in “areas with high income”. Experience from pilot tests shows that average click prices easily drop 25-50% and relevant reach increases 25-50% for the same investment, with only rude optimization of advertising zones. Calculation on more fine graded zones even show increased ROI up to 400% compared to standard procedures. Also, the high level of automation means that screen work can be reduced with more than 50%.


Screen dumps from cloud system: Systematic calculation of where customers come from, and where the retail chain can advertise with a high probability of getting customers into the shop.

Read more here:

The background to URM-systems for retail advertisers

From 1:1 to 1:many – using statistical data

Algorithms that predict expected sales revenue at a specific address using statistical data

Algorithms for estimating campaign effects of SOME-advertising and optimized budget allocation?

Results using in Facebook and Google advertising

The background to URM-systems

URM is short for ”unknown relations management” as different from traditional CRM, customer relations management. In an CRM-system you know the customer and have permission for a 1:1 communication. In a URM-system you don´t know “the customer” as anything else but a “statistical person with a probability” in a 1:many relation.

A customer journey starts via display advertising or search, continues with CRM and permission, and ‘ends’ either with long-term customer loyalty or with churn. In this process, the 1:1 relation and the permission are often systematized by SEO, but there has been missing systems and data that could systematize the creation of relations and permission via display advertising and 1:many relations. This has now become possible by the first URM-system

1:many relations have in practice so far been managed by Facebook and Google advertising, or by homepages, often in a combination with outdoor, search advertising and old school offline advertising, PR and influencers.

These channels often suffer from weaknesses. For example, optimization is done via the channel’s own media data and methods, because a channel has only the interest of selling more of the channel’s own advertising. This is contra to the interests of the advertiser, who wants to benchmark channel opportunities and compare according to reach, click and other action KPIs.

This is where steps in for the advertiser. Because offers knowledge and decision support to advertisers using advertiser-centered bid data. contains large volumes of data at 100m cell resolution, and can be used for both analysis, planning, execution and learning and, of course, automation of screen work. relates and integrates all types of data geographically and, by using statistical techniques, it models and predicts what works effectively and negatively in the advertising, and also explains ‘why’, using data in one collected report across all channels, Facebook and Google campaigns for example, all the advertising zones, campaigns, adsets, ads etc. This makes the perfect tool for dynamic testing and learning for each of the geographically limited market potentials any retail chain or social-demographically-influenced business depends upon communicating with. for retail advertisers

Using, any retail advertiser with shops, can calculate optimal advertising zones around each shop, identify target groups within the zones, gather large amounts of background data for each advertising zone and, for each target group and campaign, create and launch data-driven campaigns online. Advertisers can also, for each advertising zone, launch and execute campaigns, receive data from channels like Facebook and Google about expected reach and receive actual reach and click results. All this is done in a fully automated process that can be managed using a minimum of time.

For every round of campaigning, the investor and campaign decision maker, can learn from results and make even more accurate and precise decisions in the next campaign round. Due to the “many campaign automation features” it is also easy to set up a test pipeline, where different factors like demography, competition, ad content, adsets, channel type, timing etc. can be tested systematically to get knowledge about how each factor influences campaign KPI´s. No other marketing automation system has all these data ready at one place in one process! The pilot tests shows that using data to cut down on unnecessary advertising and improving targeting precision and accuracy works as we all intuitively know it does!

One of the single biggest advantages using, is that you can plan and execute far more campaigns per working hour, than what can normally be done. As you also get more data for analysis and learning gives you a steep learning curve for your online advertising. Like any medical research company, you can now create optimized pipelines to gain insight, effect, learning and find the fastest road to your success KPIs ahead of your competition.

Is there a future in teaching your organization how to use Yes, many things point in the direction of using statistical data for decisions on online advertising spending and allocation, for optimal location of shops in the retail network, logistical planning and in many more tasks within the functional areas of

From 1:1 to 1:many – using statistical data

The Executive Vice-President for ‘A Europe Fit for the Digital Age’, Margrethe Vestager, published in late February 2020, an initiative about EU organizing pools of statistical data to be used by EU businesses. This initiative should be seen in the light of Vestager and the EU fighting the use of 1:1 data without permissions, like Facebook and Google have been doing. By late 2019 both Apple and Google began dropping the use of 3rd party cookies, making 1:1 advertising more difficult. The trend instead  is towards contextual data to target advertising. Contextual data are often understood as “car advertising on car-sites”, but few have properly understood that contextual data can as well be low resolution socio-demographic and geographical behavioral data. Our behavior, and especially our shopping behavior, depends heavily on where we live and what we do here. You can be a houseowner or own an apartment, and your consumer behavior is VERY different regarding hundreds of product categories. You can be close or far away from a chain shop, and the probability that you buy in the shop drops with the distance between your home and the shop. makes it possible to use all these kinds of data and behavior systematically in targeting online campaigns, and the effects are, at times, very surprising.

Algorithms that predict expected sales revenue at an address using statistical data

Gilling has developed based on more than 25 years experiences in using statistical geographical socio-demographic big data in predicting retail sales “at an address”. Gilling has systematically developed the system for locating shops optimally for large Danish shopping chains like McDonald´s, Circle K and many more, and nowadays supermarket chains like Meny and Lagkagehuset. The core has been sets of low resolution big data on roads, addresses, properties, traffic, demographics and a large database of retail locations in Denmark, polled with clients own data. The algorithms used can predict retail revenue with an R2 above 90 and average uncertainty between 5-15%.

When you invest in shop locations (new, moving, closing), you run a considerable business risk, and gives a considerable reduction in this decision risk. Also, the chains can simulate expected revenue on say 30.000 locations among the 73.000 retail locations in the database.

Using does not require specific datasets available, but the more data in the more low resolution the better results will get. can in this respect also be used to test the value of datasets in reducing decision risks.

Algorithms for estimating campaign effects of SOME-advertising and optimized budget allocation?

On top of, a cloud system, any type of data can be related and integrated to work with all other data, has been developed to handle analysis, planning, execution and learning when investing in online advertising (and also offline advertising). has been developed to calculate optimal advertising zones, and documenting them with rich data. Data are sourced from all kinds of sources, also the channel data sources like Facebook and Google with campaign estimates on reach and clicks in the zones. Using these data, a retail advertiser can monitor campaign results with campaign potentials and over many runs, calibrate a model for estimation of KPI´s.

Both the probability of visiting a shop and the probability of seeing the ad, can be calculated per individual on the advertising zone.

Results using in Facebook and Google advertising can for example use and optimize on the fact that Facebook´s reach varies between 10% of population in advertising zones to 90% of the population in advertising zones on a weekly basis across Denmark (based on Facebook’s own data). These can be combined with probabilities of visiting a shop, local traffic behavior etc. Again tests show that click costs per relevant click in the advertising zone can be reduced by at least 25%, reach can be increased, especially running more campaigns closer to the shop for different target groups. At the same time, more campaigns can be tested using the automated features.

Thus uses and integrates with Facebook´s and Google´s own data and optimization algorithms. Neither Facebook nor Google optimizes by proposing which advertising zones to use. Such zones can easily vary between 1-30km per shop in a chain around in Denmark, therefore knowledge about these zones is crucial for not wasting advertising money on people that will never buy.

For more information about call or write:

Copyright: Gilling Aps, Symbion, Fruebjergvej 3, 2100 Ø

Finn Gilling, mobil 60211704, email,

The largest data- and knowledge base of danish retail – EVER BUILD! gives fast access to updated locations and data about ALL danish retail shops and chains (app. 66.000 shops, including 550 chains with app. 17.000 chains shops), in this example the Matas chain. In half a minut you have access to all the data you need about every Matas shop, and you can export these data to your own excel template to make the calculations you need.

Get to know your market: Fast inspection of potential new locations or explanations of existing shop´s revenues and
competitive situation? gives fast access to visual hot spot inspection checks of many different variables like income, traffic, age groups, consumption, houses, appartments and in general every variable that can be measured.


Professional retail investment management: Estimating revenue and ROI for new shop locations using Big Data and algorithms?
How much revenue can we make at this new location? gives access to ADI™ (automated decision intelligence).
Using algorithms prepared by integrating chain data with Gilling data, we are able to estimate potential revenue for a NEW SHOP with great accuracy and precision. For the technically minded, R2 often is above 85% and average uncertainty around 10%. easily implements all kinds of algorithms using big data, ADI™ and AI, and outputs fast calculations and easy to understand graphics.

Efficent business development potential: Simulate new shops at EVERY potential location and prioritize for development!
If a retail chain has developed a usefull algorithm, can be used to ”massrun” say 50.000 potential locations (addresses). For each location calculates potential revenue if a shop in the chain is located at the address, delivering a list of total national covering potential. These locations and their values shows where to search for optimal new locations, and are efficient tools for business development.

Gilling Products:

Retail MasterPlan™

Retail MasterPlan™

The best alternative. How much revenue can/should we create from this retail location? Where should we locate retail chain stores? How many of our retail stores can the market accommodate?

The Human Decision System™

The Human Decision System™

The moment of truthHow do humans make decisions?  How can we create algorithms for behavioural predictions that reflect the human decision process?

Big Data for sale

Big Data for sale

Integrating datasources. What data is available to support our own ADI system? Consult the Gilling Universe of Big DECISION DATA for the answers.

Retail Ranking™

Retail Ranking™

Knowing who & why. Which retail chains are the optimal renters for this location?  Simulate which of the +500 Danish retail chains can create the best sustainable business from any retail location.

Office Ranking™

Office Ranking™

Knowing who & when. How do we direct our marketing to the most probable companies to rent our available office space?  Office Ranking™ predicts your most probable prospective tenants!

Center Ranking™

Center Ranking™

Knowing who & what. What is the optimal retail profile of the shops in our shopping centre from the shoppers’ perspective? Do we have the right mix of shops?  Who´s missing, who´s redundant?

MEDIA Ranking™

MEDIA Ranking™

Optimize!… Where and how can we maximize ROI and response to our marketing investments using advertising geolocation analysis and all social and psychological profiling and thus optimise all social media activity?

Software development

Software development

On time on price. Develop effective ADI solutions based on Gilling’s standard and bespoke components and Big Data.

The Profit Chain™

How to automate and optimize your sequence of business decisions with ADI and Big Data?

Finn Gilling presenting The Human Decision System™ at Symbion Science Park, August 2016

Teach yourself how to automate Business Decision making

Use 5 minutes:

How to develop a model for ADI and Big Data

The Profit Chain™ is a general model describing all the decisions, that any business management is confronted with. As these decisions are depending on each other, The Profit Chain™ also describes the optimal sequence in which to make them, maximizing your degrees of freedom as you are using them.

To design automated decision intelligence, translate these general decisions in The Profit Chain™ into “your reality” and add goals and targets as measurement points:

An example: The “location” decision in The Profit Chain™ model translated for ADI with Big data:

Retail chains need to find the best locations for their shops. The target is an unidentified address among millions of addresses. The goal is to maximize revenue at one of these addresses with a new shop.

To automate your business decisions with ADI, work systematically through all the business decisions in The Profit Chain™, and define goals and targets. When this is ready, you can identify your measurement points (ex…target = an address related to data like the number of consumers within 500 meters to the address, or shopping traffic around the address etc.)

If you have not seen The Profit Chain™ before, here is a short 5 minutes introduction – push left/right with the arrows at the screen below and start with no. 1.

Read more and order the book on:

Gilling / The Human Decision ApS

Symbion Science Park

Fruebjergvej 3

2100 Copenhagen Ø

Finn Gilling


Tel.: +4560211704


Kontakt os

Skriv en mail til os og vi vender tilbage ASAP.