A quick guide to catalogue marketing data science

Catalogues may be living on borrowed time, but catalogue marketing data science techniques have been an important influence in other areas of marketing.

A quick guide to catalogue marketing data science
Ikea is ending its catalogue after 70 years. Picture © Ikea.
21 minutes to read

Catalogue marketing is dying out. Over the past few years, virtually all the UK’s top catalogue retailers have stopped printing on paper and successfully transitioned their businesses online, either to cut costs, meet demand, become more environmentally friendly, or simply stay alive. Even Ikea, which is said to have printed more copies than the bible, has finally given up after 70 years in the game.

However, while the remaining catalogues published by retailers are arguably living on borrowed time (at least in the non-specialist sector), catalogue marketing research has continued and has evolved significantly over the years. In fact, many of the techniques commonly employed in online retailing and marketing - including RFM - came from studies in this field.

Therefore, those businesses that still rely on catalogue marketing or direct mail (or perhaps, those whose management think they still rely on catalogue marketing), have access to decades of in-depth research to draw upon. This allows them to utilise techniques based on best practices developed by data scientists who pioneered this research field over many, many years.

Major catalogue retailers in the UK

Many of the major names in UK retail initially started life as mail order catalogue retailers. Some of these were the equivalent of the pure play ecommerce businesses of today, operating without physical stores, while others used catalogues or direct mail to supplement their high street stores.

The rise of the internet, and changes in customer shopping habits saw many of these mail order retailers cease publishing catalogues, change hands, or stop trading altogether. Arguably, this might be because they failed to adapt quick enough and move their business models online.

Most of the UK’s biggest catalogue publishers have now ended production, or are in the process of phasing out direct marketing and especially catalogues. Relatively few mail order businesses remain, having already either undergone digital transformation, or been too slow to react to the changing retail landscape.

Company Date Background
Ikea 1951 - 2021 Ikea printed catalogues for over 70 years and was the world's biggest catalogue publisher. At its peak it published 200 million copies per year. In late 2020, it announced its 2021 catalogue would be its last, citing a drop in demand due to the rise of its website.
Argos 1973 - 2021 The Argos catalogue, or "book of dreams" as it was colloquially known, lasted 48 years. Its 2021 issue, like Ikea's will also be its last, following a successful shift to online retailing.
Littlewoods 1932 - 2004 Littlewoods started retailing in 1923 and published its mail order catalogue in 1932. It merged with Kays Catalogues in 2004, creating Littlewoods Shop Direct Group, now called Shop Direct Group.
Kays 1890 - 2004 Kays Catalogues started life in Worcester in 1890 as Kay and Co. Ltd, and published its catalogue selling clothes and other products until 2004, when it merged with Littlewoods and formed what is now Shop Direct Group.
Grattan 1912 - Ending Grattan is an English clothing mail order catalogue retailer and started in 1912, and is now owned by Otto Group. It appears to be in the process of switching to online and no longer accepts requests for its catalogue, citing environmental benefits.
Freemans 1905 - Ending Freemans started in 1905 and is a mail order clothing catalogue retailer. Like Grattan, it's also owned by Otto Group. It's been moving towards the web since the mid 2000s, when it changed its logo to Freemans.com. It no longer accepts requests for catalogues.
JD Williams 1859 - Ending JD Williams (part of N Brown Group PLC) has been around since 1859 and currently specialises in women's fashion. It's transitioning to ecommerce and no longer allows customers to request its catalogue, encouraging them to use the website instead, citing environmental benefits. It produces several catalogues for its brands.
Screwfix 1980 - Present Screwfix, a specialist company which supplies hardware to the building trade and consumers, has published 140 catalogues in its 40 year history. As of 2020, it continues to publish one.

png The former home of the catalogue retailer Littlewoods. Picture by Matthew Daniels, Unsplash.

Catalogue marketing benefits

Catalogues, of course, are not without their benefits. It would be annoying and time-consuming to browse an entire product portfolio or full product line via a website, but it’s acceptable and enjoyable using a catalogue. A catalogue can be inspiring, promote products well, can make a brand feel more premium, and help set it apart from rivals.

There’s also a chance that customers who receive a catalogue may keep it and refer back to it later when they need something, or be reminded of the brand and search for them before making a purchase online. Catalogues might aid “Front of Mind Awareness”, with the hope that they also increase sales as a result, or encourage customers to place orders for different types of product and expand their basket mix.

Thankfully, there’s little that data scientists can’t measure, so it’s possible for these assumptions to be tested. This is most likely what the catalogue retailers above did before they cautiously decided stop printing their catalogues and fully embrace digital transformation.

png Picture by Semen Borisov, Unsplash.

Catalogue marketing challenges

While many of the techniques used in catalogue marketing overlap with those of digital marketing, and many of the processes used were pioneered by catalogue marketers, there are some important distinctions that make catalogue marketing much more challenging.

The main issue is with the costs. It’s cheap to send targeted emails to potential or existing customers, and little or no costs are incurred when targeting fails to hit the mark or customers are unresponsive. However, when you are paying £3-4 to deliver each copy of a catalogue to a customer by post, it’s absolutely vital that you get a decent conversion rate and that the catalogue increases CLV. Printing and mailing catalogues is expensive.

Even a seemingly impressive 10% conversion rate on a catalogue marketing campaign with a cost per send of £4 will set you back £364 per conversion (£4 + ((100-10) x £4) ), because you’re paying for every order that failed to convert as well as every one that did! At that price, and a whopping 50% margin, you’d still need a mighty big AOV or multiple orders just to break even…

According to research by the Direct Marketing Association or DMA, the typical conversion rate from catalogue marketing is in the low 3% area. As a result, you can see why direct marketers invest so much effort into segmenting their lists and trying to increase the return on investment they can generate from mailings.

Issue Impact
Production time Catalogues are time-consuming to produce. It takes a team months to produce the typical catalogue. This requires company-wide collaboration to ensure that those due to provide products, prices, imagery, content, and data, don't compromise the print slot.
Production costs As well as the costs of paying the staff involved in the creation of the catalogue itself, there are also photographic costs, paper costs, and printing costs, which will run easily into the tens or hundreds of thousands on small or modest print runs.
Distribution costs Paper is heavy and cataloguers want their product to arrive looking great, so there are also distribution, postage, and packing costs on top of the price of production and printing. When these are added up, costs could easily be several pounds per catalogue, particularly if economies of scale are low.
Price longevity Once you've printed a catalogue, you can't change the prices shown, and usually honour the prices while the catalogue is in-date. However, online retailers check and update prices on a daily basis, so if they increase or decrease prices, you could lose sales or margin.
Unscientific approach While the data science techniques for catalogue marketing have been around for decades and are robust, the lack of data science skills within some businesses means that they aren't aware of the available techniques and some companies may continue to apply an unscientific historic approach.
Lack of testing One of the key things you're taught in marketing is the importance of testing. Testing techniques came from catalogue marketing and were adopted by digital marketers, yet weirdly, some companies are still reluctant to allow this, especially where hold-out groups are concerned.
Changing demand Not only do most people refer to their phones, tablets, or computers and go online instead of picking up catalogues or telephone directories (remember those?), they're also more environmentally conscious, and the low conversion rate on catalogues means they're wasteful to print and deliver.
Product selection With every page you add greatly increasing the costs of printing and mailing, it's vital that each product pays for its space and delivers a good return on investment. There are some clever techniques available to help retailers do this, but they're often not used.
Distraction from focus Finally, there's distraction from focus. If a company's catalogue distracts the staff from working on more profitable areas of work, it could limit the company's growth and profitability.

png Picture by Kaboompics, Pexels.

Catalogue marketing data science

During a job interview, I was once told by the interviewer that “half the money I spend on advertising is wasted; the trouble is, I don’t know which half.” I bit my tongue, but obviously, in these days of self-driving cars and other innovations, this outdated view no longer holds water.

While some naive people may think you can’t measure everything, it’s certainly possible to see whether catalogues work, if you know the right techniques to apply. The end results of almost everything is measurable, and there are decades of research to draw upon.

Research falls into three primary fields: customer selection or targeting, catalogue performance analysis, and product selection. Improved profitability from catalogue marketing really comes from one or two key things - reducing costs and increasing conversion.

1. RFM analysis

The classic method used in catalogue marketing is RFM, or Recency, Frequency, and Monetary value analysis. This uses the assumption that existing customers who have shopped most recently, placed more orders, and spent more money, will be much more likely to purchase than those who have placed, say, a single order for a pair of socks in 2001, and haven’t been seen since.

When used correctly to construct mailing lists, RFM can greatly decrease direct marketing costs and help businesses target those most likely to be active customers. While the standalone usage of RFM has been superseded by more sophisticated methods, the RFM metrics themselves (not to be confused with the RFM scores) remain just as valid in even the most advanced models.

Almost everything still does, and always will, revolve around RFM data. This means the technique is just as applicable in other areas of marketing which require you to target customers, especially email marketing. It can massively increase return on investment.

2. Break even analysis

Break even analysis typically follows on from RFM and uses the scores, rather than the raw data, to analyse who purchased after they were mailed. The aim is to understand how each of the 125 RFM quintiles from 111 to 555 respond.

Each quintile will have a different response rate, with the cost of reaching higher quintiles (i.e. 555) being significantly lower than the bottom quintiles (i.e. 111). For example, it might cost you £20 to get a conversion from a 555 customer, and they might generate an average of £100 in revenue, or £20 in profit. However, it might cost you £150 to get a conversion from a 111, and they might spend just £80, and generate a loss.

The aim is to measure the response rate for each quintile from the response data, calculate the profit generated, and identify the break even point at which it becomes unprofitable to mail a quintile. You can then select only the customers with RFM scores of, say, 334 or more to ensure you generate a profit and don’t fritter away the budget contacting unprofitable quintiles.

3. Cohort analysis

Cohort analysis aims to look at groups of customers over time to identify whether there are any differences in their behaviour. It’s commonly applied to direct mail marketing, including catalogue marketing, to understand whether customers who are receiving the catalogue behave differently to those who do not.

If you publish additional brochures or seasonal catalogues, as well as your main one, you can also use cohort analysis to examine how they work in conjunction with each other, if at all.

4. Marketing tests

Catalogue marketing tests aren’t always that easy to pull off, partly because customers differ so much in their behaviour, and partly because the key stakeholders are often reluctant to include their best customers in marketing tests.

The classic tried-and-tested approach was to use an “nth”, whereby every nth customer is selected at random for inclusion within the test group. For example, if you want to include 10% of your customers in a test group, you’d select every 10th one from your data to end up with a random group.

However, even with this technique it will be necessary to check that the data are, for want of a better word, “balanced”. When judging performance afterwards, you’ll likely be comparing the behaviour of the test group with the control via a t-test, or similar statistical method. A statistically significant difference between your chosen metric across the two groups will tell you whether the test worked or didn’t.

The skewness of retail data means that sometimes all it takes is one or two unusual customers to fall in either group, and the results can look very different. Checking that there’s no statistically significant differences in your test metric over previous periods for the two groups can help avoid this, but is often computationally intensive.

5. Holdout groups

The other thing you will want to do to improve the performance of your catalogue marketing is to include a holdout group. This group of customers won’t receive any marketing, often for a prolonged period, so you can measure the performance improvement you get against a baseline group where no marketing was provided.

The snag is, some stakeholders really hate holdout groups. They get worried that they’ll lose customers by not marketing to them and don’t want to take what they see as a risk. Instead, they’re happier to spend the marketing budget like it is water, potentially wasting good money where it could be saved. Or, perhaps they just don’t like hearing that their historic approaches to marketing no longer cut it in the modern world?

6. Match Back

To measure whether customers have responded to your marketing or not, you’ll need to be able to identify who you mailed and then check who purchased, irrespective of the purchase channel they used to make their order.

This is fairly straightforward. You simply store the list of customers you mailed, then wait for a specific period representing the campaign’s active window, then check against your data warehouse to see which of the customers mailed placed an order.

This Match Back process is fine, but it can cause issues when you assume that a direct mail marketing campaign was the cause of a purchase. Customers these days have numerous touch points so attributing a sale to a catalogue or brochure, just because a customer was mailed one does not mean it was the cause of their response. It’s not ideal for attribution modeling on mailing lists.

7. Response models

When it comes to the selection of customers who will receive a catalogue, there are several approaches, with response or propensity models being the most common after standard RFM. Response models examine data from previous direct mail campaigns and examine the relationship between each customer’s features (i.e. RFM, latency, churn, CLV, and other values) and their probability of responding to a mailing during a campaign’s active window.

The same response model approach works brilliantly with outbound sales projects and email marketing, among other things, and is fascinating to implement and often very effective. Its shortcoming is that it will inherently include customers who would have purchased anyway, just as the RFM approach does. However, there’s a solution to this - uplift modeling.

8. Uplift models

The uplift model aims to overcome the drawback of response or propensity models, in that they select customers who have a high propensity to purchase, irrespective of whether they’ve been influenced to purchase because you sent them a catalogue.

Uplift models are arguably one of the most powerful tools used in catalogue marketing. One common objection is that mailings contact people who would have purchased anyway, which is true. However, uplift modeling does allow you to identify those customers who will purchase irrespective of marketing before you mail. Though, in practice stakeholders might like the comfort blanket provided by mailing them anyway.

There is one downside to the uplift modeling approach: it’s much harder to implement and typically requires two or more models to pull off. It also requires the use of test and holdout groups for solid results, and again, some stakeholders don’t like these.

Further reading

  • Cui, G., Wong, M.L. and Lui, H.K., 2006. Machine learning for direct marketing response models: Bayesian networks with evolutionary programming. Management Science, 52(4), pp.597-612.

  • Diemert, E., Betlei, A., Renaudin, C., and M-R, Amini (2018) - A Large Scale Benchmark for Uplift Modeling. Proceedings of the AdKDD and TargetAd Workshop, KDD, London,United Kingdom, August, 20, 2018. https://ailab.criteo.com/criteo-uplift-prediction-dataset/

  • Gubela, R., Beque, A., Gebert, F., & Lessmann, S. (2019). Conversion uplift in ecommerce: A systematic benchmark of modeling strategies. International Journal of Information Technology & Decision Making. doi:10.1142/s0219622019500172

  • Hossein Javaheri, S., 2014. Response Modeling in Direct Marketing: a data mining based approach for target selection. Data Mining Applications with R, chapter 6.

  • Kane, K., Lo, V.S. and Zheng, J., 2014. Mining for the truly responsive customers and prospects using true-lift modeling: Comparison of new and existing methods. Journal of Marketing Analytics, 2(4), pp.218-238.

  • Neslin, S.A., 1990. A market response model for coupon promotions. Marketing Science, 9(2), pp.125-145.

  • Radcliffe, N. 2008. Hillstrom’s MineThatData Email Analytics Challenge: An approach using uplift modeling. Stochastic Solutions. https://stochasticsolutions.com/pdf/HillstromChallenge.pdf

  • Radcliffe, N.J. and Surry, P.D., 2011. Real-world uplift modelling with significance-based uplift trees. White Paper TR-2011-1, Stochastic Solutions, pp.1-33.

  • Rzepakowski, P. and Jaroszewicz, S., 2012. Uplift modeling in direct marketing. Journal of Telecommunications and Information Technology, pp.43-50.

  • Shimizu, A., Togashi, R., Lam, A., & Huynh, N. V. (2019). Uplift Modeling for Cost Effective Coupon Marketing in C-to-C Ecommerce. 2019 IEEE 31st International Conference on Tools with Artificial Intelligence (ICTAI). doi:10.1109/ictai.2019.00259

  • Sołtys, M., Jaroszewicz, S., & Rzepakowski, P. (2014). Ensemble methods for uplift modeling. Data Mining and Knowledge Discovery, 29(6), 1531–1559. doi:10.1007/s10618-014-0383-9

  • Kabaskal, İ., 2020. Customer Segmentation Based On Recency Frequency Monetary Model: A Case Study in E-Retailing. International Journal of InformaticsTechnologies, 13(1).

  • Uysal, Ü.C., 2019. RFM-based Customer Analytics in Public Procurement Sector (Doctoral dissertation, Ankara Yıldırım Beyazıt Üniversitesi Sosyal Bilimler Enstitüsü).

Matt Clarke, Saturday, March 13, 2021

Matt Clarke Matt is a Digital Director who uses data science to help in his work. He has a Master's degree in Internet Retailing (plus two other Master's degrees in different fields) and specialises in the technical side of ecommerce and marketing.

Building Recommendation Engines in Python

Learn to build recommendation engines in Python using machine learning techniques.

Start course for FREE

Comments