The short answer
Catalogs with similar costs can reach different customers and promote different assortments, producing different purchasing responses. Spend alone does not describe those differences. Where recipient and order data are available, use the observed response as a modeling input and calibrate its incremental share with experiments. Keep campaign attributes available to investigate why performance varies.
This article is part of Asked by Marketers, a series answering real questions from marketing leaders.
Sellforte's team holds more than 1,450 meetings each year with marketing leaders in Ecommerce and Retail about Marketing Mix Modeling and incrementality testing. Each week, we anonymize at least one question from those conversations and answer it in depth, based on what marketers are actually struggling with, not what keyword tools suggest. About the series →
Why marketers ask this
A catalog team can see that some mailings sell well and others disappoint, even when circulation and production costs look similar. A model that returns much the same performance for each send is hard to use for planning. The team needs to understand which useful differences are missing from the inputs and how to distinguish recipient purchases from sales the catalog actually caused.
Why might catalog spend be an incomplete modeling input?
Spend describes the cost of the mailing, but not the customer's response to it. Two catalogs can have similar costs while differing in audience selection, assortment, pricing, or the timing of purchases.
Those differences matter to Marketing Mix Modeling. A cost or circulation series may show when a mailing occurred and its scale, yet leave little information for distinguishing a strong campaign from a weak one with a similar budget.
Keep the business definition of a catalog campaign clear. A mailing to active customers, a reactivation mailing, and an acquisition mailing to unknown recipients can have different goals and data. Combining them under one print label can conceal distinctions that the team uses to make decisions.
Cost and circulation remain important. They establish the investment and distribution scale. An additional response signal can help distinguish campaigns when cost and circulation are similar.
What data can help distinguish catalog campaigns?
Where customer linkage is available, connect the catalog identifier to recipients and their subsequent orders. Preserve the purchase dates so the input reflects how the response unfolds after the send.
The campaign identifier provides the connection between separate datasets: who received a mailing, what it cost, and which purchases fall within the agreed attribution rule. The rule must define how long purchases remain associated with the mailing and how overlapping sends are handled.
| Input | What it contributes |
|---|---|
| Catalog ID, send date, and recipient information | Identifies the campaign and the customers who could respond. |
| Orders, values, and purchase dates | Describes the observed purchasing response under a stated attribution rule. |
| Campaign cost and circulation | Provides investment and distribution measures. |
| Audience, assortment, and offer attributes | Helps investigate why campaigns with similar costs perform differently. |
| Relevant incrementality experiments | Provides evidence about how much of the response the mailing caused. |
Choose the response window around the business's purchase pattern and measurement purpose. A window used for one catalog program is not a universal standard. If purchase dates already describe a delayed response, check how the model treats that timing before applying additional carryover assumptions.
Metadata can make the results easier to investigate. For example, the team can compare campaigns by audience type or product range. Those comparisons can suggest explanations, but an assortment label alone does not prove that the assortment caused the performance difference.
How do we separate recipient purchases from incremental sales?
Use experimental evidence to estimate the portion of the observed response caused by the catalog. Customers selected to receive a mailing may have purchased even without it, especially when targeting already favors customers likely to buy.
A recipient-to-order link is therefore an attribution signal. Treating every linked purchase as incremental would retain the selection bias in that signal. Incrementality testing adds a comparison with customers or areas that did not receive the treatment, using a design appropriate to the campaign.
Match the evidence to the audience and catalog program being modeled. If targeting changes from active customers to reactivation, check whether earlier evidence still supports the incremental-share assumption. An experiment for the whole print program may also support a different conclusion from an experiment that isolates one mailing.
Sellforte's article on combining MMM, attribution, and experiments explains how experiments, MMM, and attribution can work together. For catalogs, the practical distinction is that the observed purchase pattern helps describe the response, while experimental evidence informs how much of that response is incremental.
What if we cannot link recipients to orders?
Use the distribution information that is available and keep the acquisition mailing distinct from addressed campaigns with known recipients. An unknown recipient cannot reliably enter the same customer-to-order matching process.
Distribution dates, circulation, geographic coverage, and relevant sales outcomes can still inform an aggregate analysis. The feasible detail depends on those data and the variation in activity. Do not manufacture a customer link to make the mailing fit an addressed-catalog dataset.
Coupon codes can provide a partial response signal, but they only identify purchases where the code was used. A customer may see the mailing and buy without entering the code. Treat code redemptions according to that coverage, rather than as a complete count of every sale influenced by the mailing.
Where geography and distribution allow a suitable comparison, a geo-lift experiment may help assess an acquisition mailing. That is a different design from matching known customers to their orders, and its result should retain its own scope and uncertainty.
How this looks in practice
Consider two hypothetical addressed catalogs, each costing $50,000. Assume a consistent attribution rule links $200,000 of subsequent recipient purchases to catalog A and $300,000 to catalog B. Relevant experimental evidence is assumed, for illustration, to support incremental shares of 40% and 20% respectively. These inputs are invented to demonstrate the calculation.
| Measure | Catalog A | Catalog B |
|---|---|---|
| Catalog cost | $50,000 | $50,000 |
| Linked recipient purchases | $200,000 | $300,000 |
| Assumed incremental share | 40% | 20% |
| Illustrative incremental revenue | $80,000 | $60,000 |
| Illustrative iROAS | 1.6 | 1.2 |
Catalog B has more linked purchases, while catalog A has the larger incremental contribution under these assumptions. Similar costs cannot explain that difference, and recipient sales alone would rank the campaigns differently.
The percentages in the example are not benchmarks and cannot be inferred from the purchase totals. In practice, the team needs evidence for the incremental share and should review its uncertainty before changing the mailing plan.
How Sellforte helps
Sellforte can combine catalog cost and response data with experimental evidence when building the measurement setup. The approach depends on whether recipients can be linked to purchases and which campaign distinctions the available data support. Book a demo.
Authors
Lauri Potka is the Chief Operating Officer at Sellforte and has over 15 years of experience in Marketing Mix Modeling, marketing measurement, and media spend optimization. Before joining Sellforte, he worked as a management consultant at the Boston Consulting Group, advising some of the world's largest advertisers on data-driven marketing optimization. Follow Lauri on LinkedIn, where he is one of the leading voices in MMM and marketing measurement.