Measuring returns to advertising – case study at a cloud computing scaleup

No Thumbnail Available

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Master's thesis

Date

2024

Major/Subject

Mcode

Degree programme

Economics

Language

en

Pages

41 + 7

Series

Abstract

Firms invest significant resources in advertising, yet often lack clarity of the outcomes. The main challenges in measuring returns on advertising spend (ROAS) stem from two sources. Firstly, low signal-to-noise ratio – sales is a noisy variable exhibiting seasonality and randomness and advertising spend is typically low in comparison to sales. Measuring the effect of this weak signal on noisy sales results in estimation uncertainty. Secondly, selection bias arises from both ad targeting and advertisers increasing ad spending in anticipation of increased demand, which may lead to overestimation of ROAS. These two factors make measuring returns to advertising challenging. This thesis comprises two main sections. In the first section we review four common methods utilized in measuring returns to advertising: attribution models, geo experiments, Media Mix Modelling, and Bayesian Structured Time Series model. We discuss their applicability and common pitfalls. In the second part, a Media Mix Model variation designed for search advertising is utilized to measure the ROAS of the case company that operates in the cloud computing industry. The main results reflect the common challenges of high estimate uncertainty and the potential for underlying bias. Furthermore, a secondary analysis indicates that directing advertising towards competitor keywords may yield higher returns compared to category keywords. This thesis helps us understand the challenges of measuring ad effectiveness and sheds light on advertising performance of a scaleup operating in the cloud computing industry during 2019-2021.

Description

Thesis advisor

Terviö, Marko

Keywords

advertising, ad effecs, media mix modelling, attribution models

Other note

Citation