Is Marketing Similar To Medicine?

Pranay Aryal
Jul 21, 2017 · 4 min read

This applies to research studies. But it can also apply to A/B testing, machine-learning, market research or any interventional research. The mnemonic PICO is used for this.

POPULATION — What is the population being studied? Does the sample being studied generalize to my case? In machine-learning terms, will this be an overfit?

If this study is only done on females, the result will not generalize for males. If this study is done on the western population, it will not generalize for Asians. If this study is done one one species of fish, it will not generalize to other species.

INTERVENTION — What is the intervention being done on the population or what is the exposure being studied?

The intervention could be a drug, surgery, an educational campaign, a marketing campaign, a social media post, or it could be with-holding of something like smoking cessation.

In medicine, ethical questions can arise. If I give vaccination to one group but withhold it from another group, it is not ethical because the current standard of care is being denied to a group.

Can the intervention cause harm? Has it been shown to cause harm in the past? These are ethical questions to consider.

We need to be careful about conflicts of interest — a scientist who can make money off a drug he researches could be biased, a pharmaceutical company doing research on a drug or device they wish to sell could be biased. We need to pay special attention to who has funded the research. Funding by private companies can give rise to conflicts of interest if they are also trying to sell it.

If someone is talking a lot of science about something they are trying to sell, they could be talking pseudo-science.

COMPARISON — The best way to determine causality is by having controls to compare with. This makes the evidence stronger. Compare the population on which you did your marketing campaign with the population on which you didn’t.

If you want to study the impact of a fertilizer for a few plants, have a control group on which you don’t apply the intervention so that you can compare the effect in both groups.

Ideally, the entity being studied should have equal chance of being assigned to the intervention group or the population group to avoid bias.

If a patient is being given a placebo (control group), the patient should not know if he/she is taking the real medicine or the placebo to avoid biased reporting of symptoms (if symptoms were being studied). This is known as blinding.

If you want to study the impact of an educational campaign, have a control group on which you don’t campaign and then compare. But blinding can be difficult here so be aware of this.

Beware of studies where the investigator knows beforehand if the intervention group is more likely to respond than the control group. This can give bad results.

That is why randomized controlled trials (RCTs) have the highest strength of evidence. ‘Randomized’ means the entity being studied has equal chance of being assigned to the control or intervention group (a biased human should not hand-pick the two groups) and ‘controlled’ means that there is a control group to compare with.

OUTCOME — What is the outcome after the intervention? Is it something which can be easily measured (this should be determined at the outset)? Adverbs like honesty, simplicity, greenery are not good as outcome because they are abstract and cannot be measured.

Have an outcome that is simple to measure, that gives very less intra-observer and inter-observer variability, outcomes that don’t depend on opinions or judgment of the person measuring it.

Ideally, the person who is measuring the outcomes in intervention and control groups should not know which group the individual being studied is. This is also blinding. Knowing the groups can bias the results.

Of course you will need statistics to see whether the difference that you observe (if you observe a difference) is because of real difference or because of chance. Usually the results are reported in terms of p-values.

For example, the results might say that there was significant difference between the intervention and the control groups with a p-value of 0.0001. This means that there is only a 1/10000 (very less) odds that this difference could have occurred because of chance rather than, because of the intervention, which is a good thing.

Ideally the p-value should be less than 0.05

Sometimes negative results can be insightful. Say a surgeon came up with a better technique of surgery using the latest equipments. This technique is compared with the subset of patients who get the normal standard surgery (control group). The study did show a difference between the groups. This could lead us to conclude that the more expensive technique has the same results as the current technique so we would stick with the current standard of care.

So I hope you will remember PICO.

)
Pranay Aryal
Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade