An increasing number of documented novel coronavirus outbreaks have helped us understand when, where, how and why this disease is spreading. At a wedding in Jordan in March, the father of the bride, who had fever and cough a few days before the event, greeted his guests with hugs, cheek-kisses and handshakes. In the process, he unknowingly infected 76 of the 360 attendees. In China, a Buddhist — who was infected but had no symptoms at the time — rode to a religious event with their congregation on a bus where the air was recirculated but not exchanged. Over a third of the passengers on the bus were infected, though only one person sitting next to a window developed the disease, covid-19. In a restaurant in China, three families sitting across from an air conditioner were infected, while others sitting to the side of it were not.

With transmission, the devil is often in the details. If the father at the wedding had been a guest rather than the host, would that outbreak have occurred? If the bus had better ventilation, would that outbreak have been averted? If the air conditioning in the restaurant had been off, might some of those families been spared infection? Understanding the granular details of each outbreak by conducting an “epidemiological autopsy” for all confirmed cases is critical to begin characterizing and categorizing where and why transmission is happening. This information can then, in real-time and on an ongoing basis, be used to adjust our strategy and refine our response.

We now know several ways to prevent covid-19’s spread, including wearing masks, physical distancing, testing and isolating people who may have been exposed or have symptoms, and avoiding settings that are indoors, poorly ventilated or crowded. If most people did most of these things most of the time, we would have vastly fewer cases and probably enough reduction in transmission for the epidemic to die out.

But that is not what is happening. Transmission continues to plateau, wax, wane and periodically resurge. Part of the reason is that officials and public health experts continue to lack enough real-time data on where, why, how and to whom transmission is happening. Instead, we make assumptions based on a trickle of journal articles and media reports which, for all we know, may describe incidents that are the exception rather than representative of most of the virus’s spread. And we’re left repeating blanket messages to double down on the preventive measures we know work and sometimes speculatively suggest new ones — such as having people talk less in public, wear better masks or open more windows — that should reduce the chance of transmission, but without knowing if these interventions address the actual reasons the virus is still spreading.

Investigating where and why each case happened would help to understand where the public health system’s gaps are and what we need to do better or differently. Are most cases propagating from a few “superspreading” scenarios in places such as nursing homes and jails? Or are cases happening randomly in the community? Are cases in a town happening in a particular store, or just from a particular person? We also need to know how transmission is happening. Is it bad ventilation? Poor hygiene? Improper mask use? Lack of distancing? Do we need better masks? Or is there something else? There’s also the question of who. Are cases mostly among essential workers? If so, which particular types of workers seem to be most affected?

The answers to these questions and others lead to dramatically different conclusions on the corrective actions needed to prevent further spread — and they can change the strategy applied in each community. All epidemics are fundamentally local and driven by site-specific issues, such as types of essential work, transportation and density of housing. The reasons transmission thrives in one area may be different from why it persists in another. And even in the same location, what drove the spread of the virus in, say, April may be very different from what is causing it now.

We used this approach effectively in the West African Ebola epidemic, which one of us worked to fight. Once blanket measures — such as community engagement, promotion of preventive tactics, burying the dead safely and contact tracing — were in place, we analyzed every new infection to understand where and why it happened and then used that information to improve our response and push transmission rates down even more. For example, in one of the last places with Ebola cases, many new infections were caused by bodies of deceased Ebola-infected people not being buried safely. Our knee-jerk assumption was that communities that earlier in the epidemic had been resistant to changing the way they buried their dead were still reluctant. But when we looked closer, we realized that communities were, in fact, calling for assistance for safe burials, but the teams we organized as part of the response were reaching them many hours later. The problem wasn’t convincing communities to follow safe burial practices; they were already on board. What we needed was more burial teams, with dedicated vehicles ready to respond promptly to calls. We set that up and immediately saw transmission decline further.

We need to use the same approach to counter covid-19 and use this data to tighten gaps in our response and adjust our strategy. It can also help us guide messaging so the public understands how to avoid transmission. Most people want to do whatever it takes to protect themselves and their loved ones. Rather than just repeating slogans like “six feet apart” or “mask up,” officials and public health experts should provide people with tangible examples of how transmission is happening that they can use to understand how to adjust their habits. Stories based on actual instances of transmission would probably stick more than numbers, charts or generic slogans, and provide more accurate insights than the hearsay or exceptional outliers that end up filling the vacuum of information.

Insights on why transmission is persisting can underscore where different tactics may be needed. For example, if spread is mostly happening among essential workers, we then need to focus efforts on providing them with personal protective equipment, making it possible for them to stay home without losing income or similar measures. If transmission is persisting among undocumented groups, we need to then consider how to engage them in ways that build trust and assure them that getting tested or seeking care will not cause problems related to their immigration status.

This data can help identify the subsets of the population who are most affected in each region or city and should be prioritized by health officials for further intervention. For example, if the rapid antigen tests now becoming available could be targeted to the groups or settings driving most new cases, they could help reduce transmission below the threshold of each infected person infecting one additional person needed to shrink the epidemic even potentially without deploying daily population-wide testing. This will also be important for eventual vaccine deployment. If transmission is primarily mediated by particular subsets of the population, “herd immunity” with vaccination will only work if sufficient coverage is achieved within these groups.

In recent weeks, some health departments have begun listing tallies of where cases likely occurred, but this data needs to go a step further and also try to discern why transmission happened despite attempts to reduce spread. The Centers for Disease Control and Prevention issues investigation reports, but these are typically published months later, only cover a few outbreaks, and they appear to be at risk of tampering by Trump administration officials.

Even if the final product isn’t as exhaustive as the CDC’s reports, this type of investigation needs to be done for every case in real-time so the results can immediately feed into location-specific adjustments in responding to the virus. Newly diagnosed people are already supposed to be interviewed in this manner as part of contact tracing efforts that have not been keeping pace with cases in some states. We need more people trained to carry out these investigations, and hospitals should also integrate them into the routine intake when covid-19 patients are admitted.

Because the coronavirus spreads invisibly through the air, and often asymptomatically, it may not be possible to clearly establish the where and why in all cases. But for many, clear exposures can be identified, such as being at a crowded house party, working in a food-processing plant or living with other people with confirmed infections. In aggregate, these investigations can paint a narrative of transmission in each area and what needs to happen to reduce it.

We are at a crucial juncture in responding to this epidemic as winter is approaching. While we need to redouble our efforts on all fronts, a clearer view of where and why covid-19 is spreading would go a long way toward fighting the virus.