Finding silent exhibitor accounts before the renewal call goes badly

The call lasts four minutes. The rep opens with the island position they have been holding, the exhibitor says they have taken space at the competing show in March and their board signed it off in June, and there is a pause where the rep works out that this happened eleven weeks ago and nobody noticed.

Afterwards the account gets coded lost, reason budget, which is untrue and unfalsifiable. The post-mortem concludes that the team should have stayed closer to the account.

Silent exhibitor accounts are findable, and staying closer is not how you find them. Two thousand four hundred accounts and twelve reps is two hundred accounts each, and no amount of diligence covers that. What covers it is a measurement, because the thing that happened between the last edition and the June board meeting is observable in your systems and it has a number attached.

What should you measure when an account goes quiet?

The obvious approach is days since last contact, and every CRM has a field for it. Almost every implementation gets it backwards by counting outbound activity.

A rep who logs four calls, two emails and a LinkedIn message to an account that answered none of them has produced a last-contact date of yesterday and an account that has said nothing for six months. The field is measuring your team's activity, which is a real thing to manage and has nothing to do with the exhibitor's state.

What predicts is exhibitor-initiated contact. An action the exhibitor chose to take, costing them time, directed at you. Everything else is noise generated by your own process.

The four clocks worth running are the ones where the exhibitor moved first.

Inbound contact. A reply to an email, an inbound call, a meeting they asked for, a question about next year. Replies are the hard part, because they need to be captured from the mail integration and most CRM activity logging records the send rather than the response.

Portal session. A login to the exhibitor portal outside the periods when you have forced them to log in. Sessions during the mandatory manual-upload window measure compliance. Sessions in February on a September show measure interest.

Service or upsell order. Any spend beyond the contracted space. Furniture, extra power, a catalogue entry, a bag insert. Money moving voluntarily is the strongest of the four. Note that this is order activity and not the support ticket record, which measures something closer to your own delivery and needs normalising by stand size before it means anything (G16).

Marketing response. A click through to a real page, a form completion, a webinar registration. Weak individually and useful in aggregate, and only if you exclude opens, which measure image loading and mail client behaviour.

Take the most recent event across all four and count days from it to today. That is the account's quiet days.

Percentile of the base, worked

Absolute thresholds do not travel between shows, because a quarterly-cycle show and an annual one have completely different natural gaps. Set the threshold from your own distribution.

Take a portfolio with 2,400 exhibitor accounts. Compute quiet days for every one of them at the same reference date, which should be a fixed number of weeks after show close so that the measurement is comparable year to year.

Suppose the median comes out at 38 days, the seventy-fifth percentile at 84, and the ninetieth at 121. The ninetieth percentile puts 240 accounts on the list, which is a workable number, and it is workable by construction because you chose the percentile to make it so. Choosing it against what your reps can actually get through is capacity planning for the call list and belongs to G22.

Now check whether the flag is worth anything. Follow those 240 accounts to the renewal and compute their churn rate against the base. If the base churn rate is 28 per cent and the quiet cohort churns at 46 per cent, the flag gives a lift of 1.64 and the 240 accounts contain 110 of your eventual losses. Whether that is good depends on what else you have, and it is a genuine number produced from data you already hold.

An account sitting at 214 quiet days against a median of 38 is not borderline. It is five and a half times the typical gap, and something happened.

Against the base, or against themselves?

The percentile approach has an obvious failure that shows up the first time a rep reads the list.

Some exhibitors are structurally quiet. A long-standing account with a two-person marketing team who book the same 36 square metres every year by return email have never contacted you outside the renewal window in nine editions. They will sit above the ninetieth percentile permanently, they will be on the list every year, and after the second year the list loses credibility.

The fix is to give each account its own baseline. Compute the median gap between exhibitor-initiated events over the account's prior two editions, then express current quiet days as a multiple of that.

Account X has a typical gap of 22 days and now sits at 214, a ratio of 9.7. Account Y has a typical gap of 140 days, because they genuinely only speak to you at renewal, and also sits at 214, a ratio of 1.5.

Both are at 214 quiet days. X has changed and Y is behaving exactly as it always has. A list built on the absolute figure contains both. A list built on the ratio contains X, which is the account where something is worth finding out.

I would run both and rank on the ratio, keeping the absolute figure as a displayed column so a rep can see the raw fact. Accounts new enough to have no baseline fall back to the base percentile, which is the correct default for a first or second edition exhibitor.

What silence is evidence of

Bolton modelled the duration of a customer's relationship with a continuous service provider in Marketing Science in 1998, treating cumulative satisfaction as an anchor that each new service experience updates. Her estimates put satisfaction and relationship duration in a positive relationship whose strength depends on how much history the customer already has, with long-standing customers weighting their accumulated view heavily and new information relatively lightly. The part worth carrying over is the stock. A commercial relationship has a level, built up over years, that gets updated by each interaction.

You cannot see that level. You can see whether the account still bothers to interact, and my argument is that recency of voluntary contact is the cheapest available proxy for it, because an account with a healthy stock finds reasons to get in touch and an account whose stock has run down does not.

Netzer, Lattin and Srinivasan formalised something close to this in Marketing Science in 2008 with a hidden Markov model of customer relationship dynamics, where customers occupy latent relationship states, transitions between states are driven by encounters with the firm, and the state governs buying behaviour. Observed silence is what a transition to a lower state looks like from the outside when you have transaction data and nothing else.

The practical consequence is that a quiet account is not a communication failure to be fixed by more emails. It is evidence about a state, and the response is to find out what changed, which needs a conversation with a person and not another campaign.

Where this stops

Everything above depends on your systems recording exhibitor-initiated events, and most do this badly in one specific direction.

Replies to a rep's personal mailbox, phone calls answered on a mobile, and conversations at another industry event never reach the CRM. The accounts most likely to have those unlogged interactions are your largest, because they have a named account manager who works by relationship. So the quiet-days measure has a systematic bias: it will overstate silence exactly where the relationship is strongest, and understate it for mid-sized accounts handled through the portal.

You can partly correct this by asking each account manager to review their own top twenty before the list goes anywhere. That takes an hour, it removes the false positives that would otherwise destroy trust in the output, and the corrections themselves are useful because they tell you which channels are missing from your logging.

Quiet days also sit in the middle of the lead time ranking that G14 builds, which is worth knowing before you decide how to route the list inside a renewal intelligence process.

The second limit is that silence has more than one cause. An exhibitor who has gone quiet because they signed with a competitor and an exhibitor who has gone quiet because the person who owned your account left the business look identical in the data. The second is often recoverable and the first usually is not, and no measurement distinguishes them. That is what the phone call is for, and it is a reason to make the call early enough that the answer still matters.

Start by pulling every exhibitor-initiated event you can join to an account for the last eighteen months, take the maximum date per account, and plot the distribution of days since. Look at where your ninetieth percentile falls. If it is somewhere sensible, you have a working measure. If it is at four days because your marketing platform is logging opens as responses, you have found the first thing to fix.