Skip to content

Winning back lapsed exhibitors starts with knowing which ones to call

Renewal intelligenceUpdated 2026-08-189 min read

In short

Winning back lapsed exhibitors works when the file is cut before it is worked. Remove closed businesses, accounts absorbed by a group that already exhibits, and anything four or more editions gone. Score what remains on first lifetime value, editions completed, editions since last appearance and the coded reason for leaving, then call the top decile.

The win back report at the end of Q1 says 214 calls made, 9 accounts recovered, 61,000 of space sold. It reads like a modest success and gets a modest amount of attention.

Then somebody joins the call log to the billing history. The 214 accounts that got called averaged 11,000 of lifetime spend across their whole history with the show. Sitting in the same lapsed file, untouched, are 80 accounts averaging over 100,000 each. Nobody phoned them, because they were further down an alphabetical export and because the reps worked the accounts whose names they recognised.

The campaign did not fail. It was never pointed at anything. A lapsed exhibitor file has enormous variance in it and the only question that matters before the first call is which end of that variance you are working.

Most of the file is not callable

Take 1,240 accounts that exhibited in one of the last five editions and are absent from the current one. Three groups come straight out before anyone scores anything.

Businesses that no longer exist, 41 accounts. Companies acquired into a group that already exhibits with you, 88 accounts, where the stand did not disappear so much as get absorbed and calling them is a conversation with an existing customer about a different topic. And accounts that last exhibited four or more editions ago, 306 of them, where the buying team has turned over completely and you are cold calling a company that happens to have a row in your CRM.

That removes 435 of 1,240, so 805 remain. Which is the first useful thing the exercise produces: a win back target set as a percentage of 1,240 was set 35 per cent too high before a phone was picked up, and somebody will be held to it.

Two of those three cuts depend on knowing why the account left, and the code list that makes that computable is its own piece of work that belongs with G31. If your lapse reasons are free text, do the removals by hand for one cycle. Forty minutes with the companies register handles the closures.

What should a lapsed exhibitor score use?

Score the 805 on four inputs, all of which sit in the systems your renewal reporting already runs on.

Lifetime value on the first tenure. Total gross spend across every edition the account held, including sponsorship and services. Rank it into deciles across the 805 and score one to ten.

Editions completed. How long they stayed. An account that held six editions and left has told you the product worked for them once. An account that came once and never returned has told you the opposite.

Editions since last appearance. Recency, scored steeply: one edition ago scores ten, two ago scores six, three ago scores three, four ago scores one.

Coded reason for leaving. Recoverable reasons such as a budget cut score ten. Ambiguous ones such as poor lead volume score five. Structural ones such as leaving the category score zero.

Weights of 0.40 on value, 0.25 on tenure, 0.20 on recency and 0.15 on reason are where I would start. Value carries the most because the whole point of the exercise is to stop spending senior rep time on 11,000 accounts, and because it is the input you measure rather than infer. The reason code carries the least because it is the softest field in the set.

Working two accounts through the score

Two accounts from the 805, both of which a reasonable person would call.

The first spent 148,000 across five editions and last exhibited one edition ago, coded budget cut. Value puts them in decile nine, so nine. Tenure of five editions against a cap of six is 8.3. Recency at one edition is ten. Reason is ten. The score is 0.40 times 9, which is 3.60, plus 0.25 times 8.3, which is 2.08, plus 0.20 times 10, which is 2.00, plus 0.15 times 10, which is 1.50. Total 9.18.

The second spent 21,000 across two editions, last appeared three editions ago, coded poor lead volume. Value decile three scores three. Tenure of two against six is 3.3. Recency at three editions is three. Reason scores five. That gives 1.20 plus 0.83 plus 0.60 plus 0.75, or 3.38.

Nine point two against three point four. Both accounts look identical on an export sorted by company name, and one of them is worth a rep's morning.

The scoring is deliberately crude and every weight in it is arguable. It beats the alternative because the alternative is alphabetical order and whoever the rep happens to remember.

How deep should the win back list go?

Sort the 805 by score and take the top decile, which is 81 accounts.

Those 81 hold 8.4 million of the 24.6 million of lifetime revenue represented in the whole lapsed file, so 34 per cent of the historic value sits in 10 per cent of the accounts. That concentration is the argument for the cut, and it is worth computing on your own file rather than taking mine, because if your top decile only holds 15 per cent of the value then your file is flatter and the case for a short list is weaker.

Eighty-one accounts is roughly a fortnight for one senior person working properly, which means reading the account history, checking whether the contact still works there, finding out what the competing show did last year, and then calling. That is eight or so accounts a day, and it is a different unit of work from a renewal call on a live account, which G22 queues separately. How many slots the wider team has for outreach of any kind is a separate calculation and belongs with G23.

The reason to hand this to one senior person instead of spreading it across the floor is that a win back call is a harder conversation than a renewal call. The account left. Somebody has to open by acknowledging that, and a rep who has been there two years and inherited the account cannot.

The price you come back at is a separate decision

The instinct on a win back call is to lead with a discount, and the published work says that instinct is half right and expensive if you follow it all the way.

Thomas, Blattberg and Fox studied reacquisition in the Journal of Marketing Research in 2004 and did something structurally important: they modelled the decision to come back and the length of the second tenure at the same time, in one framework. Their pricing conclusion follows directly from that. The optimal strategy is a low price to reacquire, then higher prices once the customer has returned.

Modelling reacquisition on its own gives you the opposite advice, because a deep permanent discount maximises the number who come back and quietly maximises the number who come back worthless. Getting nine accounts through the door on a 40 per cent standing discount is a worse outcome than getting six at full rate, and the report that counts recoveries cannot see the difference.

For an exhibition the practical form is an offer that expires with the edition. A better hall position for the return edition, a waived lead retrieval package, a sponsorship credit, or a discount that is explicitly one edition only and written into the contract as such. What that returning account is actually worth across its second tenure, and how much offer you can afford against it, is G30's arithmetic.

Recency does not decay in a straight line

The recency weighting above drops from ten to six to three to one, and the steepness is doing real work. It is a crude version of a question the customer base literature handles properly.

Fader, Hardie and Lee set out the framework in the Journal of Marketing Research in 2005, combining a Pareto/NBD model of transaction timing with a gamma-gamma model of spend per transaction, and drawing iso-value curves that group customers with quite different purchase histories but similar expected future value. The idea underneath is that a gap in purchasing means different things depending on how frequently the customer used to buy. A long gap from a frequent buyer is strong evidence they have gone. The same gap from an occasional buyer is barely evidence at all.

That maps onto exhibitions cleanly in principle and badly in practice, and the reason is worth being blunt about. An annual show gives you one transaction per account per year. Five years of history is five observations, and most lapsed accounts have two or three. The Pareto/NBD machinery is built for customer bases with dozens of transactions per customer and it has almost nothing to work with here.

So take the idea and leave the model. The idea is that recency should be read relative to the account's own pattern. An account that held six consecutive editions and missed one is a different proposition from an account that came in 2021, skipped 2022, came in 2023 and has now missed two. The first has broken a habit. The second never had one. A weighting scheme that scores both as lapsed one or two editions ago is throwing that away, and a single extra column counting consecutive editions held before the gap recovers most of it.

Where this stops

The score ranks accounts by what they were worth and how likely they are to still be reachable. It does not estimate the probability that a call recovers them, and those are different quantities.

You cannot estimate that probability from the file you have, because nobody has run a campaign with outcomes logged against a comparable set of accounts. The first properly targeted cycle is the thing that produces the data, which means the first cycle should log the offer made, the contact reached, the outcome, and the date, against every one of the 81. Skip that and you will be running the same score on judgement three years from now.

The reason code is the softest input and it carries a known bias, because exit reasons given to the person who sold to you are polite and budget is the cheapest answer in every business. It is weighted lowest for that reason and it should probably be weighted lower still if you have never audited it.

The last limit is censoring. Accounts that lapsed before your last CRM migration may have no billing history at all, so their lifetime value scores as zero and they sink to the bottom of a list they might have topped. Check how far back the spend data actually goes before you trust a decile rank built on it.

Export the lapsed file with total lifetime spend and editions held, sort by spend descending, and read the top thirty names to the sales director. You will get an immediate reaction to at least a third of them, and that conversation tells you more about the callable half of your file than a scoring model built in isolation ever will.

Questions people ask about winning back lapsed exhibitors

Which lapsed exhibitors are worth calling?
The top decile of a scored file, and few others. From 1,240 lapsed accounts, remove 41 closures, 88 absorbed into groups that already exhibit and 306 that last appeared four or more editions ago, leaving 805. The top 81 of those hold 8.4 million of 24.6 million in historic lifetime revenue.
How should a lapsed exhibitor win back score be weighted?
Start at 0.40 on first lifetime value, 0.25 on editions completed, 0.20 on editions since last appearance and 0.15 on the coded reason for leaving. Value carries the most because it is measured and not inferred, and because the point of the exercise is to stop spending senior rep time on 11,000 accounts.
Should a win back offer be a discount?
Partly, and only for the return edition. Thomas, Blattberg and Fox modelled reacquisition and second tenure duration together in the Journal of Marketing Research in 2004, and concluded the optimal strategy is a low price to reacquire followed by higher prices once the customer is back. For an exhibition that means an offer written to expire with the edition.

Related reading

All renewals articles